We’re delighted to welcome Brian Anderson to the blog. Brian is a gamer and a publishing consultant. He blogs and tweets. This post first appeared on fishhouses.wordpress.com.

Maybe this is a personal thing, but I never found those CD-ROMs you’d buy at the store effective for learning a language. To me, they always felt like putting on blinders. I’d follow them from point A to point B, and my interest always lagged behind me. It lagged even when I had convinced myself that since the digital people who lived at point B spoke a certain way, I could speak like them and I’d become one of them.

The earliest approaches to computer assisted language learning (CALL) were brutally sequential. We still see that legacy in the some of the most prominent CALL products. This approach is not dissimilar from that of rail shooters like Area 51, in which gamers are carried through a sequence of scenes.

area51
Area 51, an early rail shooter.

If I had to pick one single feature of CALL that contributes the most to a negative user experience, it’s the closed nature of the majority of CALL applications, particularly in a time where our lives are increasingly digitally infused, and our cognitive horizon has expanded to encompass such extensions like Google search, Wikipedia, our email inbox, Twitter, and more. It’s one reason that learners feel trapped in a learning application in a way they they never would with a book. When learning is placed on rails, learners are either passively conveyed or left at the station.

Again, this is subjective, but I hate rail shooters. Similarly, the rare times I have been successful as a language learner, it’s when I’ve used many different sources, some books, magazines, radio, movies, following the natural path of my inclination, and supplemented that with whatever scaffolded solutions I had on hand (i.e., language learning textbooks).

The Discrete Syllabus

A lot of the backlash against adaptive learning I see in ELT comes from an understandable and longstanding weariness with the itemized approach to language education. A reliance on linguistic features we can quantify (vocabulary, grammar) lead us to the often erroneous conclusion that correlation is causation i.e., since learners of this level of proficiency tend to use this set of words, we can bring other learners to this level by teaching them this set of words.

For lack of a better model, we’ve transferred this to the adaptive learning environment. We are still telling learners to go straight to grandma’s house.

path
“GO TO GRANDMOTHER’S HOUSE — AND STAY ON THE PATH” – The Path, an art game by Tale of Tales. What do 99% of players immediately do?

Language Learning in the Age of Recommendation

A rails-based approach to language learning is troubled by the new environments and new possibilities offered by our web-based lives. More and more aspects of language learner engagement are becoming measurable: personal data, the learner’s social network (lovingly provided through Google, Twitter, and Facebook APIs), geolocation, patterns of engagement with a site, affiliation to certain topics, linguistic features of their online correspondence, even their emotional state (wearables like Google Glass, which will start collecting affective data based on eye movements and other biological indicators). In the near future, we’ll start seeing the same sassy robots that tell us “X users buy Y”, or construct playlists of cat videos for us on Youtube, will be smartly constructing i+1 environments for learners.

When that is all put together we may discover some new truths about language learning, but it’s more likely that the big data systems that are ingesting and processing this data will generate new and disposable custom truths at such a rate that we will cease to care anymore about the internal structures of language learning. As online platforms offering real-time analytics and recommendation engines become more ubiquitous, we may find ourselves content to “black-box” the language learning process and concern ourselves more with tangible results. In other words, the increasing complexity and sophistication of our online applications is well suited to serving the sort of language learning as described by Diane Larsen-Freeman.

It’s possible that we’ll discover that providing the student with highly relevant, engaging content will be more effective than crafting learning experiences on rails (à la our current, item-response-theory-influenced “adaptive learning”). Maybe we’ll find the sweet spot between the two: we’ll find effective ways of spinning up scaffolded, customized learning paths within this rushing environment of constantly-updated, machine-learning-recommended content. Paths that, like a well-constructed video game, make you feel like this was your own personal journey all along, and that you were following your fancy, not a syllabus on rails.

These paths are not entirely learner-generated; nor are they dictated by a teacher. Instead, they emerge through interaction, the result of a complex negotiation of a learner with the learning environment. Much like, in a successful classroom, they emerge between the student and the teacher.

These are the paths that learners actually complete. And when they get to the end, they can say: “I was headed there anyway.”

skyrim
Skyrim, one of the most open of open-world games.

9 Comments

  1. I am sorry but I can’t figure out what the point of this “post” is. I couldn’t find a thesis sentence and the concluding remarks don’t point backward at anything I read earlier.

