Abstract

Abstract: LanguageBug: a speak-first approach to language learning is a mobile-based software designed to teach Novice-level, Brazilian Portuguese to self-motivated, adult learners. LanguageBug focuses on helping learners build the speaking skill, but also presents metacognitive principles based on growth mindset to address the problem of foreign language anxiety. The main features of LanguageBug are: exercises (structured prompts to engage learners in speaking practice), self-assessment videos (where learners can record themselves speaking Portuguese to assess their formative progress), and principles (which scaffold learners to understand and apply efficient language learning strategies). The learning theories and design principles that support LanguageBug are: Constructivism, Cognitive Load Theory, and Embodied Cognition.

Keywords: language learning, foreign language anxiety, metacognition, growth mindset, CALL, MALL.

Prologue

The outsider’s challenge

From my standpoint in instructional design, it is delightful, though extremely hard to try to understand language learning in depth.

Experienced linguists and Teachers of English to Speakers of Other Languages (TESOL) are certainly more qualified than me to debate questions in language learning/teaching with rigorous scientific accuracy.

In fact, even insiders may find this a hard task, since some fundamental questions remain open-ended in the field of second and foreign language learning and/or acquisition.

For example, in a long-standing debate known as the critical period hypothesis (Lightbown; Spada, 2006, p. 93), linguists have yet to reach consensus if learning languages is biologically linked to age or not.

Similarly, it may be difficult to agree on what makes someone a good language learner, on how foreign accents develop, or on why children may seem to learn languages so fast.

A burning question

In my lifelong involvement with language learning and brief experience as a language teacher, one question has always intrigued me: how might we make language learning more efficient?

Of course, this question is not new.

In fact, in a pursuit of better practices in language teaching, several methods have come and gone over the past hundred years, some in total “philosophical opposition to others” (Brown, 2000, p. 16)

But learning languages is often still perceived as being “harder than other subjects” and even “demotivating” (Ward, 2014), or causing “decreased motivation, frustration and anxiety” (Bernat; Gvozdenko, 2005).

In the realm of educational technology, I found that most language learning software are reproducing typical classroom approaches, where there could be more ambitious innovations (see: Landscape Audit).

From CALL to MALL

What is the current landscape and the history of language learning technology? Can we say that the language learning websites, applications, and services recognize that language learning is complex and nonlinear?

It is relevant to analyze how computer software for language learning have evolved over the decades to identify its gaps.

The history of CALL

Warschauer & Healey (1998) roughly divide the history of Computer-Assisted Language Learning (CALL), into three stages: behavioristic CALL, communicative CALL, and integrative CALL.

1. Behavioristic CALL (the 1960s and 1970s)

In this period, computers were seen as “mechanical tutors which never grew tired” (Warschauer; Healey, 1998, p. 57). They served to prompt some automated, linear, fixed exercises to learners.

Most of these exercises were “repetitive language drills, referred to as drill-and-practice (or, pejoratively, as ‘drill-and-kill’)” (Warschauer; Healey, 1998, p. 57), as illustrated on Image 1.

Image 1 - Drill & Kill

2. Communicative CALL (late 1970s)

This period is marked by the rejection of a pure behaviorism. Rather than just drill exercises, computers would also feature discussion prompts and other forms of conversation starters that students could use to engage in authentic communication.

“Proponents of communicative CALL stressed that computer-based activities should focus more on using forms than on the forms themselves, teach grammar implicitly rather than explicitly, allow and encourage students to generate original utterances rather than just manipulate prefabricated language” (Warschauer & Healey, 1998, p. 57)

Additionally, the software would then begin to experiment with elementary forms of interaction with learners, such as mini-games, multiple-choice tests, audio recording, and others.

3. Integrative CALL (1990s)

It reflects a shift in language teaching theory and practice from a “cognitive view of communicative teaching to a more social or socio-cognitive view” (Warschauer & Healey, 1998, p. 58).

In other words, it puts emphasis on language use in authentic context, rather than on language forms. Task-based, project-based, and content-based exercises begin to be part of the integrative approach.

Additionally, learners begin to make choices in the Integrative CALL period. Computers would give them control over both their learning pace and their individual path within the software.

4. Still Integrative? (2000-now)

The Integrative CALL period is the last period covered by Warschauer & Healey (1998). From their standpoint in 1998, a lot of hope and fascination was being directed to the “multimedia networked computer.”

Almost two decades after the publication of “Computers and language learning: An overview”, claims and projections made by Warschauer & Healey (1998) seem to accurately describe the current context, as in:

“students learn to use a variety of technological tools as an ongoing process of language learning and use, rather than visiting the computer lab on a once a week basis for isolated exercises” (p. 11).

On the other hand, it was not clear to them what to expect from such variety of technological tools. Also, new academic terms were created to describe the current context better.

New terminology

In the acronym CALL, the word “computer” is usually associated with desktop or laptop machines. The acronym does not cover the current range of digital devices that may be used for learning purposes.

The literature variously refers to CALL (Computer Aided Language Learning), CAI (Computer Assisted Instruction), TALL (Technology Assisted Language Learning), distance learning, on-line learning, and more (Bourgerie, 2003, p. 3).

Notably, the acronym MALL (Mobile Assisted Language Learning) has become widely accepted over the past decade, for dealing with the particular affordances and challenges of mobile devices:

Notwithstanding its benefits, MALL also poses related challenges. For instance, inherent in the portability of mobile media are reduced screen sizes, limited audiovisual quality, virtual keyboarding and one-finger data entry, and limited power. (Chinnery, 2006)

Conclusion

By analyzing the history of CALL and the current perspectives of MALL, we may think that the notion of a computer as a mechanical tutor (as seen in the Behaviouristic CALL period) is mainly in the DNA of language learning software.

Most services incorporate a single particular method and attempt to guide learners through the whole language learning experience as if they were enough to all learners (see: Landscape Audit).

As a result, most language learning software does not help learners understand how to engage in effective practice. The “networked computer” now offers a sea of distributed, individualized resources that compete to be the best.

A sea of resources

If your target language is widely spoken, it is likely that you will not have any problems finding learning resources on the Internet. Still, learning this language may often be challenging and even demotivating.

Why? What is missing? How might we be able to fill this gap? To answer these questions, we should first try to understand the broad landscape of resources that are currently available.

Let’s begin by distinguishing these services and materials between two categories: authentic resources and language-based resources.

Authentic resources

Definition

Materials that have a primary purpose other than helping people learn languages or providing any language assistance.

Example

An English learner may watch Citizen Kane or read The New York Times to build up reading and listening skills. However, neither the film nor the newspaper was produced with the primary purpose of instructing language learners.

Availability

The realm of authentic resources for language learners is essentially endless. It ranges from all sorts of user-generated content (posts on a blog, Twitter page, or Facebook profile, comments, …) to corporate generated content (newspapers, media channels, company websites, …).

Language-based resources

Definition

Materials created with the primary purpose of helping people learn languages or providing any language assistance.

Example

A workbook called “Learn English Now! - Beginner” might include several texts in English, but they would all be tailored to suit the expertise level of the learner. Vocabulary lists, explanations, assessment questions, and discussion prompts might follow these texts.

Availability

Language-based resources for language learners are extremely abundant on the Internet. To illustrate that, Table 1 offers a rough division of these resources, services, and products into functional categories.

Classic dictionaries	Merriam-Webster, Cambridge, Dictionary.com
Crowdsourced dictionaries	WordReference, Urban Dictionary
Translators	Google Translate, Babylon
Forums & Social Media	WordReference, Linguaholic
Blogs	FluentU, Espresso English
Video Channels	Rachel’s English, English For You
Alternative methods	Mimic Method, Fluent in 3 months
Synchronous teaching	iTalki, Verbling, Colingo
Asynchronous courses	Rosetta-Stone, Duolingo, Livemocha
Language Exchange	HelloTalk, My Language Exchange
Pronunciation Databases	Forvo, howjsay.com, inogolo

Table 1 - Language-based services divided into functional categories

Due to many overlaps, such categories may look slightly arbitrary. For example, some blogs may also have YouTube channels, just like some products may offer more than one service.

It is important to state that this list is not exhaustive. Besides the few examples within parenthesis, it is likely that many other products could be added to almost all categories.

… so what?

After analyzing such an overwhelming availability of language-related resources, we should ask: how is all this information being used? When does one begin to actually learn a language?

These questions are usually taken for granted and, as a result, language learners may face challenges and/or dilemmas without any assistance. By informing people why they’re doing a certain thing they will be empowered to make decisions on what and how to use the services.

Choosing among resources

Image 2 illustrates the challenge of choosing one path among a sea of resources. As information gets more prominent and distributed, it becomes harder for individuals to make a informed decision.

Image 2 - So many tools out there!

Using the chosen resources

Most of the listed services and resources are filled with information on languages, but do not offer clear instructions on how to build up language skills from that information.

Even the majority of language courses do not deliberately explain how to use the content they provide (see: Landscape Audit).

As a consequence, learners become responsible for structuring their own training or practice time, based on the language learning strategies they know and on the beliefs they hold.

Synthesis

This section has indicated that language learners who are relying on online resources need to constantly make decisions on how to structure and spend their practice time to learn languages.

Even though the amount and diversity of language learning resources on the Internet is vast, users are usually left alone to make these important decisions without any formal assistance.

Therefore, this section highlights a need for providing users with information to make informed decisions on how to use the available resources to learn languages in a more efficient way.

Ill-structured practices

When language learners want to make use of the available information on the Internet, what should they do? In other words, how do people train or practice their foreign/second language skills?

This section shows that structuring a practice/training time can be challenging to most people. Also, the quality of this training can impact the results that language learners will obtain.

The challenge of performance

Two examples (among many others) make it simple to understand the challenge of language learning practice: music and sports.

Music

Imagine an individual who is willing to learn to play a given musical instrument, for example, the acoustic guitar. This person can surely find a lot of information about that instrument on the Internet, such as:

• how to play all notes and chords using the acoustic guitar,
• the history of the acoustic guitar,
• song tabs and chords for acoustic guitars, etc.

However, merely accessing, reading, and understanding all that material will not make this person a good guitar player.

Sports

Similarly, picture a person who wants to become good at a certain sport or type of exercise. Let’s say the target sport is weight-lifting. Uncountable sources of information will provide this person with:

• good techniques and best practices on weight-lifting,
• an array of different exercises to strength each body muscle,
• slow-motion videos of athletes lifting weights, etc.

Still, this person will need to actually go to the gym and lift weights to become stronger or better at weight-lifting.

Deliberate Practice

Scott and Ghinea (2014) argue that engaging in deliberate practice is an important way to foster language expertise. They primarily discuss computer programming, but also refer to many other areas.

According to them, it is crucial that learners be in control of their own practice. The acquisition of expertise happens best when maintaining “an ongoing, reflexive, and self-regulated learning process” (p. 169).

Scenario

Let’s imagine a language learner whose goal is to improve his/her language skills. This person has decided to devote some time to doing that right now. Where should he/she begin?

Selecting the activity

Of course, in the field of language learning there is never only one answer to any question. In fact, there are many activities in which this language learner may opt to engage.

He/she may decide to watch a movie in the target language, for example, knowing that this would be a good way to improve his/her listening skill.

Activities	Skills
Watch a movie	Listening
Sing-along to a song	Speaking
Read a news article	Reading
Write a text/post	Writing
Translate a text	Reading, Writing
Read a book	Reading
Contact another speaker	Speaking
Read a text aloud	Reading, Speaking
Listen to a song	Listening
...	...

Table 2 - Deliberate Practice: activities and their target skills

In Table 2, we can see many other examples of typical, self-regulated language learning activities. This list is, of course, non-exhaustive, as many other options certainly exist.

Structuring the activity

Each option in Table 2 brings up new questions to the language learner. If, for example, he/she decides to sing-along to a song, the immediate question that arises is: “which song?”.

To make this decision, the learner will have to think about songs in that target language. Even if we assume this is an easy task for this particular learner, other questions may keep coming up.

Singing Along

How to choose a song that will make me learn the most?

Am I motivated to sing the songs I know in my target language?

Will I be learning relevant vocabulary from the song I choose?

Is the vocabulary in this song overly easy/complicated for my level?

What should I focus on: pronunciation, accent, vocabulary, or meaning?

Is it better to try to sing the whole song, or go sentence by sentence?

Singing how many times would be enough to learn from the song?

