Task-based Language Teaching, SLA, and AI: A Conversation with Rod Ellis

In this first episode of The Language Innovators Podcast (recorded on December 29, 2025), Linh Phung and Nikolas Wolfe interview Professor Rod Ellis, one of the most influential scholars in Second Language Acquisition (SLA), to explore Task-Based Language Teaching (TBLT) and the growing role of AI in language education.

Together, they discuss the fundamentals of TBLT, implicit/ explicit learning, and the potential and dangers of AI in language education and beyond.

This interview is hosted by Eduling: The Hub for Task-Based Teaching and Learning, an AI-powered platform designed to make meaningful, task-based language teaching accessible to every language teacher and learner. Eduling brings the most empirically validated approach to language learning, Task-Based Language Teaching, to life through intelligent agents, expertly designed tasks, and authentic communication experiences.

Eduling website

The Eduling app

👤 About Professor Rod Ellis

Professor Rod Ellis is a world-renowned linguist and a leading expert in Second Language Acquisition. He is best known for bridging SLA theory and classroom practice, particularly through his influential work on Task-Based Language Teaching, implicit and explicit learning, and individual differences in language learning.

Rod Ellis is distinguished Professor Emeritus at the University of Auckland (New Zealand) and Curtin University (Australia), Senior Professor at Anaheim University (USA), Fellow of the Royal Society of New Zealand and one of the most highly cited scholars in applied linguistics with a citation count of 152,668 on Google Scholar at the time of interview. 

💡 Topics & Questions Discussed

• What is Task-Based Language Teaching, and how has it been applied worldwide?

• How different is TBLT and other approaches such as content-based and project-based approaches

• What is a task?

• Do Natural Order, Monitor Theory, Universal Grammar have anything to do with TBLT?

• How do we design a task-based training program for languages without a written script?

• Is TBLT alone sufficient for language acquisition?

• What role can AI play in language learning and task performance?

• What aspects of human communication are still missing in chatbots and AI systems?

• Is it possible to acquire a language using only AI or computer-generated interaction?

• How can we work with AI now that it’s here to stay?

Video

Transcript

Linh Phung: Thank you so much for spending one of the last hours of the year with us. I wanna introduce you to Nik, Nikolas Wolfe, who is the co-founder and now Chief Technology Officer at Eduling. And perhaps you have heard about it. It's a platform that now we have branded as the hub for task based teaching and learning, and so thank you so much for spending time with us today!

Rod Ellis: You're very welcome. I can't imagine anything I'd like to talk about more at the end of the year than task based language teaching.

Linh Phung: So it's lucky for us then!

Nik Wolfe: No hints of sarcasm at all

Linh: So I just want to give a quick introduction in terms of what Eduling is, and I would love to give an introduction of you so that later on when we share the recording people also have this information. Eduling: the hub for task-based teaching and learning aims to make meaningful task-based teaching and learning accessible to every language learner and teacher our AI powered platform brings the most empirically validated approach task-based language teaching to life through intelligent agents and expertly designed tasks. By combining advanced language technologies with authentic communication, Eduling transforms language learning into an engaging and effective experience. 

And I would love to introduce Professor Rod Ellis, who is a world-renowned linguist and leading expert in Second Language Acquisition (SLA) known for bridging SLA theory with practical language teaching especially through his influential work on task-based language teaching, implicit/ explicit learning, and individual differences in language learning. He is distinguished Professor Emeritus at the university of Auckland in New Zealand, and senior professor at Anaheim University in the US, fellow of the Royal Society of New Zealand, and one of the top cited scientists in the field of applied linguistics. And for me, I had the honor of taking Doctor Ellis’ courses during my doctoral studies and conducting my dissertation on task engagement under his supervision at Anaheim University.

Rod: Just one little thing to add I am an Emeritus professor in two universities. You mentioned university of Auckland, but actually my latest position is really Emeritus professor at Curtin University in Perth Australia.

Linh: Yes yes. So sorry for not mentioning it but I know that Dr. Ellis was working at Curtin University in Australia and many other universities. I remember also in China and Temple University in the US. So we have a series of questions for you but I guess that we can also chat if we have follow-up questions or if you have questions for us in terms of what we do at Eduling. And actually maybe I would love for Nik to talk a little bit about him. You and I have known each other for a long time, but maybe Nik can introduce yourself a little bit.