    And before you take the rails analogy any further CD-Roms were typically branching systems (based on Hypermedia). They would take you to a point A and you would have the option of going off in a number of directions B, C, maybe D. Yes, it was a closed system but the basic designs of closed systems haven’t changed a lot they have just gotten much bigger, and more complex.

    CD-Roms were in fact much more branching than any book that I have ever read. That doesn’t make CD-Roms better but they were not mere rails. They are much better compared to a small ralroad.

    I guess you guys have gotten busy with your day jobs…..

    1. Hi Mike,

      It’s true that here at ELTjam we are indeed very busy with our day jobs! However, we enjoyed this post and felt that it made some interesting observations about adaptive learning, learner control and the paths which learners are encouraged to take.

      While we welcome a range of commentary, both positive and negative, we would ask that people are generally respectful to fellow writers and colleagues.

      It seems that you have a lot to say, Mike, and so if there is something you feel strongly about then please do approach us as we’d definitely consider publishing a guest post of your own. But again, it’s important that we keep this blog a supportive and constructive environment for all our contributors.

      Jo

      1. Sorry, bad day. I was much too harsh. The writer didn’t deserve such a treatment. Let me change my tact.

        The writer claimed:
        “A rails-based approach to language learning is troubled by the new environments and new possibilities offered by our web-based lives.”

        First, I don’t understand from this article what is meant by a rails-based approach to language learning. I am unfamiliar with this term. Is it new to ELT? I don’t think it was well defined in this article. If he meant by this, “a linear approach” I believe the writer was mistaken with regard to the example he gave. CD-Roms typically are branching systems.

        Second, after falling short of defining this new term (in an ELT context) he began to criticize a rail-based approach to language learning. Could he give a concrete example of this failure? In other words, what exactly is a rail-based approach and how does it fail us?

        Finally, is the writer suggesting an alternative? How would this alternative be different?

        I apologize in advance if I missed the definition or the example that supports the writer’s case in this posting. I look forward to learning more about this rail-based analogy (or approach, or methodology).

        1. Hey Mike,

          The headline is a play (I think) on words mixing ‘ruby on rails’ programming with language learning pedagogy. The hidden meaning is that ‘code’ is the source of creation, rather than a ‘rail’ or ‘course’. I thought it was quite clever.

          Regarding, ‘evidence’. Well, we don’t allow the student to create the data so ‘the evidence’ is based upon what we want to collate – right?

          I get really annoyed. Here in Brazil, the average learner who has been taught by ‘mainstream teaching’ is around intermediate, no matter how hard they try. And what i noticed is, they lack the ‘tools’, which is the ‘code’ behind the art. So, it appears to me we need to make the learner the ‘artist’, not the copycat. Yep – we can copy and paste, fork, iterate – but, only if the ‘teacher’ can. And they can’t.

          In New Zealand (I follow ed in NZ and Brazil) they are now talking about the merger of subjects, between English and History for example,so, a method of code switching. And to get there, we need collaboration, sharing, and transparency. It’s a complex subject, and in the end it comes back to democracy. I liken it to ‘political parties’, the big publishers dominate perspective through media, and therefor they dominate the market, with little or no ‘listening skills’ – meaning, the public has no say. Who are the public? They’re the learners who can’t afford the app, or the textbook, or the expenditure associated with ‘traditional education’. So, tailoring education to exclude the poor, is really counter-education, and the global publishers.push try as hard as they can to fight democracy and the education of the poor, eco-pedagogy as described by Freire, the community based educational learning model.

          Mmmm a pedagogy that has ‘purpose’ is not associated ‘for profit’, and is democratic, peer to peer, iterative in construction – it includes the leaner, and the writer suggests that our current hierarchy does not, or at least that’s how I interpreted it. 🙂

          1. @Robert,

            Thanks for trying to parse this for me. If that is the case I was thrown off my the author’s introduction which didn’t include a specific mention of code but did include a mention of sequential learning (a rail???). I look forward to getting this cleared up by the author.

            It bothers me a bit that I can’t figure this out and everyone else seems to have done so.