Is this song adequate for my proficiency level? Or is it too fast/slow?

...

Table 3 - Questions when structuring a sing-along to a song practice time

Of course, not all language learners will go through each of these questions. Table 3 only serves to illustrate that there are many decisions that this learner will face, even if unconsciously, within a single activity.

Consequences

It may be challenging for this learner to deal with so many questions and decisions. Such uninformed autonomy may lead him/her to bad decisions, and ultimately, to an ineffective practice time.

Also, rather than focusing exclusively on the language training, this language learner would probably have multiple concerns. After all, his/her tasks during this practice activity include:

1. answer the above questions to structure the practice activity
2. evaluate these answers in a reflective way
3. play/pause the chosen song, possibly many times
4. follow the lyrics and sing along

Focused Practice

From the perspective of the Cognitive Load Theory, human working memory resources are limited. Therefore, instructional designs should attempt to make learners concentrate on the specific learning processes.

Extraneous Cognitive Load

In the present scenario, tasks 1, 2, and 3 are not directly linked to language learning. They are called extraneous cognitive load.

This extraneous cognitive load uses too much of the learner’s working memory resources, instead of productive learning (Sweller, 2010, p. 43).

As a result, not only could those simple tasks become very hard to execute, they would also become much less efficient and less pleasant.

Intrinsic Cognitive Load

On the other hand, task 4 is where the intrinsic cognitive load occurs. In other words, it is the only task that is relevant to the goal of “schema acquisition and automation” (Sweller, 2010, p. 43).

Devoting most of the learner’s attention to task 4 is, therefore, crucial to making this whole practice time an efficient learning experience. As Sweller puts it,

“working memory resources should be devoted to dealing with intrinsic cognitive load rather than extraneous cognitive load because schema acquisition is directed to the interacting elements associated with intrinsic cognitive load.” (Sweller, 2010, p. 43)

Synthesis

This section highlights that language learners may need to reduce the extraneous cognitive load in their language learning practices, while also increasing the intrinsic cognitive load.

Practice environments can be designed explicitly to address these needs. By doing so, when a language learner engages in a practice activity, his/her complete focus will be devoted to language learning, rather than any decision-making or operating other tools.

Target Learners

LanguageBug targets specifically adult language learners.

This population has typically a complex profile when it comes to language learning. Here, adults learners are going to be characterized as self-motivated, but discouraged, distrustful, and maybe too serious.

Such negative framing is of course not what describes all adult language learners. Adults can also learn languages because of their prior success with language learning, or their great abilities in this field.

This section, therefore, serves to illustrate the target learners who would most struggle with language learning, due to their prior unsuccessful experiences, negative beliefs, or discouragement.

Motivation

According to Gardner and Lambert (1959, 1972), there are two main classes of orientations that characterize a language learner’s motivation to learn another language (L2):

• integrative orientation
• instrumental orientation

Noels et al. (2000) explain these classes of orientations:

“First, the integrative orientation refers to a desire to learn the L2 in order to have contact with, and perhaps to identify with, members from the L2 community. This orientation can be contrasted with the instrumental orientation, which refers to a desire to learn the L2 to achieve some practical goal, such as job advancement or course credit.” (Noels et al, 2000, p. 59)

Andragogy

“The five assumptions underlying andragogy describe the adult learner as someone who: 1. has an independent self-concept and who can direct his or her own learning, 2. has accumulated a reservoir of life experiences that is a rich resource for learning, 3. has learning needs closely related to changing social roles. 4. is problem-centered and interested in immediate application of knowledge, and 5. is motivated to learn by internal rather than external factors.” (Merriam, 2001, p. 5)

Knowles himself has stated that andragogy is less a learning theory of adult learning and more of “a model of assumptions about learning or a conceptual framework that serves as a basis for an emergent theory” (1989, p. 112)” (p. 5)

Self-motivation

Adult learners are most likely to be self-motivated, that is, to be sure of the reasons why they want to know a particular language. For example, an employee who has the chance to work in a different country will know why knowing that foreign language is needed.

In another example, someone looking for a job may look better qualified by knowing how to speak a foreign language. In cases like these, one understands the benefits and goals of learning a certain language, therefore, is willing to take action and invest time, energy, and even money into it.

But even those self-motivated learners sometimes do not succeed in learning foreign languages. It may happen for many different reasons, as lack of emotional support from peers and teachers, unrealistically high expectations, and inefficient language learning techniques, etc.

Discouragement

A person that has “failed” to learn a foreign language in the past feels discouraged and frustrated. This frustration may get even more harmful if this person believes that “language learning ability is dependent on some immutable, innate talent” (Mercer, 2012, p. 22), which is a typical scenario. Mercer & Ryan (2010) project that:

“It is possible that FLL [Foreign Language Learning] is a domain in which the fixed mindset may be particularly prevalent, given the widespread belief in the importance of natural talent or aptitude in successful language learning” (p. 444)

It might be the case that most people that have “failed” in language learning were actually making lots of progress, but were not able to notice that. Most language courses are too pragmatic to explain that language learning has many inherent challenges, and it is expected to be difficult to overcome some of these challenges. In other words, they lack metacognition and growth mindset.

Distrustful

Besides adult learners being most likely to suffer from frustration and discouragement from a prior, unsuccessful attempt to learn a foreign language, they have to deal with a general discredit for the simple fact that they are adults, therefore, less capable of learning languages than children.

The Critical Period Hypothesis, on which there is no consensus among linguists, is that “there is a time in human development when the brain is predisposed for success in language learning” (Lightbown; Spada 2006, p. 93). It is estimated that this critical period would usually end around puberty.

Believing in the Critical Period Hypothesis, or in any myth that adults are bad language learners, may impact the learner’s self-beliefs. In fact, “negative or unrealistic beliefs can lead to decreased motivation, frustration and anxiety” (Oh, 1996; Kern, 1995), which will likely affect the quality of learning experiences.

But while “the general consensus is that the younger learner has stronger powers of mimicry and retention, there is no evidence to suggest that adult learners are slower in terms of absorbing new information” (Johnson, 2015).

Too serious

Besides their self-beliefs and mindsets, it might be the case that shyness affects more adults than children when it comes to learning languages. Fear of feeling and looking silly, resistance to trying new approaches, and concerns with making mistakes can explain a harsher, stricter approach.

For example, when it comes to speaking skills, it is hard for most adult learners to experiment with produce sounds that go beyond their own phonetic repertoire. This makes it challenging for them to mimic the accents and the pronunciation in foreign languages.

Fortunately, one of the principles of Andragogy is that adult learners are guided by learning goals and tied to logical principles. By explaining to such adult learners the benefits from “loosen up”, it is expected that their capacity to playfully explore their language abilities will expand.

Content

LanguageBug covers Novice-level, Brazilian Portuguese. It focuses on helping learners build the speaking skill.

Brazilian Portuguese

LanguageBug addresses problems and issues that are genereal in the field of language learning. This thesis project could cover any language, as a consequence. However, I chose Brazilian Portuguese.

First, I am a native Portuguese speaker and a content-wise expert.

Secondly, I live in the United States, so it is very easy to find potential user testers who had never spoken Portuguese. Any person (my peers, for example) who is willing to learn a little Portuguese may benefit from contributing to my design process.

Novice level

As a language teacher, I have dealt many times with Novice speakers than with speakers in any other level. As a language learner, I have reached at least the Intermediate level in 4 different languages. Therefore, I opted to make target Novice learners only, so that I can build on these experiences.

Speaking

The most common approach is to divide language learning into of four main skills: reading, writing, listning and speaking. This project addresses mainly the speaking skill, but also tackles listening and reading.

Novice speakers

According to the American Council on the Teaching of Foreign Languages (ACTFL, 2012), Novice speakers usually face problems regarding intelligibility, functionality, participation in a conversational exchange, speed, vocabulary, syntax, and pronunciation.

Secondary content domains

This project also covers metacognition principles and mindset strategies as vehicles to foster successful language learning.

• Metacognition: usually how polyglots and hyperglots accomplish distinguished results in their language learning experiences.
• Mindsets: emotional strategies to overcome shyness, embarrassment, foreign language anxiety, and perfectionism.

Content items

LanguageBug covers basic phrases, greetings, cognates, main verbal structures, and self-introduction sentences. Part of the vocabulary involved in the process of acquiring such competencies will depend on the learner’s goals and interests.

Context / Settings

These are the ideal conditions for which LanguageBug is designed to work best: private environment, unaccompanied, mostly mobile, distraction-free, optimized time.

Private environment

Learners will have a better experience at a private environment where they can freely speak aloud without any concerns. Houses and private office rooms are examples of ideal study places.

Unaccompained

Studying alone is a great way for language learners to loosen themselves to speak freely with no fear of making mistakes. It also helps them maintain focus and to keep their pace.

Mostly mobile

Learners will be able to use LanguageBug on different platforms, both mobile and computer-based. The goal is to facilitate access to the training practices, independently on which device the learner owns.

Also, mobile devices are key entry points to online services. Today “19% of Americans rely to some degree on a smartphone for accessing online services” (Pew Research Center, 2015).

Distraction-free

It is a common practice to use apps during free time, like while watching TV or having dinner. In fact, 30% of American smartphone owners check their phones even when at a meal (Lookout Mobile Security, n.d.).

However, language learning practices require lots of focus from learners. Therefore, their exclusive attention should be devoted to LanguageBug; otherwise, it will not be possible to reach the full experience.

Optmized time

Language training in LanguageBug consists of short, timed practice activities, which should range from 1 to 5 minutes. Therefore, even short breaks (i.e. at work) can be used for these meaningful, intense practices.

Goals

The primary goal of this project is to help learners that do not know any (or close to any) Portuguese to become novice-level Portuguese speakers. It targets specifically the building of speaking skills.

The following table was built based on guidelines from the American Council on the Teaching of Foreign Languages (ACTFL, 2012). It lists and exemplifies the specific linguistic objectives of LanguageBug.

Speaking (Portuguese)
Learners will be able to manage successfully a number of uncomplicated communicative tasks in straightforward social situations.	Example: Self-introduction, description of objects and people, buying items, greetings, etc.
Learners will be able to handle interactions needed for basic survival in a context where the target language is spoken.	Example: Basic personal information, basic objects, activities, preferences, and immediate needs.
Learners will be able to respond to simple, direct questions or requests for information in the target language.	Example: Giving directions and asking for directions, asking for prices, conversational questions.
Learners will be able to speak intelligible, short (and sometimes incomplete) sentences in the target language.	Example: pronunciation, syntax, and vocabulary understandable by sympathetic interlocutors.

Table 4 - Specific Learning Objectives

Ultimately, LanguageBug users will also understand better how to structure an effective, focused language learning practice.

Therefore, they will be better prepared to build up their language skills using other resources (both language-based and authentic).

Landscape Audit

As described in the first section, the landscape of language learning tools is vast and diverse. This section analyzes only a few language learning mobile applications, since LanguageBug belongs to this category.

This is also a broad category that includes a large group of services. They were chosen based on relevance and popularity, according to the number of downloads on Apple Store.

Disclaimer

These services are long, complex, and may include paid features that I was not able to access. Therefore, these are not exhaustive analyses of their functions, but personal takes on the features I have explored.

Product Analyses

Babbel

“Babbel is an online language learning software and e-learning platform available in various languages since January 2008. Thirteen languages are currently offered: Dutch, Danish, English, French, German, Indonesian, Italian, Norwegian, Polish, Portuguese, Russian, Swedish, Spanish and Turkish. According to babbel.com, it has over 20,000,000 users from more than 190 countries”

Duolingo

“Duolingo is a free language-learning platform that includes a language-learning website and app, as well as a digital language proficiency assessment exam. Duolingo is ad-free and offers all its language courses free of charge. As of February 2016, the language-learning website and app offer 54 different language courses across 23 languages; with 28 additional courses in development. The app is available on iOS, Android and Windows 8 and 10 platforms with over 120 million registered users across the world.”

Memrise

“Memrise is an online learning tool with courses created by its community. Its courses are mainly used to teach languages, but are also used for other academic and nonacademic subjects (such as trivia, video game trivia, and pop cultural). Memrise uses flashcards augmented with mnemonics—partly gathered through crowdsourcing—and the spacing effect to boost the speed and ease of learning.”