Nik: I'm maybe a little starstruck here because I've had seen papers and had your textbooks for a while and when I realized that Linh's PhD advisor was Rod Ellis, I was like oh really. It’s funny cause you don't often get to put names with the faces. You just read these papers and they’re just like names that often occur in them.

Well by by training I am a computer scientist but I focus on language technologies and my background is mostly in speech recognition. I worked on language learning like using speech based technologies at Amazon and Google and I've kind of tried looking at this through the lens of big tech, and I find that people aren't really paying attention to the research in SLA and when I met with Linh last September and we had a conversation about Eduling, I thought that you know here's an opportunity to actually try something that isn't just sort of applying whatever people think technology should be in an app and and gamifying it and then kind of expecting the rest to follow. They're actually taking sort of a research-oriented approach first and looking for, working backwards to find the technologies that would work for it. So that's kind of my background.

Rod: Thank you very much for that introduction.

Nik: My my question for you was: Do you know what your citation count on Google Scholar is if you had to guess?

Rod: I'm I'm vain enough to check every now and again so I know.

Nik: Okay but for for the record it's 152,668. Pretty good in terms of influence, but yeah just a silly question.

Linh: So I would like to ask the second question here in the list that I sent to you. What is task-based language teaching and how has it been applied worldwide from your research as well as your experience?

Rod: All approaches to language teaching have to address really two questions, and that is: how do you organize the content of your language teaching program, and secondly, how are you actually going to implement the program.

So task-based language teaching has things to say about both the design of a syllabus, the content, and also the implementation. The essential thing about the design is that, whereas most traditional language teaching takes as a starting point, the various bits and pieces of language that you want to teach, and they constitute the content of the syllabus, in task-based language teaching you simply identify tasks that you want the students to perform, without trying to specify what particular language you are trying to teach through the performance of those tasks.

And then regarding implementation, there are various ways of implementing tasks, but in general, a key feature is that there should be a primary focus on meaning as learners perform the task, and also that there should be opportunities taken to draw learners attention to form.

So task based language teaching, like all approaches, addresses those two issues. First of all, how do you actually design the content of the course, and then secondly, how are you going to implement the tasks.

Now there are differences with regard to both of those aspects of task-based language teaching. For example, some people argue that in order to select tasks you need to conduct a needs analysis in order to try to find out what tasks your students will need to perform in real life. That is the approach that Long and advocates of Long's TBLT approach have adopted.

Others, including myself, have suggested that in many teaching situations, learners don't really have any particular target tasks that they will need to perform in English in the target language, and therefore you need some other principle to decide what tasks to select. And probably the principle that you need to draw on is students’ interests, what particular topics would they be interested in performing tasks about. So there's that difference there.

And then with regard to implementation, there are also differences. Some people, for example, argue that while tasks have been performed, teachers should not engage in any correction of students, but just simply let them perform them. Other advocates of task-based language teaching argue that corrective feedback serves as a very important way of focusing learners attention on linguistic form, and therefore has a role while they are actually performing a task, while the primary focus remains on trying to communicate.

So you can see that task-based language teaching is not monolithic, that is to say, it is not just one thing. There are different versions of task-based language teaching, but central to all of them is the idea that you base your program and individual lessons on specific tasks, and then you have to think about how you're going to implement those tasks that will facilitate language acquisition.

Linh: Yeah, thank you so much for that answer, and I'm so glad that you mentioned meaning and form. And to me, that is fundamental, that students have to focus on meaning, but also their attention has to be drawn to form.

So I just want to ask a follow up question. You talk about the starting point of designing a program by using tasks, right? But what is the difference between the task-based approach and, say, the content-based approach or the project-based approach?

Rod: The project-based approach and the content-based approaches can be task-based, but they don’t have to be task-based. For example, in a project-based approach, or even a content-based approach, your starting point might not necessarily be tasks. It could be the language that is associated with a particular content, or the language that is associated with the particular project that you want students to do.