  2. Great post Brian. A couple of thoughts:

    1) Your idea that we may end up “black-boxing the language learning process” is somewhat similar to what has happened in machine translation, where the older rules-based approach using syntax and semantic analysis is being replaced by a big data statistical approach (e.g. Google Translate) where no one really cares too much about the rules of language, but instead just rely more on brute force statistical relationships.

    2) Regarding the rails-based approach and its limitations: as you know I’m hugely sympathetic and I agree that we’ll need a hybrid where, although there are some rails in place, they are supportive rather than restrictive, thus scaffolding an open environment but also placing certain constraints in place. I think there do need to be constraints and structure in order to keep learners in the flow state, rather than wandering around, getting distracted and losing focus. Recently our team has been experimenting a bit with designs for a “learning layer” that sits on top of any webpage and provides an interface between the learner’s interest and their language study. It’s very early stages still, but it’s sure fun to play around with the designs….

    3) We need to be careful not to conflate adaptive learning with the discrete-item approach. It’s the very earliest stages of adaptive learning and just because one company had decided to do it a certain way doesn’t mean much – we need to refuse to allow Knewton to define our terms for us. I tried to show another way to provide teacher-driven adaptive learning in those ridiculously long posts I wrote here a while back, but I wasn’t very effective it seems because everyone ignored that part and just focused on Knewton again. Oh well.

    4) So what happens in the classroom? Just like we have a bad habit of thinking “adaptive learning = discrete-item approach”, I think we also tend to make the “SLA technology/software/apps = teacher replacement” equation all too often. I’m not implying you are saying this in your post Brian, but I feel that it’s usually an implicit assumption that the teacher is being replaced; whether that is good or bad, it seems to be the underlying theme. Maybe I’m reading too much into the discussion, but I still am convinced that we’ll see optimal results from a combination of synchronous teacher-driven classes, and asynchronous self-study by the students. In other words, the classic blended, flipped classroom approach. If the asynchronous part turns out to be your “sweet spot” hybrid case (and I think it will), then we can connect that into the human, interactive classroom experience, I think we’ll have a solid step forward.

    1. Good point! I didn’t make much reference to teachers at all here, primarily because this essay came more from my experience as a consumer of educational products, and less from my experience as a teacher.

      “Supportive” vs. “restrictive” is a nice framing. And love the idea of a ‘learning layer’. I have a deep fondness for Learning With Texts, and would love to see a less clunky version that could integrate with the browser.

  3. Spot on.

    An app goes so far. I think we’ve all used them for a minute or two and once we´ve figured out what they do, we leave them never to return. CALL doesn’t ‘power up’ the learner, instead it provides a band aid to real teaching and learning. What’s missing here, is the ‘self directed’ or ‘informal’ learner who discovers and solves problems through code with the teacher and other learners. Coding is the language behind the app, the source, the enabler. But, it’s a language that teachers don’t know or use such are the purveyors of traditional hierarchical models of professional development.

    A model that begins with the teacher, not the student, will always focus on the data, because the data and the test results and the assessments that the educational institutions require, justify their methodology, and high fees through data. Without test results they can’t prove anything. So, the focus needs to move away from the teacher, and publisher, and app back to the student. Data can be manipulated, depending on how you want to use it, depending on the content. But, the content comes from the teacher, and I think that needs to change. If the student is responsible for the content, the data, they’ll be motivated to see what they can do with it. To create collaborative generative art from the data with peers and ‘share it’, ‘badges’ of learning that visualize new found skills. And provide art, light, creativity for the school and teachers.

    New methodology, empowers the student for life, not a test. An IELTS exam is designed for testing, not communication. A DELTA test uses the same methodology, top down, hierarchical teaching which immediately limits the learners creativity and engagement through the premise ‘we know more than you’. Which of course, is not true.

    1. Thank you Robert. I think you hit the nail on the head — more content from the learner, less from the teacher.

      Very interesting point about data. It’s rarely used to improve the quality of the learning experience for the end user (i.e. the learner) and more often used to optimize the value for other stakeholders (schools, employers, etc.). Who are we optimizing for?

More comments

Leave a Reply

Your email address will not be published. Required fields are marked *

TwitterLinkedInFacebook

Other related posts

See all

Am I a Content Creator or a Writer?

Deconstructing the Duolingo English Test (DET)

My English learning experience – 6 lessons from a millennial learner