Rosetta-Stone

“Rosetta Stone Language Learning is proprietary computer-assisted language learning (CALL) software published by Rosetta Stone Inc. The software uses images, text, and sound to teach words and grammar by spaced repetition, without translation. Rosetta Stone calls its approach Dynamic Immersion (a term which has been trademarked).”

Babbel

About: "Babbel is an online language learning software and e-learning platform available in various languages since January 2008. Thirteen languages are currently offered: Dutch, Danish, English, French, German, Indonesian, Italian, Norwegian, Polish, Portuguese, Russian, Swedish, Spanish and Turkish. According to babbel.com, it has over 20,000,000 users from more than 190 countries" (from: https://en.wikipedia.org/wiki/Babbel)

App Analysis (iOS v. 5.3.1)

Screen-shots of the initial screens of the app

Learning

In spite of Babbel's claim to have 'reinvented language learning', this App is not very innovative. It consists of a series of translation-based exercise prompts, meaning that the bulk of the learner's work is made of tapping on buttons. The animation after each click/tap delays the learning experience and increases the extrinsic cognitive load. Besides that, there is no clear content structure, and sometimes Babbel even prompts users to translate content that has not been introduced yet.

• Speaking: Babbel deliberately asks learners to speak aloud, but only occasionally.
• Content: Words and expressions that do not seem to be structured on any logical category.
• Estimated duration: The first lesson covers a very reduced amount of words and expressions (only 'Oi', 'Tchau', 'Qual é o seu nome?', and 'Tudo bem?', meaning 'Hi', 'Bye', 'What's your name?', and 'How are you?'). Acquiring fluency could take up to years if learners rely only on Babbel.
• Relevance: Words and expressions that are relevant for a beginner Portuguese learner.
• Focus: Actions take too long to complete due to animation delays, which reduce focus.
• Right: Learners can only advance in the lesson after choosing a correct answer, which triggers Babbel to add a green border to the right answer box and speak the content that is inside it.
• Wrong: Babbel does not punish learners for wrong answers, it simply does not advance in the lesson.
• Challenge: With easy and repetitive questions that slowly follow each other, it is not challenging.
• Metacognition: It does not explain its method, or give users choices to shape their learning path.

Self-Efficacy tools

• Beginner?: Users can choose between two levels of proficiency before starting to take courses.
• Counters: It provides a quantified overview of the learner's performance in each lesson.
• Beginner's Course Structure: It consists of 23 lessons that have different titles and purposes.
• Review manager: Babbel keeps track of content that the learner may need to review.

Usability

Audio output

Audio is usually only played after the user chooses the correct answer to the matching exercise.

• Bad recording settings: Sometimes audio was clearly recorded without a pop filter.
• Accents: I was able to identify non-native Portuguese speaker accents a few times.

Speech recognition

Speech recognition is a complex feature that is to produce errors and slow down the pace of the whole learning experience. On the other hand, it can enrich the responsiveness and cleverness of an App, if working correctly and seamlessly. In Babbel's case, only the errors and slowness resulted from it.

• Errors: I could not make the app recognize my speech, with or without earphones with buil-in mic.
• Siri: After a few failed attempts, even Siri popped up, which slowed it all down even more.
• Continue without speech recognition: This feature may be essential, since the speech recognition feature will mostly likely break, making it impossible to advance in the course with it active.

Keyboard input

The exercises in Babbel are only based on matching and tapping, so there is no keyboard input.

Other features

Literal translation: The literal translation of a few expressions is sometimes displayed.

Meaningless pictures: Increase extrinsic cognitive load and do not help learners in any way.

Conclusions

In summary, Babbel feels very slow. It is hard to remain concentrated when the words/sentences repeat to exhaustion and each question is followed by an animated transition. Just like most language learning apps, Babbel trains users on the tapping of buttons that choose appropriate translations for the words in their target language, rather than on concrete linguistic skills. Finally, its speech recognition feature is not reliable and is likely to result in a frustrating experience.

Duolingo

About: "Duolingo is a free language-learning platform that includes a language-learning website and app, as well as a digital language proficiency assessment exam. Duolingo is ad-free and offers all its language courses free of charge. As of February 2016, the language-learning website and app offer 54 different language courses across 23 languages; with 28 additional courses in development. The app is available on iOS, Android and Windows 8 and 10 platforms with over 120 million registered users across the world." (from: https://en.wikipedia.org/wiki/Duolingo)

App Analysis (iOS v. 4.6.0)

Screen-shots of the initial screens of the app

Learning

Duolingo mainly consists of a series of mini-games that introduce single words or sentences as the user progresses. It is highly anchored in the right/wrong binary, that is: users advance within the course as far as their answers are correct. Mostly, all new content is introduced through translations, therefore choosing the right answer means finding the accurate translation. As a result, users will become highly trained at associating words in their target languages with words in their native ones, which is not the same as becomeing fluent speakers in their target languages.

Easy games, with little relevance, mixed with dense, short grammar explanations

• Speaking: Users are rarely prompted to speak words/sentences in the target language.
• Content: The first lesson introduces words such as 'man', 'woman', 'bread', and 'apple'. The course is structured in different topics, which are either vocabulary-related ('animals, 'fooda') or grammar-related ('plurals', 'tu / você'). 'Phrases' is one of these topics, meaning that most times, new content is introduced on a single-word base.
• Estimated duration: In the first lesson, only a few words were introduced: 'the', 'a', 'boy/girl', 'man/woman', 'eat', 'drink', 'water'). Acquiring fluency at this rate would probably take up to years.
• Relevance: Sentences such as 'She eats bread' and 'I'm not a boy' are introduced in the first lesson, even though they are not nearly as relevant as self-introductory sentences, for example, such as 'My name is____' or 'I am from ____'.
• Focus: Having to continuously press the continue button and often type in words can be distracting.
• Right: Correct answers makes some of the items on the screen become green while sound is played.
• Wrong: The progress bar shrinks and more exercises are assigned when users give 'wrong' answers.
• Challenge: The games are initialy very easy and sometimes even the pictures give away the answers.
• Metaphor: Duolingo compares language learning to the life and growth of a living animal.
• Metacognition: Duolingo rarely/never attempts to teach/discuss best learning practices/strategies.
• Translation: Sometimes, translating single words can produce wrong/oversimplified associations. For example: verbs in different English conjugations ('eat/eats') are not always equivalent to their Portuguese counterparts ('como/come').

Self-Efficacy tools

One of Duolingo's main strenghts is the amount of design features created to keep the language learning process happening. A variety of tools help users create and maintain their habits and, therefore improve their self-efficacy skills.

Daily goals, strength bars, structured content and day streaks.

• Pick a goal: The app can adapt to how much time/effort users intend to devote to learn a language.
• Are you a beginner?: Users can be placed in more advanced levels based on their prior knowledge.
• Topics: Users can decide which content they want to learn. App keeps control of options and needs.
• Weak skills: Duolingo takes note of each user's 'weak skills' and prompts them to practice those skills.
• Strength bar: It features in a very clear way which topics are most/least strong in the user's memory.
• (#) days streak: The app counts for how many days users have been accomplishing their daily goals.

Usability

Audio output

Audio is used as exercise prompts (for example, you hear a voice saying 'Hello' and you have to match its Portuguese translation, 'Olá'). Also, when you select an answer, the app speaks the word you chose. That being said, none of the observed exercises were particularly or exclusively designed to build the users' listening or pronunciation skills.

• Non-native accents: Sometimes, words are pronounced an accent that sounded Spanish.
• Multiple voices: Typical male/female voices from different people are played throughout the course.
• Slow mode: Users can choose to hear the audios in a very reduced speed by hitting the turtle icon.

Speech recognition

There was no speech recognition feature in the first lessons of the Portuguese course.

Keyboard input

- Typing takes too much time

• Typing before learning: Duolingo often asks users to type in content that has not been introduced.
• Typo: Words with only a few wrong letters are considered correct, and the typo is identified.
• Accents: Portuguese words typed in without special characters (i.e.: 'é') are considered typos.

Other features

Diversity: Duolingo shows characters from multiple identities, cultures, and ethnicities speaking the user's target language. In this example, it is fascinating that there is a Muslim character represented on Duolingo's Portuguese course, despite the fact that the Muslim population in Brazil is very reduced. By doing that, Duolingo not only shows respect for minorities, but also contributes with the fight against the stereotypification of the Brazilian population.

Conclusions

Duolingo is well-known, cleverly designed, and its game-based exercises are responsive and aethetically pleasant. On the other hand, Duolingo does not offer any significantly new approach from the perspective of language learning. Its whole mechanism is mostly useful to help users memorize small pieces of content by matching words/sentences with their counterparts in the target language. Its most strong design features are the self-efficacy tools, which help users keep track of their progress, create a routine of practice and remain motivated to learn their target languages.

Memrise

About: "Memrise is an online learning tool with courses created by its community. Its courses are mainly used to teach languages, but are also used for other academic and nonacademic subjects (such as trivia, video game trivia, and pop cultural). Memrise uses flashcards augmented with mnemonics—partly gathered through crowdsourcing—and the spacing effect to boost the speed and ease of learning." (from: https://en.wikipedia.org/wiki/Memrise)

App Analysis (iOS v. 2.1.11)

Screen-shots of the initial screens of the app

Learning

Memrise is focused on helping users memorize content, mainly single words. There seems to be a complete absence of a higher-level plan to help users build their foreign language skills. Mainly, the exercises consist of behaviouristic tests of memorized content, without almost any regard to pronunciation, listening, speaking. Language learning through Memrise is veiewd as an empty pursuit of a 'right' answer, which may blind the learners to the benefits of trying, practicing, and making mistakes.

• Speaking: The app does NOT prompt the user to produce any form of speech in the target language.
• Content: The number of words and the semantic structure of the Portuguese course are unkown.
• Estimated duration: Daily goal is to learn 5 new words, which reflects a complete lack of ambition.
• Relevance: 'I would like some cofee, please' is taught in the first lesson, but 'My name is ___' is not.
• Focus: Pressing 'Continue' repeatedly takes some of the learner's attention from learning.
• Right: Correct answers are rewarded with points, animations, and progress within the lesson.
• Wrong: mistakes are marked red and need to be immediately corrected to continue the course.
• Challenge: Single-word, multiple-choice, translation questions are by essence not very challenging.
• Metaphor: The growing plant is a clear and visual appealing metaphor for 'growing memories'.
• Metacognition: 'Words Learned' (short-term memory) and 'Words Mastered' (long-term memory)

Self-Efficacy tools

The goal of each lesson, of each exercise, and the overall long-term goal of the course are not clear to the user. As a consequence, it is hard for users to evaluate their learning processes and remain focused on their goals. What the the app does offer is a few important tools to help maintain the schedule of exercises.

• Word listing: Displays a list of words that the user has been introduced to, which quantifies the learning process and helps the user keep track of what he/she should know.
• Meaningless progress bar and meaningless points: The points earned by the user and the progress of each lesson or exercise do not tell anything.
• Adaptive reminders with daily goals: App reminds the user to perform the daily activities.
• Low long-term goal: Learning only a few words a day means that becoming an intermediate Portuguese speaker may take users many years.

Usability

Audio output

The app uses audio as one of the main forms of exercise prompts. For example, the user listens to the audio and finds the appropriate translation among several options. Most of these prompts, at least in the beginning, consist of single words, or very short sentences. Also, the introduction of a new word consists of displaying the written side by side with its translation and a button to hear it spoken

Screen-shots showing how the app uses the audio output

• Bad sound quality: It sounds like not all audio has been recorded in a studio. It is possible to hear some undesired echo a few times.
• Non-native accents: Apparently, not all the people who recorded the audios are native Portuguese speakers. I could sense a Spanish accent a few times.
• Multiple voices: More than one person has recorded the audios in Portuguese, so the learner may get used to different pronunciation styles.

Speech recognition

There is no speech recognition function in this app. Or, at least, I have not encountered an active speech recognition function in any course or exercise level I have reached.

Keyboard input

The app asks or requires the user to input information through a keyboard several times, as a way to assess and correct the user's answers to exercise prompts.

The availability of letters is limited to the ones used in the correct answer, plus a few 'wrong' options

• Reduced letters: Memrise does not use the regular iOS keyboard when the user has to type in information. Instead, it provides a simpler keyboard layout, with reduced options. There is one upper-case and one lower-case key of each letter, even though the answer is not case sensitive.
• 'Skip' option: The user may choose to proceed with the exercises without typying the answer.
• Slows down the learning process: This approach reduces the speed of the interaction between the user and the app, that is, even when the user immediately knows the right answer to a certain question, typying it in will take him/her at least a few seconds.