So, content-based language teaching and project-based language teaching are likely to be task-based, but needn’t be. It depends on how they are implemented in the classroom. In fact, usually content-based and project-based language teaching involve a mixture of task-based language teaching and probably more traditional types of language teaching that are language-focused.

Nik: But if you’re designing, I mean, in most classroom-based setups, you have essentially a syllabus or a curriculum. And, at least in maybe the more traditional classroom-based contexts, there are certain types of grammatical structures and vocabulary that get added to it. And I know that that’s sort of the explicit teaching approach. But if you’re doing a purely task-based curriculum, would the individual items in the curriculum essentially be tasks, or is there something … 

Rod: The design of the content of the program centers around the selection of the tasks that you want students to perform, and then whatever the linguistic needs of that task are.

Nik: How do you approach the needs for each task? Like, if there’s a task that requires a certain language or a certain grammatical form, I guess how do you seed that with simpler tasks?

Rod: Well, I mean, this takes us into a discussion about different types of tasks. One important distinction about tasks is whether they are unfocused or focused. An unfocused task is a task where you’ve decided to include that task with no particular idea about what language that task will create opportunities for learners to learn. That’s an unfocused task, and by and large, task-based language teaching is primarily concerned with unfocused tasks.

But another type of task is a focused task, which is where you select something that has all the properties of a task, but it’s been designed in such a way that it might elicit, say, the use of a particular grammatical structure. So you might want to elicit the use of the simple past tense, or you might want to elicit hypothetical conditionals. In that case, you would choose a task that provided opportunities for learners to produce those grammatical structures, to use those grammatical structures.

It does not follow, of course, that when the students do the task, they will actually use the particular grammatical feature that you designed the task to provide opportunities for them to use. It doesn’t follow that they will. They may or they may not, right?

Irrespective of whether a task is focused or unfocused, the task has to satisfy a definition of a task. A task is a language teaching activity that meets four criteria. The criteria that I have used are, one, that the primary focus must be on meaning. In other words, when learners do the task, they should be really focusing on simply trying to communicate to achieve the outcome of the task.

Secondly, in order to provide opportunities for communication, there has to be some kind of gap, because real communication always involves dealing with some kind of gap. There are perhaps two principal gaps, possibly three,

but two main gaps. One is an information gap, where one person has information and the other doesn’t. The other is an opinion gap, where students share opinions about a particular issue, et cetera.

The third criterion is that students are free to use their own linguistic resources. There’s no attempt to direct them to use some particular grammatical structure or some particular set of vocabulary items. That is true whether the task is focused or unfocused, because even in a focused task, you do not tell the students what the focus of the task is. The idea is that the task itself will naturally elicit the use of the particular target features.

So, using their own linguistic resources is the third criterion. And the fourth criterion is that the outcome must be a communicative outcome, not a linguistic outcome. In other words, you evaluate the task in terms of whether students have managed to, for example, tell a story if it’s a narrative task, or complete a map if it involves listening to information that you have to put on a map, etcetera. There has to be some clear communicative outcome. You do not evaluate it in terms of whether students have used language correctly.

Linh: Do you know that there's an acronym for it. Recently I read an article and your criteria actually have an acronym the MGOO: M for meaning, G for gap, O for own resources, and O for outcome, and I like that a lot because now it is a little bit easier to remember the four criteria.

Nik: It's a bit of a stretch yeah like the own resources 

Rod: People vary in how they like acronyms. I think we have we have acronymic people and non-acronymic people. I am definitely a non-acronymic people.

Linh: Okay so you're not impressed.

Rod: They drive me crazy because you know I always have to go back and find out what they're referring to.

Linh: I want to ask a follow up question actually because in that article also they ask teacher to evaluate a reading comprehension activity and I mean using your four criteria and most teachers would say that a reading comprehension activity by completing a multiple-choice question or different multiple-choice questions satisfies the four criteria of a task. Would you say that it is a task? 

Rod: I don’t think that there is a communicative outcome. I think if it’s a multiple choice, the outcome is the choice, the option that you’ve selected in each multiple-choice comprehension question. So the goal is simply to choose the correct answer in the multiple choice. I don’t see that there is a communicative outcome.