Other features

Crowdsourced hints: Other users may submit images to help learners understand the meaning of words. Although this is not crucial to the learning process, it might foster in its users the feeling of belonging to a community.

Conclusions

Due to its complete lack of skill-based exercises, Memrise may not be considered a language learning application. Rather than that, it is a tool for memorization that offers some language-specific collections. Memrise may be used by language learners who want to increase their vocabulary, however, memorizing individual words without context is likely to be harder than learning multiple-word sentences within a context or narrative.

Rosetta-Stone

About: "Rosetta Stone Language Learning is proprietary computer-assisted language learning (CALL) software published by Rosetta Stone Inc. The software uses images, text, and sound to teach words and grammar by spaced repetition, without translation. Rosetta Stone calls its approach Dynamic Immersion (a term which has been trademarked)." (from: https://en.wikipedia.org/wiki/Rosetta_Stone_%28software%29)

App Analysis (iOS v. 2.6.1)

Screen-shots of the initial screens of the app

Learning

Rosetta Stone is one of the only language learning services that does not rely on translation. Rather than that, it will always only speak the learner's target language in its exercise prompts. To respond to those exercises, learners may have to either match the audio clue to a picture, or produce a sentence of their own. Learning occurs as the complexity of these exercises increase, by both adding more words or removing the written prompts.

Exercises are of either matching or pronunciation type.

• Speaking: Learners are prompted to speak at different times during the lessons.
• Content: The first lesson covers a surprisingly huge amount of words and expressions ('reading', 'eating', 'drinking', 'running', 'he', 'she', 'they', 'cooking', 'swimming', 'writing', -ing form, 'man', 'woman', 'boy', 'girl', plural).
• Estimated duration: The beginner level should take approximately 50 hours to be completed.
• Relevance: The first lesson does not include the most basic self-introductory sentences.
• Focus: Reduced extrinsic cognitive load, with nothing on the screen, besides the exercises.
• Right: Right answers are marked with a green 'checked' sign, while a happy sound is played.
• Wrong: Wrong answers are marked with a red 'x', while a brief, 'sad' sound is played.
• Challenge: Exercises get harder as the number of words increase and written clues are reduced.
• Metacognition: It does not explain its method, or give users choices to shape their learning path.

Self-Efficacy tools

Clear, linear course progression, with lesson reports

• Classical structure: Units are divided into lessons: core, vocabulary, pronunciation, and writing.
• Navigation within lesson: App lists exercises within each lesson in either green or red squares.
• Lesson reports: After each practice, learners are 'graded' on their performance.

Usability

Audio output

The method created and used by Rosetta Stone relies completely on audio, that is, the App speaks in the target language and we are required to perform actions, which are based on either matching or speaking.

• Slow: Sentences are spoken so slowly that it does not sound like natural speech.
• Accents: I was able to identify a few non-native accents in some of the audio tracks.
• From text + audio to audio only: As the lesson progresses, exercises get a little harder when written words are not shown and the user is forced to match the right picture based only on the audio clue.

Speech recognition

The speech recognition feature in Rosetta Stone is very efficient and conveys a feeling of design cleverness and responsiveness. Since this feature was designed to perceive the speech produced by non-native speakers of Portuguese, even slightly wrong utterances tend to be validated.

• Not completely accurate: I tried mumbling some words indistinctly, and my answers were occasionally marked as correct
• Fast: Enabling and processing the microphone input occur in a surprisingly fast way.
• Pronunciation: Learners hear and practice each syllable of a few of the words learned in each unit.

Keyboard input

There was not a keyboard input in the exercises I explored. However, due to device limitations, I was not able to access Rosetta Stone's writing section, where the keyboard input would probably exist.

Other features

Pictures and meaning: There are a few times in which it may be hard to distinguish the objects or people in a picture. For example, when only the subject of the sentence changes ('She is swimming' vs. 'He is swimming' vs. 'They are swimming'), it may take a few extra seconds to identify the correct picture based only on the quantity, gender, or age of people on it.

Conclusions

Rosetta Stone is astonishingly well designed: it is stable, responsive, direct, and appealing. The content is also incredibly well structured and the complexity/difficulty increases as each lesson advances. Even the speech regonition feature, which is usually slow and unstable in other apps, is seamlessly integrated to the learning experience of Rosetta Stone.

In fact, it may be demotivating to compete with a multi-million dollar company that has been producing language learning softwares since 1992. But, on the other hand, the method of Rosetta Stone has not evolved much over the decades, and may even look outdated, from the uncountable, cheesy ictures to the recurrent game of “tap and match”.

Narrative

There is a very rare species of bugs: the LanguageBug!

The most particular fact about LanguageBugs is that the smell of Foreign Language Anxiety attracts them.

When a LanguageBug bites a person, within only a few hours this person will have an infection.

However, it is not a bad infection: it is an infection that develops superpowers.

Foreign language learning superpowers!

Storyboard

Annotations

[1] As explained in the Target Learners section, according to the conceptual framework of Andragogy, adult learners are identified as self-motivated and self-directed (Knowles, 1989; Merriam, 2001).

[2] [3] Context: 30% of American smartphone owners check their phones even when at a meal (Lookout Mobile Security, n.d.), but LanguageBug is designed to be challenging and to require complete concentration.

As described in the Design Rationale section:

[4] LanguageBug focuses on helping language learners build up their speaking skills. Exercises are intensive speaking practices that prompt learners to produce utterances in the target language.

[5] Learners can access the rationale behind each Principle and each Exercise within LanguageBug. This metacognitive information helps them reflect on their approach to language learning.

[6] The self-assessment video feature is designed to help learners notice their progress. As their speaking skills increase after each practice, they can see their evolution video after video.

As described in the Landscape Audit section:

[7] Speaking: most language learning applications rarely prompt learners to speak. Speech prompts are also limited to simple sentences or even single words, resulting in reduced progress in speaking.

[8] Rather than prompting users to speak, these language learning apps are mostly based on matching mini-games. Learners merely tap on words to find appropriate translations for (usually irrelevant) sentences.

[9] Right/Wrong: after each wrong answer, learners lose a few seconds correcting and re-submitting their answers. Their working memories stop dealing with language learning and begin to operate the app.

Expected results

[10] Learners using LanguageBug speak several times the same utterances. After a few practices, they will be able to repeat those utterances by heart and assertively, even in a social context.

[11] Learners using other apps build up skills that mostly useful within those apps. In a social context, it will be hard for them to express what they have learned in a timely way to engage in a conversation.

[12] Self-assessment videos make learners aware of their progress. As a result, they will become more motivated (Halvorson, 2013, p. 23) and are likely to continue learning their target languages on LanguageBug.

Exercises

In a nutshell

Exercises are focused practices that prompt and scaffold learners to produce communicative utterances in their target language.

Within the narrative

Infected learners also develop the capacity to talk to other language bugs. Each bug is interested in one particular topic or action, such as knowing about the family, the house, or the routine of learners.

Association with existing resources

In the past, language schools typically used printed materials (Liu, 2009, p. 521) and “audio-cassettes or audio CDs of pre-recorded listening materials” (Kukulska-Hulme; Shield, 2007, p. 2) to engage learners in speaking practice.

LanguageBug exercises are essentially adaptations of these materials, with a few significant improvements:

• The pace of activities is very fast, to make it very challenging,
• Exercises prompt learners with both visual/written and audio clues, and
• Principles will scaffold learners before practicing through activities.

Functions

• Engage learners in speaking practice from the beginning,
• Introduce learners to sentence structure without using grammar,
• Help learners build relevant and personal vocabulary, and
• Explain to learners why it is relevant to learn the contents in each exercise.

Example: Self-introduction

What is it?

“Self-introduction” is the most important exercise within LanguageBug. It introduces learners to multiple sentences that are extremely relevant to any language speaker, such as “My name is ___*” and “*I live in __”.

By completing sentences with personal information, such as the color of their eyes (“My eyes are ____”), learners will also acquire vocabulary that is relevant to their individual background and experiences.

Why?

• Essential sentences for communication and survival,
• Introduces learner to relevant, conversational vocabulary,
• Help learners become comfortable presenting themselves, and
• Implicitly teaches learners different sentence structures.

Script

• Show self-introduction sentence in Portuguese with a blank space.
• Audio: play audio of this incomplete sentence in Portuguese.
• Show the translation of that sentence in English.
• Audio: play audio of this incomplete sentence in Portuguese.
• Audio: play audio of the whole sentence (with sample answer).
• (If new vocabulary) Introduce learner to possible responses.
• Ask learner to repeat after the audio, with his/her information.
• Play complete audio again, and ask learners to repeat.
• Repeat all steps with the following sentence.
• Show and play the audio of the two sentences in Portuguese.
• Ask learner to repeat those two sentences with his/her info.
• Repeat procedure, increasing the number of sentences.

Scalability

Within the scope of this thesis project, LanguageBug only covers basic Novice-level content, but there are certainly more words and phrases that could be integrated to the exercises. In addition to that, it is always possible to keep improving the selection of words and expressions based on feedback from learners and user observation.

There could also be many other ways to foster speaking practice, besides the ones that LanguageBug features at this moment. Getting to know other approaches to language learning and teaching may provide insights on the creation of new exercises. The architecture of LanguageBug would make it easy to integrate them.

Finally, bugs could be further developed as characters with particular interests, so that learners could have different conversations with different bugs based on their interests. For example, a learner wishing to speak Portuguese in professional settings could learn relevant content by interacting with a “Business Bug” character.

Current exercises

• Adjectives
• Building-Blocs
• Family
• Greetings
• House
• How often?
• Numbers
• Shortcuts
• Where?
• With whom?
• Yourself

Related FAQ

• No corrections!?
• What about accuracy?

Theoretical foundation

Constructivism

The constructivist notion that knowledge is constructed, not transmitted (Smith; Ragan, 2005) supports the design of Exercises. Learners interact with the same Exercises several times and keep updating their perceptions and mental models.

This is part of what Piaget describes as disequilibrium and accommodation (Martinez, 2010). Rather than being a passive receiver, learners have an “active position of trying to make sense of the world” (Martinez, 2010, p. 198).

Exercises help learners “regulate the sequence of relevant actions” (Reiser; Tabak, 2014, p. 47). The goal is to reduce guidance gradually, so that “learners appropriate this guidance and begin to regulate their own actions” (Reiser; Tabak, 2014).

Embodied Cognition

LanguageBug assumes the embodied perspective on learning and instruction:

“Simply stated, the embodied perspective takes seriously that our physical bodies actually play a very important and, until quite recently, overlooked role in how we think and act in the world” (Lee, 2014, p. 6)

The body plays a crucial role in the process of trying to speak words/sentences in a foreign language. It does so by:

• hearing and processing unfamiliar, meaningless, speech sounds,
• attempting to understand, memorize, and mimic those sounds, while
• making adaptions or even playing with their vocal apparatus.

The goal of Exercises is to provide learners with a safe environment to embody their cognitive process.

The fast-paced sequence of words does not allow learners to “think for too long,” in the traditional sense of the word. On the contrary, it understands that “manipulating the body manipulates how we think” (Lee, 2014, p. 10), therefore, it foster actions.

Cognitive Load

According to the Cognitive Load Theory (Sweller, 2010), working memory has limited processing capacity. Therefore, instruction should only present information that is relevant to the goal of “schema acquisition and automation” (Sweller, 2010, p. 43).

Based on those principles, Exercises only display what is relevant to the goal of language learning. This means that words, sentences, instructions and translations are present in these practices, but interactive buttons, images, and keyboards are not.

Full list of Exercises

See: Appendix 1 - Exercises

Principles

In a nutshell

Principles are mini-lessons that walk learners through guidelines, useful strategies, best practices, or hints that could increase their efficacy in learning languages.

Within the narrative

Language-Bug infection causes learners to develop “language learning superpowers”, which they cannot manage well at first. Principles train users to use these superpowers to learn languages.

Association with existing resources

There are a few TED Talks about language learning. For example:

• Benny Lewis: Rapid Language Hacking
• Young & Jaiswal: One Simple Method to Learn Any Language
• Benny Lewis: Hacking language learning
• Sid Efromovich: 5 techniques to speak any language
• Chris Lonsdale: *How to learn any language in six months

In each of these talks, the lecturers do not teach any actual content in a particular language (i.e.: vocabulary in Portuguese), but present their strategies to help language learners become better language learners.

Similarly, Principles within the LanguageBug app are sections that walk learners through practical strategies to learn languages without necessarily using the Portuguese language whatsoever.

Functions

• Scaffold learners to use the LanguageBug appropriately.
• Make language learners aware of their learning processes.
• Use informal language engage learners in metacognitive thinking.
• Question widely accepted assumptions in language learning.

Example: Speak Fast

What is it?

“Speak Fast” is one of the core principles of LanguageBug. This means that users will need to follow this guideline (that is, to actually produce speech in a fast-paced way) when performing the exercise practices.

Why?

Speak Fast may be an efficient strategy for a language learning practice for several reasons, such as:

• It is harder to speak faster, so learners need to fully focus.
• Real life or native speech is fast and compacted.
• Less time doing each exercise means doing more exercises.

Script

Users will be introduced to the principle of Speak Fast through timed instructions combined with action prompts, following this script:

1. Show user a simple sentence in English for a few seconds.
2. Ask the user to read this sentence aloud.
3. (User reads sentence aloud at a natural pace)
4. Fast transition: ask the user to perform 3 again, but faster.
5. Fast transition: reinforce that reading should be fast.
6. Ask the user to keep reading sentences at the same fast pace.
7. Show other sentences in English.
8. Instruction: “Now, repeat after the audio.”
9. Show/play a few more sentences in English.
10. Transition to words and sentences in Portuguese.
11. Instruction: “Keep that sweet spot!”

Scalability

Creating an exhaustive list of good language learning principles is both impossible and undesired for many reasons, such as:

• there are unlimited learning strategies that could be effective,
• different strategies may work better/worse with different learners,
• new research and publications may challenge current assumptions,

As I develop my general understanding of how people learn languages, some of the present principles may become outdated or invalid. In other words, it is very likely that principles will be added, removed, and expanded over time.

Current principles

Currently, there is not a scientific rationale with evidences to validate each of the following principles. Strategies that I have learned as a language teacher are the current LanguageBug principles, as it follows:

• Speak Out Loud
• Speak Fast
• Avoid writing
• Keep Focused
• You will not remember

Related FAQ

• Is it a method?
• No corrections!?

Theoretical foundation

Growth Mindset

“A language learning mindset reflects the extent to which a person believes that language learning ability is dependent on some immutable, innate talent or is the result of controllable factors such as effort and conscious hard work.” (Mercer, 2012, p. 22)

There is little research integrating Growth Mindset with language learning (Noels; Lou, 2015). Despite that, it is clear that there are many benefits in having a Growth Mindset, that is, believing that anyone can develop skills and abilities over time.

According to Halvorson (2010), “psychologists refer to the desire to get better - to develop or enhance your skills and abilities - as a mastery goal.” (p. 61). When learners have a mastery goal, they take action when facing a challenge (p. 62).

One of the primary goals of Principles is help learners frame their practices within Growth Mindset or mastery goal. Noels & Lou (2015) show that fostering such attitude may be quite simple using electronically delivered messages (p. 49).

Metacognition

“At the simplest level, metacognition is thinking about the contents and processes of one’s mind” (Winne; Azevedo, 2014, p. 63).

Mercer & Ryan (2012) advocate for a Growth Mindset in language teaching and learning. The authors also state that

“although a growth mindset can encourage a learner to work consciously and actively towards improving their own abilities, this may only be effective if the individual also feels that they are equipped with the skills and know-how to do so. Thus, teaching practices, instructional techniques, and learners’ metacognitive strategy knowledge could be important dimensions that influence the actual effectiveness of a particular mindset” (p. 442).

Nesbitt (2013) also state that language learners should have “access to activities that help increase their strategic competence”. According to this author, this sort of metacognitive knowledge is both a need and a request of CALL users.

In a small-scale study, Bozorgian (2014) has found that “less-skilled learners benefit more from metacognitive instruction to develop listening comprehension” (p. 3) than more-skilled learners. The target audience of LanguageBug is Novice speakers, so that addressing metacognitive strategies within the app is even more appropriated.

Bransford, Brown & Cocking highlight that metacognitive knowledge is one of the distinctive characteristics of experts (Bransford; Brown; Cocking, 2000, p. 47). Therefore, metacognitive Principles may also contribute to learner’s long-term expertise building.

Self-assessment Video

In a nutshell

At the end of each practice, learners can record their own speech in a self-assessment video, so they can witness their ongoing progress.

Association with existing resources

In the professional and academic domains, foreign language speakers are often required to take oral proficiency tests. These tests may be administered in-person or through video conferencing tools.

For example, the New York University (NYU) may require its prospective international students to take language placement tests, offered by the American Language Institute (NYU ALI, n.d.).

Similarly, the American Council on the Teaching of Foreign Languages (ACTFL) has been offering a popular test called the ACTFL Oral Proficiency Interview (ACTFL OPI) for over 20 years (ACTFL, n.d.).

Certified test takers analyze and rate the performance of test takers in both examples. Within LanguageBug learners are their own test takers: not trained to look for accuracy, but certainly able to identify their own progress.

Functions

The main function of the Self-assessment video tool is to allow learners to perform their own formative assessment. A few other features are also highly relevant:

• Register speech samples continuously.
• Compare current and past speech samples to assess progress.
• Help learners become comfortable hearing their FL speech.
• Allow learners to share their speech samples with their contacts.

Example: User Test

Sample

During the user tests, I simulated the self-assessment video feature using the Wizard of OZ method (see: User Tests, Results). As a result, there are videos of both users speaking Portuguese after each exercise practice.

Nicole’s “Yourself” exercise self-assessment video can be accessed at the following URL: https://www.youtube.com/watch?v=BNAl6thkpkA.

Regular assessment: Standard framework

If a group of proficiency test takers analyzes this video, they will certainly notice that Nicole’s performance is marked by evident traces of beginner speech, such as strong accent, slow pace, unusual rhythm, etc.

As a consequence, Nicole would be labeled a “Novice” speaker, and her performance would probably receive a low grade. In this case, Nicole’s speech was compared to a certain (so-called) standard speech.

Self-assessment: Progress framework

When Nicole watched the same video, she certainly noticed her speech differed from the Native speech introduced within LanguageBug. It is likely that she could not identify these differences with precision.

But she was also able to notice all the learning she had accomplished in only a few minutes: from not knowing how to say any sentence in Portuguese to knowing a few sentences, even if not perfectly spoken.

In this case, Nicole is comparing her current self to herself before the practice. Progress is the ultimate goal. If Nicole did another exercise practice, it is likely that her performance would be even better.

Scalability

User-generated videos bring several new affordances to the design of LanguageBug. As a result, there could be several enhancements to the self-assessment video feature, such as the following:

• Integrate multiple videos into a single “timeline of progress” video,
• Allow users to follow the videos on their network of learners, and
• Measure/identify the time and content of each video to generate data.

None of these features has been tested our outlined at the moment.

Current Options

• Record: records a video after each practice.
• Watch: plays any video recorded by the learner.
• Share: outputs the URL of the video on YouTube.

Related FAQ

• No corrections!?
• What about accuracy?

Theoretical foundation

Informal learning

The LIFE Diversity Consensus Panel’s “Learning in and out of school in diverse environments” report state that

“learning takes place not only in school but also in the multiple contexts and valued practices of everyday lives across the life span” (Banks et al, 2007, p. 5).

When it comes specifically to languages, the same report suggests that

“students learn more when they are allowed and encouraged to use the variety of language resources available to them.” (p. 22).

Initially, LanguageBug learners have very limited language resources. Despite that, they should be provided with a chance to experiment with their new language and to observe the results. Self-assessment videos is the safe space to do that.

Self-directed Learning

In a self-directed learning experience, “students fully determine their own schedule and pace” (Oberg & Daniels, 2013, p. 179). The purely self-paced and independent studies are naturally challenging and may involve burdens for the learner.

As a result, motivation is one of the keywords in self-regulated learning. Research shows that conducting independent studies may produce high levels of cognitive engagement, meaningful connections and increased motivation (Kinzie, 1990).

Halvorson (2010) states that

people are motivated to do anything as a function of (1) how likely they are to be successful and (2) how much they think they will benefit from it. And of course, the more motivated you are, the more likely you are to reach your goal. (p. 23)

Self-assessment videos show learners that their investments of time and effort result in concrete improvement. As a result, these learners are likely to remain self-motivated.

Is it a method?

No. LanguageBug is not a language learning method, but an approach to language learning. Language learning methods are rigid, fixed instructions to learn languages. An approach, on the other hand, is a flexible and adaptable set of practices and beliefs.

The problem with methods

The word “method” usually comes to mind when discussing how to make language learning. Yet, language learning and teaching methods can be too strict and purist.

Brown (2002, p. 10) lists several reasons to give up with such “obsession” with methods. This list includes the following items:

• It is hard to empirically test, and verify language pedagogy,
• Methods are usually too prescriptive and overgeneralized,
• Methods become indistinguishable as learners reach advanced levels.

A unified approach has been substituting individual methods. This shift begins with the Designer Methods and results in the Post-method Era.

Designer methods

In the 1970s, “in a burst of innovation” (Brown, 2000, p. 10), several language teaching learning methods were created. Nunan (1989) calls these the designer methods.

The designer methods consisted of Community Language Learning, the Silent Way, Suggestopedia, Total Physical Response, and a few others.

EXAMPLE: Total Physical Response

To understand what these methods meant and still mean, and how to approach them, let’s analyze the Total Physical Response method, also widely known as the TPR method.

Functioning

TPR is one particularly unconventional example of a designer method, created by Dr. Asher (1969), psychologist at San José State University.

Archer observed that children “in learning their first language appear[ed] to do a lot of listening before they speak” and that “their listening is [was] accompanied by physical responses” (Brown, 2000, p. 30).

Based on these observations, Archer proposed the TPR method. Its functioning relies on very specific and limited roles in the classroom:

• teachers give commands in the target language (“Sit down!”),
• students react to these commands physically (they sit down).

No other actions or activities are part of the TPR method, which means that all language learning should occur exclusively through physical reactions to voice commands.

Pros

A great deal of physical engagement makes TPR a really fun method for some students. Also, TPR can be “especially effective in the beginning levels” (Brown, 2000, p. 30).

In the field of embodied cognition (Lee, 2014), TPR might offer an interesting case study. For example, it would be fruitful to assess how much these body movements might contribute to memory retention.

Cons

Would it be possible to teach an abstract word, such as “socialism”, through TPR? What about idioms, such as “Bite off more than one can chew” or “Elephant in the room”? Probably not.

TPR is notably limited regarding what can be taught through it (Widodo, 2005, p. 240). There is obviously not one body movement associated with every word or phrase in a language.

Balance

Today, it has become rare to find an exclusive TPR classroom, where speaking, writing, and reading are not allowed. Still, some strategies and activities taken from the TPR method remain popular among teachers.

The post-method Era

Designer methods such as TPR should be applauded “for their innovative flair” and have indeed left behind their legacy of findings, achievements, and teaching strategies. However, we now acknowledge that they were “not the godsend that their inventors and marketers hoped they would be” (Brown, 2000, p. 25) in light of their limitations.

Brown (2002) coined the expression Post-Method Era (p. 9) to refer to the moment in which practices that come from different, fundamentalist methods can be integrated into a single approach.

Approaches

While in the 1970s language teaching consisted of many competing methods in Brown’s view (2000, p. 13), it is “now more aptly characterized by a relatively unified, comprehensive ‘approach’”.

Galante (2014) defines this as a “shift from using methods in the purist sense to recognizing that the nature of language learning is complex and nonlinear” (p. 58).

An approach, therefore, is a consistent body of knowledge (strategies, best practices, …) that supports a particular path to language teaching and learning. It does so without being too restrictive or normative.

Image 3 summarizes this whole trajectory from the designer methods to the post-method era. It ends with the acknowledgment that “there is no single right way to learn languages”.

Image 3 - Methods & The Post-Method Era

No corrections!?

LanguageBug does not assess or correct what learners are saying. In fact, it does not have any audio input during exercises, which means that it does not even “hear” its learners at all. Why?

Foreign Language Anxiety

Research has found links between perfectionism and high foreign language anxiety (Gregersen; Horwitz, 2002). As a result, learners should not begin to worry about having a perfect pronunciation.

LanguageBug understands that making mistakes is an essential part of learning any language. Learners deserve to know that from the beginning of their Portuguese learning journey.

Right/Wrong

When it comes to speaking only, it may be hard to define what is right and what is wrong. Variations in speech are common even among native speakers of the same language, as a result of:

• the same language in different countries,
• regional differences and dialects,
• accents in different social groups,
• formal/informal speech practices, etc.

Samuel & Larraza state (2015) that accepting “wrong pronunciation” is not an error. It may actually be a useful adaptation to an environment of linguistic variations (p. 51).

Accents

Accents can be a distinctive feature to express and maintain cultural identity. For that reason, not all foreign language learners seek accent reduction (Hartshorn, 2013).

Some learners prefer to keep their foreign language accent. Others may struggle to reduce it. In any case, it may be an act of prejudice when a speaker rejects or dislikes a “different” way of speaking.

Corrective Feedack

Correction is not common-sense among researchers in the field of language learning and teaching. In fact, Corrective Feedback (CF) is the center of several debates. Sung & Tsai (2014, p. 38) have listed some of the fundamental questions around CF:

• does CF assist in language acquisition?
• what types of CF are the most efficient?
• what factors can influence the effectiveness of CF?

Language learners may also prefer “clarifications and elicitations” over corrections (Sung; Tsai, 2014, p. 39). Such types of feedback help them “find correct answers themselves” (p. 39).

Conclusion

Most language learning services rely on the right/wrong binary to function (see: Landscape Audit). Mistakes result in a big, red “X” sign that prevents learners to press “continue”. Sometimes, learners may be even “punished” with a reduction in the progress bar.

LanguageBug adopts a different strategy: it praises effort and welcomes mistakes. By doing so, it fosters risk-taking and reduces the chances of increasing anxiety. In summary, it understands that mistakes are part of the process of getting things right.

What about accuracy?

If LanguageBug never corrects, how can learners know if they are speaking it right? How could learners improve if they are the only ones assessing themselves? The answer is that accuracy is not a priority within LanguageBug for several reasons.

Corrections

First, Corrective Feedback (CF) is not a common-sense in the field of language learning (Sung; Tsai, 2014). CF may even cause learners some foreign language anxiety (Gregersen; Horwitz, 2002)(see: No Corrections?!).

Variations

Also, it is hard to determine what is accurate in speech. One single language may have several spoken variations, all correct. Creating any hierarchy of desirability among variations would be problematic.

Repetition

LanguageBug learners can perceive and correct their own mistakes, at their own time. Learners become able to produce more comprehensible and accurate utterances as they keep practicing. This self-assessment mechanism relies on the Constructivist concept of Accommodation (Martinez, 2010) (see: Is it behaviorist?).

Expected level: beginner

LanguageBug aims to help non-speakers of Portuguese become Novice speakers. We should expect Novice speakers to face difficulties with intelligibility and pronunciation (ACTFL, 2012). Addressing those issues should be a priority only during the Intermediate and Advanced stages.

Conclusion

Accuracy is not a core goal for Beginner and Novice learners. Also, issues are surrounding corrective approaches, including the likelihood that it will cause some anxiety (Gregersen; Horwitz, 2002). For all these reasons, LanguageBug does not interpret the accuracy of speech as one of its priority goals.

Formative Evaluation

It is a recommended practice to conduct a formative evaluation during all stages of instructional design, according to the ADDIE model framework (Smith; Ragan, 2005). In light of that, I have created a customized, open-access, website that stores and displays all information related to this thesis project.

Such website became at the same time my central workplace and a channel to showcase my progress. It has offered my community of collaborators (my peers, instructors, and committee members) an ongoing and unlimited opportunity to access and review my work during the whole design process.

Image 4 - Visual representation of the ADDIE Model

In this section, I describe the main functions, features, and mechanisms of this website. In addition to that, I explain and illustrate how creating and using this website has impacted my design decisions at different times and stages.

Git, Jekyll, GitHub Pages

Facing the need to have a simple, easy, and fast way to communicate with my community of collaborators, I began to search for software and tools that could be helpful. What I first realized was that I needed some versioning control to document my process. This was particularly important because, besides working on an extensive design document, I would also program my entire software prototype.

“What is “version control”, and why should you care? Version control is a system that records changes to a file or set of files over time so that you can recall specific versions later”. (Chacon; Straub, 2014) Through research, I confirmed that Git was a suitable tool for my needs:

“Git is a free and open source distributed version control system designed to handle everything from small to very large projects with speed and efficiency” (Git, n.d.)

My first attempt was to create a private repository on BitBucket (a repository platform that runs Git). Some of my committee members, who are particularly proficient in technology and know how to operate Git, were added as contributors to that repository, so that:

• they would be able to see, comment, and change any file at any time,
• while I would be able to keep track of those changes and comments.

Image 5 - Initial project repository on BitBucket

However, after only one week, I realized that repository was not generating any engagement between my committee members and I. This made me realize that using BitBucket and interacting with only the committee member who could operate Git would not be enough for a valid, continuous evaluation. Therefore, I kept looking for alternatives and found out about Jekyll and GitHub Pages:

“Jekyll: A tool for creating static HTML websites from a set of templates and data. Jekyll is used by GitHub to create static web pages for GitHub hosted projects”. (Wagstrom; Jergensen, 2012, p. 5)

I consulted with my I.T. specialist committee members Daniel Negri and Ludmilla Aires, who both agreed this tool would suit my needs.

For about two weeks, I focused all my effort to learn how to install, operate and build a thesis project website using those tools. This experience fulfilled a set of goals I have had since the beginning of the thesis course:

• to learn to operate version control systems,
• to re-connect with coding, especially for the web, and
• to give the access to my design process and products to anyone.

This process is further described on Blog#7 New Thesis Website.

Centralization

As I realized the potentials of GitHub Pages combined with Jekyll, I began to design my website to be the central workplace and project management tool for my thesis. Here I briefly describe the functions and sections I have created and how they have helped me structure, organize, and collect feedback on my process.

Me

This section was designed to provide my basic biographic information, plus the list of committee members, their roles and contributions to the project. For example, by accessing the page titled Nicole Pallares, it is possible to see the written feedback provided by this committee member and an informal log of the meetings we have had.

Image 6 - Example of a committee member’s page

Demo

The “Demo” section displays the responsive prototypes I have created. While the final thesis paper (or design document) needs to be static, this website allows users to interact with my design iterations created on MockingBot or using Reveal.js.

Image 7 - Reveal.js prototype

Doc

By accessing the “Doc” page, users can see all sections of the design document, that is, of the final thesis paper. Each section comes from an individual local file and produces a particular permalink.

Image 8 - Screen-shot to show how document writing was done

This enabled me to work separately on each section, which has simplified and increased my workflow. Getting feedback from my community of collaborators also became easier. For example, if I have been devoting time to the Landscape Audit section, I can send a simple URL link to my community of collaborators, and they will directly access the Landscape Audit page on any device.

Logs

The “Logs” section was designed for accountability and management purposes. First, it provides a general landscape of my work:

• my current goals,
• a list of tasks and achievements
• the scheduled events
• my role in the next class meeting.

All this information is stored on a weekly basis, providing longitudinal information of my process. Through these logs, my committee members and I can determine how well I am splitting my work time/energy.

This section also stores brief reports from all meetings. It prevents me from forgetting or losing valuable hints, notes, pictures, and sketches produced during these formal or informal encounters.

Finally, the “Logs” section contains the most important upcoming deadlines and an entire Blog (or Reflection Journal) on which I freely discuss my design process, feelings, achievements, and uncertainties.

Networked Software Development Ecosystem

Since this website is hosted on GitHub Pages, all my thesis project files, including these very words, are open for anyone to contribute to, access, modify, re-use, or share.

In other words, LanguageBug qualifies as a Free / Open Source Software (F/OSS). The underlying ideologies of F/OSS (Stallman, 2002; Coleman, 2013), open licensing (Lessig, 2004) and Open Educational Resources (OER) (Atkins; Brown; Hammond, 2007) have, indeed, guided my thinking about this project from the beginning.

Wagstrom and Jergensen (2012) call this architecture of development “social coding”. The authors state that tools such as GitHub are “changing the way open source is perceived and how it is practiced” (p. 1)

Being open to social coding is a merely good start, which does not guarantee the participation of other individuals on the development of LanguageBug. But doing make me understand better the mechanisms of F/OSS, which will eventually help me escalate to an actual social coding practice. In fact,

“there is still much to learn about the roles that individuals play in open source development and how we can best ensure that these projects are successful and that individuals get the support they need to continue to grow” (Wagstrom; Jergensen, 2012, p. 11)

Getting Feedback

At different occasions, the feedback I have received from my community of collaborators was crucial to the improvement of my designs. Most of this communication has happened through either private message channels (email, Facebook, and SMS messages) or on my Slack channel, to which all my peers had access.

Image 9 - Post of initial storyboard on my channel on Slack

For example, when I shared my first draft of a Storyboard, I received a lot of attention from my peers (five people have commented on it in less than 24 hours). They seemed to agree that the storyboard was visually appealing, but did not describe well my solution and context:

hani: Your storyboard looks great Gui! It completely describes the potential advantages your app has as opposed to traditional methods in language learning for self-motivated learners. I am just wondering if maybe one or two more boxes to elaborate more why these differences exist to compare to other apps. I also love the way you develop your landscape audit!

nicole: it feels a bit more like a comparison than a storyboard. Still super nice actually but if you add a bit more detail into what actually happens inside (yours) and inside (other app) you would have two storyboards that compare which is actually pretty awesome.

These suggestions and comments were taken into account when building the new version of the storyboard (with the help of my talented committee member, Amanda Letícia). As it can be easily seen on this image, the amount of information to describe how both LanguageBug and most of the other applications operate has increased significantly.

Image 10 - How my storyboard has changed after formative evaluation

Conclusion: work-in-progress draft

The formative evaluation of the design of LanguageBug was accomplished by using Git, GitHub Pages, and by sharing each individual step of my design process with my peers and friends. I have obtained valuable information and suggestions from experienced designers which different expertise.

As a side effect, such an intense flow of suggestions and ideas has made me get used to treating all pieces of my project as work-in-progress drafts.

Although constantly lacking a sense of completion could be a problem for a final academic project, I consider this a positive side-effect. It has helped me understand I have “permission to explore” and even to fail, which aligned me with the Embrace Ambiguity design mindset (IDEO.org, n.d.).

UX Workshop

As part of the thesis course schedule, on March 8th, I was part of a User Experience Workshop with my peers and the UX Designer Shombit Chatterjee. This event was particularly meaningful because it was the first time I received feedback from peers on an actual design iteration for LanguageBug.

Reveal.js

At that point, I had just chosen Reveal.js as my prototyping platform. In its essence, Reveal.js is a presentation framework, not a prototyping tool. Therefore, unlike any actual prototyping tool, Reveal.js does not have:

• a graphical user interface (GUI), which means that all of the development had to be done using HTML5, CSS and JavaScript.
• built-in graphic assets (buttons, icons, etc.), and
• mobile function pre-sets (sliding, side menu, etc.)

Image 11 - Prototyping through Reveal.js

On the other hand, Reveal.js offers some valuable affordances, which is why I had chosen it as my primary tool for the moment:

• It allows for responsiveness to almost any user action,
• all aspects of the prototype are completely flexible and customizable,
• unlike most prototyping tools, it supports integration with audio files,
• it runs on any browser and platform, just like a regular website page.

My questions

Still unsure about the efficacy of Reveal.js in this context, I was looking for feedback on how to make the usability of LanguageBug align with the majority of apps.

I was wondering at the time:

• Is my prototype intuitive? Is it easy to use?
• Why doesn’t it look/feel like a mobile app?

I was also questioning if my prototype would engage users or not. My questions were:

• Will my peers respond to the command prompts?
• Is the flow of texts happening too fast?
• Will they be confused by so much information?
• Does my prototype make it clear where I’m heading to?

My assumptions were that following the design was not intuitive and that learner would need to process most of the information in the sensory memory. I was also assuming that user might constantly think “where should I click?” and “would I need to press anything while speaking?”, so that these concerns would affect their learning efficacy.

My presentation

The only section of LanguageBug I had prototyped was the Speak Fast principle. Principles are a set of practical instructions given in English to scaffold users on how to best operate the LanguageBug app.

The particular purpose of Speak Fast is to make learners speak aloud at a challengingly fast speed according to the application prompts.

Image 12 - Initial screen on Reveal.js: a statement without instructions

My peers, who have never seen me teaching using those strategies, opened this prototype and immediately began to follow its instructions. Quickly, they began to respond to the prompts by speaking.

Image 13 - 2nd screen on Reveal.js: first command instruction

As the principle practice continued, they reacted appropriately to the new prompts and began to speak all those words louder and faster. The whole test lasted only about 30 seconds, but was extremely effective.

Image 14 - 3rd screen on Reveal.js: reinforcing instructions

Feedback

I was very surprised and satisfied with what I was seeing! My peers had immediately engaged with my prototype and looked very excited about that whole experience. Their enthusiasm and compliments were crucial to solving my doubts and relieve my initial fears:

• My prototype is intuitive and easy to use.
• Not looking like a common mobile app does not seem to be a problem.
• My peers have responded to the command prompts with enthusiasm.
• Too much information increased their focus and was not confusing.

Most importantly, the overall purpose of my thesis project finally became clear to them, after months of unsuccessful theoretical conversations. It also became clear to me the importance of prototyping as much and as early as possible.

Conclusion

The participants of that workshop highlighted a few design issues and made some suggestions, such as:

• The flow of instructions and content was probably too fast for most users, therefore it could be helpful to make it adjustable,
• the formatting of the instructions should differ more drastically from the formatting of the sentences so that users could immediately differentiate them.

Image 15 - Before workshop: content/instructions distinction

Image 16 - After workshop: content/instructions distinction

These and other suggestions were really helpful and guided the next steps of UX development. On the other hand, these issues seemed to have minor importance in comparison to how happy I was feeling after having my design intentions approved by my peers.

Individual Project Focus

Another special moment of formative feedback on my project was the Individual Project Focus workshop I conducted on April 5th. This event was also part of the thesis course schedule, therefore, my thesis instructor and some of my peers attended it.

Presenting and discussing my design challenges with them provided me with insights on how to structure my user testing.

Differentiating

At that time, I was building a solid landscape audit of language learning mobile applications. After experimenting with and analyzing four different services I was able to show that there is a very homogeneous zeitgeist when in mobile language learning:

• most apps were based on tapping and matching mini-games,
• such mini-games rarely prompted learners to speak,
• they were not challenge or required too much focus, and
• the pace or flow of learning was very slow in all of them.

Image 17 - Most language learning apps work on very similar ways

In addition to that, I was able to identify that some language learning apps were relying on speech recognition as a way to give feedback to learners. In my analysis of those speech recognition features, I noticed that:

• processing human speech takes too much time and memory,
• apps would break on a regular basis due to that, and
• the pace or flow of learning would decrease even more.

Image 18 - Speech recognition may cause errors

My questions

Since the beginning of my development process, it was evident to me that I wanted to distance LanguageBug from these specific common practices. It seemed to me that they were overly complicated and did not help learners achieve their goals.

However, Maaike Bouwmeester, my thesis instructor, made an interesting point about feedback and responsiveness. She made me realize that if I neglect both the speech recognition and the matching mini-games, it means that I was conceiving LanguageBug as an entirely non-responsive app.

In other words, LanguageBug would never be providing actual feedback, because it would never be listening or receiving any information from the learners.

Although my first thought was that this whole idea was perfectly aligned with my language learning beliefs, Maaike’s point was that the user experience would be significantly less pleasant. In fact, I was neglecting some general rules/guidelines of interactive design. According to Saffer (2009), good interactive designs are usually:

Trustworthy: “Before we’ll use a tool, we have to trust that it can do the job” (p. 60).

Clever: “intelligence without smugness or condescension. … And it also implies delight.” (p. 65).

Responsive: “We need to know that the product ‘heard’ what we told it” (p. 64).

After realizing that, I was left with one important question: “Is it possible to provide non-corrective, or maybe even ‘dumb’ feedback in a clever and responsive way?”.

My presentation

In attempting to get valuable insights on this question, I had first to explain the origin of my strong beliefs on how language learning could work best. I began by sharing my experience as a language teacher at a school in which there was not any grammar, any testing, any correction and any homework.

Image 19 - These are now my strong beliefs on language learning

Then, since the approach used at that school was very unusual, I conducted a quick demonstration of how my classes used to happen.

Image 20 - Content covered during the brief demonstration

The way I conducted those classes is still what most guides of my design decisions. Also, having tested that particular approach with hundreds of students over many months really simplifies the process of evaluating the design of LanguageBug.

Then, I presented some quantitative data on the learning that had just occurred and collected general feedback from my peers. They seemed to have enjoyed the experience, and most of them agreed it was fairly challenging.

Image 21 - Quantitative data on the brief demonstration

The following step was to help them realize that I had not provided any actual feedback or performed any assessment. No matter how they were speaking Portuguese, I would praise their effort and encourage them to perform other actions in sequence.

After this entire contextualization, I introduced my peers to the actual design challenge and began to moderate a class discussion.

Image 22 - The design challenge I was posing

This entire presentation can be accessed on Individual Project Focus.

Feedback

Most of the initial feedback I received consisted of questions to clarify that approach or comments on the demonstration in which they have participated. My peers and instructor shared that:

• Written clues were either considered helpful or distracting,
• Longer sentences would sometimes make them feel anxious,
• The demonstration had required most of their focus, and
• It was nice to speak Portuguese at the same time other learners were also speaking.

Regarding my specific design challenge, everyone seemed to be unsure of ways to foster design responsiveness without making the learning approach corrective.

Therefore, I decided to propose a solution: self-assessment videos, which the user would record within the app and use to constantly evaluate their own performance and development.

Comments and questions gravitated around those main points:

• How would learners correct pronunciation mistakes of specific words?
• I should create a logic model around this self-recording idea.
• What assumptions am I making when I suggest this feature?
• Would that kind of video really be motivating to learners?
• Could these videos make learners feel shy and embarrassed?
• How about forcing learners to watch their latest performance before initiating another lesson practice?

All these helpful questions and comments have significantly impacted my design process. They made me realize that I had a solid suggestion for a design feature, for which I still needed a more substantial justification.

It became evident that to fill in such theoretical gaps, I could benefit from planning and performing some user testing with those questions in mind.

User Tests

The self-assessment video feature works on the assumption that if language learners are given the chance to take risks and to even make mistakes, their practice time will become more effective. Therefore, rather than constantly letting learners know if their speech is right or wrong, LanguageBug would prompt them to record a video of themselves after each practice.

Learners would then be able to evaluate their own learning progress and assess their language proficiency. The purpose of conducting user tests was to collect initial feedback on this feature and to generate more design questions.

Primary goal

My peers and instructor made me aware of other hypotheses and some concerns related to the self-assessment video feature, such as:

• Would this kind of video really be motivating to learners?
• Could these videos make learners feel shy and embarrassed?
• Would learners be interested in watching themselves speaking?

The primary goal of the user tests was to collect initial information that could guide me on verifying these hypotheses.

Secondary goals

Conduct user tests would also provide me with general insights on usability, such as:

• evaluate if the mechanics of the application are working well,
• if the features are easy to access and user-friendly,
• understand how users feel while using the app,
• discover additional gaps I would like to address, and
• find out other perceptions and reflections on the app.

Research method

I performed a “Wizard of OZ” simulation of the LanguageBug application. According to the Handbook of Human-Computer Interaction (Helander, 2014), the Wizard of Oz’ method happens when “the interface is simulated with the aid of human confederates” (p. 987). Additionally,

This is particularly useful in cases where a complex task machine would have to be built in order to test an actual implementation of the design” (Helander, 2014, p. 987)

My roles as the “Wizard of Oz” were:

• to walk users through the different simulated app screens,
• to speak Portuguese at all times an audio would be played,
• to give command prompts to the users in an established order,
• to praise effort and encourage learners to keep trying, and
• to record the learners performance after each practice.

During the whole test, I would conduct user observation. I would also refrain myself from:

• providing additional information on the prompts,
• explaining the sentences beyond their translation, and
• making any sort of correction or interference.

Brief interview

After the tests, I conducted a brief one-on-one interview with each participant, in which I asked open questions such as:

• how are you feeling?
• how did you like this experience?
• do you have any feedback or concern you want to share?
• would you try and learn a language in this way?

Post-test questionnaire

A few days after the test session, I sent the self-assessment videos to each of the users with a post-test questionnaire to assess:

• how much learning they believe they have accomplished,
• how they felt when they saw themselves speaking Portuguese,
• how they would evaluate their own performance, and
• if they would expect their performance to be better after a second practice.

Recruitment strategy

Since LanguageBug targets adults learners who do not speak Portuguese, there were many eligible user testers in the DMDL/G4L programs. I selected user testers based on responses on a quick online survey (see: Appendix 2 - Survey), which assessed:

• prior experience with or exposure to the Portuguese language,
• interest in learning Portuguese, and
• self-beliefs on their language learning capacity.

Location

The user tests were conducted at NYU MAGNET.

Duration

Each user test session lasts approximately 1 hour.

Materials

A low-fidelity prototype built on Google Slides was used during the user tests. The slides provided users with the actual content from LanguageBug: Portuguese.

Image 23 - Screen-shot of the material used on the user tests

Script

In the user tests, I covered the following content:

• Greetings
• Adjectives
• Numbers
• Self-introduction

For each of these content items, there was a slide that contained words and expressions in Portuguese. I showed users each of these slides and performed the following actions:

Speaking

1. Read aloud the first word (in Portuguese).
2. Ask the user to try to mimic my speech (read in Portuguese).
3. Repeat this action the following words, until the list is completed.

Translating

1. Read aloud the translation of the first word (in English).
2. Ask user repeat that translation after me (repeat in English).
3. Repeat this action with the following words, until the list is completed.

Practicing

1. Ask users to read (in Portuguese) all words by themselves.
2. Ask users to translate (to English) all words by themselves.

Recording

1. Let the user know I would start recording.
2. Start recording.
3. Repeat steps of Practicing.
4. Repeat steps of Practicing, but without slides.

Results

I conducted the user tests with two of my peers and committee members who agreed to be recorded and identified: Nissa and Nicole. Both tests occurred as predicted on the script (see: User Tests) and I here present my reflections on the results.

User Observation

Performance

The two users behaved and performed very similarly during the test, even though they have different native languages (English and Spanish) and language learning experiences.

Indeed, their performance was above the average, compared to any other Portuguese student I have had. First, their pronunciation was very accurate, especially for learners who were taking a first lesson in Portuguese. Second, during Practicing and Recording sessions (see: User Tests), their retention of meanings/translations was approximately at a 100% rate.

Focus

Both users seemed relaxed, joyful and often laughed during most of the test. In fact, on the post-test questionnaire, Nicole has mentioned:

Nicole: “I remember having a lot of fun (although I don’t look it in that video o.0)”

Image 24 - Indicators of focus/concentration

But when asked to perform tasks by themselves (during the Practicing and Recording sections of the test), they also showed clear signs being focused or challenged:

• Nissa and Nicole would stop smiling,
• Nissa would often touch his face or pull his hair, and
• Nicole would slightly hit the table with her hands).

Recording

Beginning to record Nissa’s and Nicole’s speeches did not seem to affect the excellent quality of their performances.

Notes

When practicing the self-introduction sentences, Nissa decided to write down the information he would need to complete the blank lines.

For example, in the sentence “Meus olhos são __”, which means “My eyes are (color)”. Nissa had learned that “castanho” means brown and applies to the color of his eyes.

However, that word was not written on the slides. Nissa had asked me how to say “brown” in Portuguese and so I told him “castanho”. Therefore, he was only introduced to the sound of that new word, not to its spelling.

On Nissa’s sketch page, he wrote the word “castanho” intuitively as “casteño”, which resembles Spanish and even includes the “ñ” character, non-existent in the Portuguese alphabet.

Image 25 - Nissa’s notes

Nicole did not ask or attempted to write down her answers at any point. One of the reasons was that, besides the new vocabulary in Portuguese, Nissa was also introduced to his weight and height in the metric system, while Nicole already knew these values.

Brief Interview

In brief and informal conversations after the user tests, Nissa and Nicole have stated that they enjoyed their experiences. They both shared that they were happy with how much progress was accomplished in such a short time-frame.

In addition to that, Nissa shared that it would have been nice if he had reviewed all the content after the Practicing and Recording sessions. According to him, when he was speaking and translating by himself, he had many questions (such as “Am I pronouncing this correctly?” and “Is this the appropriate translation?”). However, since after one content practice to another we have moved to another content practice, he did not have a chance to have those questions answered.

Upload of videos

On the day of the tests, I uploaded the recorded videos on Youtube and shared the links with each user. The playlists with such videos can be accessed through the following links:

• Nissa: http://tinyurl.com/Nissa-LBug
• Nicole: http://tinyurl.com/Nicole-LBug

To protect the privacy of the users (who agreed to be recorded), users are only identified by their first name and the videos were shared as “unlisted” (a direct URL link is necessary to access them).

Questionnaire

The questionnaire was sent two days after the user tests.

Interest

It called my attention that Nissa and Nicole have both declared that their interest in learning Portuguese had increased after the tests. This may indicate that the approach used by LanguageBug can be very effective for motivating learners.

Image 26 - Both users feel they have learned during tests

Self-consciousness

When prompted to describe how they felt when watching themselves speaking Portuguese, Nissa and Nicole both mentioned “pride” and a certain “self-consciousness”.

Nicole: proud yeah, funny also, I sound so formal haha

Nissa: I felt sort of self-conscious, kind of embarrassed to watch myself, but also kind of proud of the actual language work

This embarrassment was expected, since it was a completely new language for both users.

Not great… yet

It is interesting to observe that although Nissa’s and Nicole’s performance were among the best among any student I have had, neither of them has qualified their performance great.

Image 27 - Both users feel they could have performed even better

One logical explanation for that could be that Nissa and Nicole had noticed that their speech was not perfect at all times. Indeed, the two users spoke Portuguese with very clear traces of a non-native, beginner speaker (hard-to-understand accents, slow pace, unusual rhythm, etc.).

This might have made Nissa and Nicole compare their speech with the perfect and desired speech, that is, a native speaker’s speech (e.g. my speech). By doing so, they became aware of their shortcomings, rather than extremely proud of their outstanding progress.

Image 28 - Confidence in their own language learning skills

On the other hand, they believe that their performance could be improved by using LanguageBug a few more times, which indicates that they trust their language learning abilities and feel capable of reaching even better results (that they would probably qualify as great).

Sharing videos

• Have you shared the videos of your speech with anyone else?
• Would you be comfortable sharing your language learning progress with a network of people?

Nissa and Nicole have both answered “Yes” to these questions. Nissa added that:

Nissa: I would definitely be comfortable sharing learning progress with a network, as long as I could choose if it was totally public to the whole network or if it was only within my friends (who I had selected).

In addition to that, Nissa has shared with me on a private messaging channel that one member of his family felt very happy and proud to see him speaking Portuguese.

Conclusions

One immediate conclusion is that user tests such as this one are extremely helpful and should be done more often. Besides this, a few design suggestions and considerations have emerged, such as:

• All new content should be displayed on a written form, including information that varies from user to user, and
• Users should be given the chance to review the content after having practiced and recorded their performance.

Most importantly, the self-assessment video feature seemed to have produce feelings of shyness, but also pride. Both users were happy with the opportunity to review and assess their own performance, which indicates that this feature could actually be incorporated on LanguageBug.

Both users have shared their performances with other people, which evidences that there is a social aspect in the self-assessment video feature. This aspect could be further explored in other user tests.

Outcomes

Although this is the end of the thesis course, it is still the beginning of LanguageBug. As a result, it is challenging to evaluate the outcomes of this project at this moment. The fully working prototype was just released, and it would be necessary to conduct many user tests before making any conclusive statement.

Therefore, I here draw personal conclusions based on my experience presenting this project at the ECT Design Expo - Spring 2016.

Working Prototype

Prototyping an actual learning technology was one of my primary goals since the beginning of the thesis course. With the help of committee members, I manage to develop a fully working prototype and display it at the expo. People who visited my booth were not only informed about my project, they were also able to use it, and even learn a few sentences in Portuguese using it.

Simple and limited

Most people who visited my booth received the prototype with empathy and engagement. However, this prototype was very limited at the same time. Lacking some social component and looking like an automated slide presentation, learners seem to lose interest and engagement after only a few seconds. This highlights a need to implement more interaction so that users would remain engaged.

It got them speaking

Even at this early stage, LanguageBug has efficiently addressed the goal of making learners speak. All visitors who tried the prototype managed to speak at least a few sentences in Portuguese. It is unclear how they would react if they were using the app by themselves. Still, this highlights that a speak-first approach may be an efficient way to address the problem of foreign language anxiety.

Metacognitive discussion

Finally, most of the people with whom I interacted (booth visitors, user testers, and those who saw me presenting) seemed convinced that the speak-first approach could be effective. Even more importantly, they seemed interested in learning and discussing further. Therefore, the goal of engaging people in metacognitive conversations about language learning was also accomplished.

Reflections

After devoting so much time, energy, and effort to the LanguageBug project, I am sure that reflections will continue to emerge for years. The thesis blog (or reflection journal) has contained sincere, formative thoughts since the beginning of the thesis course. Here I present my current insights and impressions on my design process.

Research & Development

As a thesis student working on a Design & Development of Media for Learning, I was responsible for managing both

1. the research on the problem I was addressing, and
2. the development of a product that would address this problem.

I experienced a lot of difficulty and anxiety balancing these two tasks.

Mostly research

Most of that difficulty in managing my time resulted from my difficulty in making decisions and complying with deadlines.

During the first semester, I had already a clear notion of what I wanted to develop. However, I tried to focus on validating my ideas from a scientific perspective. As a result, I did too much research and advanced too little.

Mostly prototypying

In the second semester, I began to develop my thesis project website and my first prototypes. I spent most of my time dealing with codes and prototyping software, having meetings with developers, and sketching visual illustrations.

As a result, finishing my final paper on time was especially challenging.

Other tasks

In addition to research and development, I had course-related responsibilities such as:

• Updating my reflection journal from time to time;
• Setting up and meeting with a committee of contributors;
• Conducting class presentations and class activities;
• Maintaining a routine of accountability for advisors and peers;
• Preparing a final presentation for the ECT Design Expo; etc.

Each one of these tasks helped me develop valuable skills. However, I might have set up wrong priorities at different points. As a result, I experienced a lot of difficulties balancing between these course-related responsibilities and my core goals (research and development of my thesis project).

Lessons learned

As a consequence, I have learned some remarkable lessons, such as:

• early prototyping may be a better first step than doing research;
• dialogue with other individuals from different areas is necessary;
• user tests are more valuable than prototyping development; and
• when possible, speaking directly to your target audience is enlightening.

These and other lessons I have learned during the development of my thesis project are part of most design manuals. But, only through my own experience and struggle, I was able to validate them.

Gaps

This project has significant gaps, some of which are listed on this page. Some of these gaps could easily be addressed, while others would require much more research and user testing. Some great design features are still to be prototyped, planned, or even imagined.

Scientific rationale

Over the course of only one year, I was not able to find scientific evidence and develop a valid rationale to support all Principles. Therefore, these Principles remain as my strong beliefs, taken from my professional experience as a language teacher.

Feedback

Currently, self-assessment videos are the only form of assessment available for learners. Without affecting the fundamental principles of the LanguageBug approach, other forms of feedback could be imagined and/or developed.

Peer interaction

At this moment, LanguageBug does not provide learners with any chance to interact with other learners (besides assuming that learners would share their self-assessment videos with individuals in their networks).

Responsiveness

Encouragement prompts are displayed during exercise practices. However, these prompts are not responsive, but automated, pre-established messages that do not depend on the learners’ performance. Saffer (2009) states that

“we need to know that the product ‘heard’ what we told it” (p. 64).

In other words, clever designs are delightful, while the current so-called “dumb” encouragements in LanguageBug are not.

Personalization

There is an easy way to optimize the experience of LanguageBug: removing of blank spaces (“__”) from exercises. The app could prompt learners to answer some basic question and fill those blanks by itself.

Levels

This could be a way to structure repetition. In other words, levels could help learners engage with the same exercise practices many times. For example, as learners progress, the goals of each exercise would become increasingly challenging.

Statistics

Include a dashboard with more quantified data is desirable. Learners would then have a better sense of what they have accomplished, which would increase their motivation to remain engaged.

Plans

Since when I became a language teacher, I have wondered about ways to make language learning more efficient and less frustrating. My involvement with this thesis project was deep from the beginning, and I do not plan to change this after graduation. As an instructional designer and a passionate language teacher, I intend to continue developing LanguageBug.

Network

I am not by myself anymore, as I was at the beginning of my thesis course. Now, there is a team of at least three people (Amanda Leticia, Daniel Negri, and Ludmilla Aires, all thesis committee members) who are interested in keeping this project active.

Accomplishments

Similarly, I have now a considerable amount of structured research to justify my decisions and a working proof-of-concept prototype. This means that it is much easier now to advocate for this project than it was a year, or even six months ago.

Release

A working prototype of LanguageBug is already accessible to anyone on http://languagebug.org. But I would really like to make arrangements for a more structured release of a Beta version of the app to the general public.

Open-source

LanguageBug is licensed under the GNU Affero GPL License, therefore, anyone can also access, download, and even redistribute the source code of the app through github.com/guibueno/guibueno.github.io/.

A next step would be to approach different communities of interests (language enthusiasts, instructional designers, …) to look for other collaborators.

Academy

Finally, a one-year thesis project seemed to be more than enough time to develop the LanguageBug app, but it was not. I consider the possibility of applying for a Ph.D. program where I could further develop the research and the design of LanguageBug.

Not at this very moment, though. Maybe in a few years?

Appendix 1 - Exercises

Exercise: Adjectives

Get to know basic adjectives to describe people. Why?

• You will notice that these adjectives are everywhere.
• It is easy to memorize pair of words, such as antonyms.
• Your descriptions become richer with more adjectives.

Content

Exercise: Building-Blocs

Say how you are, what you are doing, how you feel, etc. Why?

• You can learn present, past, and future, all at once.
• It is possible to communicate without knowing grammar rules.
• You will notice this sentence structure everywhere.

Content

Exercise: Family

Talk about your family and people will know better who you are. Why?

• Talking about your family makes people know better who you are.
• Learn relevant vocabulary: family members, professions, etc.
• Ready-made sentences will get you started to describe people.

Content

Exercise: Greetings

Learn several formal/informal expressions to greet and introduce yourself. Why?

• There are many variations of greetings. Get to know most of them!
• Begin to personalize your new language: focus on what pleases your ears.
• Learn day-to-day expressions and you will not sound formal to a speaker.

Content

Exercise: House

Why?

• Learn relevant vocabulary: rooms, places, positions, etc.
• This is your first step to learn how to ask for directions.
• Get comfortable with using ‘there is/are’ constructions.

Content

Exercise: How often?

Learn how to add a frequency component to your actions. Why?

• Fours little words that you will be using all the time.
• Mention what you always, never, sometimes, or usually do.
• Habits make your routine, which is part of who you are.

Content

Exercise: Numbers

Count from 1 to 100, almost without thinking. Why?

• Many sentences in ‘Yourself’ require knowing numbers.
• Count as fast and as automatically as you do in your native language.
• Numbers are always the same. It is just about memorizing the words.

Content

Exercise: Shortcuts

Learn suffixes and increase your vocabulary with cognates. Why?

• Study/Studying = Estudar/Estudando: ‘ando’ is the Portuguese ‘ing’.
• Learn how to conjugate verbs without ever seeing a table of conjugations.
• Long words are usually very similar (cognates) in most languages.

Content

Exercise: Where?

Know how to add a place into your actions and descriptions. Why?

• Begin with just a few, very useful and common expressions.
• It is easier to memorize opposed pairs, such as here/there.
• Learn expressions. Preposition rules are endless and complicated.

Content

Exercise: With whom?

Know how to add other people into your actions and descriptions. Why?

• Begin with just a few ready-made, very common expressions.
• Learn flexible variations, such as ‘with my (sister, …)’.
• The word ‘com’ is used very similarly as the word ‘with’.

Content

Exercise: Yourself

Learn several sentences to describe who you are. Why?

• Talking about yourself is the most relevant content to know.
• It covers a lot of vocabulary, sentence structures, and even grammar.
• You’ll be able to speak for ~1min in your target language by yourself.

Content