Yes, there’s a focus on meaning, but there really is no gap, right? There’s no real gap there, because you’re given the text and you simply have to identify the meaning of the text. Where’s the gap? Is there an information gap? Is there an opinion gap? No.

And thirdly, they don’t really have to use their own linguistic resources, do they? Well, they do to a degree, because they have to read the text and understand the text, so you could argue that that one is met, right?

But it seems to me that actually this raises what I think is another important thing: many language teaching activities may satisfy some of the criteria in my definition, but not all of them. And I don’t think that a multiple-choice activity really satisfies all of them. I don’t think that, in particular, the outcome. It doesn’t satisfy the outcome criterion.

Linh: Maybe they they are not achieving a very specific real life like outcome in some way, reading for a real purpose right? If there is a specific purpose in the reading.

Rod: Task-based language teaching is purposeful because, basically, the idea is to achieve the outcome of the task. So the outcome or achieving the outcome, or trying to achieve the outcome is your purpose for doing the task. And that outcome must be communicative. It could be telling a story; it could be filling in information on a map.

I think when you have tasks that the outcome is clearly not linguistic like filling in information on a map or filling in information on a chart, the outcome is quite clearly communicative there. But I think you can also argue that if the outcome is linguistic telling a story or making an argument then that is also quite clearly a communicative outcome.

Linh: Yeah, yeah, I see. Thank you so much. So, we would like to ask a broader question in terms of second language acquisition in our context; it's also in the context of technology development. So, what do you think are the most important principles of second language acquisition for classroom instruction and technology development, things that we absolutely have to incorporate or consider?

Rod: Well, that's kind of two questions. I'm not too sure I can answer the technology one. I'm too old to talk about modern technology in great depth. I don't pretend to be an expert I'm not. Right.

Regarding general principles, I think that you have got to relate your principles to the two aspects of task-based language teaching that I've mentioned. One is the specification of content, and secondly, the actual way in which you implement individual tasks, the methodology of task-based language teaching. And therefore, we have to have principles that relate to those two, and that's what happens in a lot of written work on task-based language teaching.

For example, with regard to task selection, one of the very key issues irrespective of whether you're going to base the tasks on target tasks like in Long, or whether you're going to base them on students' interests as in the TBLT that I'd be more concerned with, you have to decide which task to do early on, which to do later, etc. There has to be, as always in all types of language teaching approaches, a process of grading and sequencing tasks. One of the challenges, I think, in task-based language teaching is identifying the principles that will underlie the grading of tasks and sequencing of tasks. I think that this is perhaps one area that task-based language teaching needs to actually give a lot more thought to.

The usual principle that is applied for grading tasks is the notion of task complexity. The notion of complexity figures in the grading of all types of content for a language program, including grammatical structures. So, in a traditional syllabus, one would try to teach what one considers to be easy structures first, then slightly more difficult, and so on. And the same in task-based language teaching: one would want to start with what one might consider to be easy tasks for beginner-level learners, and what would be more appropriate tasks for intermediate-level learners and for advanced learners.

But I think the notion of task complexity is not very well worked out at the moment. There are various factors that probably do influence the complexity of tasks, such as the amount of information that needs to be processed, whether the task requires reasoning or doesn't require reasoning, or whether the task requires talking or communicating about something you can see or about something that you can't see. These are factors that people have suggested can be used to identify task complexity. I'm not really terribly convinced with those. I think that, probably, in order to think more clearly about task complexity, we ought to think more about types of tasks rather than the individual design features of tasks.

For example, it does seem to me that another important distinction about tasks is the difference between input-based and output-based tasks. If you're dealing with beginner-level learners, then clearly you need to use input-based tasks, tasks that do not require them to produce because they have not yet developed sufficient linguistic resources in order to produce. Therefore, the tasks have to be listening tasks, or possibly reading tasks, etc.

Another distinction that I think is important in terms of types of task is the distinction between information-gap and opinion-gap tasks. In general, information-gap tasks can be thought of as easier for students to do than opinion-gap tasks. It's much easier to just describe things than it is to develop arguments and reasoning for things. So my own feeling is that, probably, the whole issue about grading tasks is something that task-based language teaching needs to continue to look at, and I don't think it's going to be solved by looking at individual design features of tasks like plus-minus reasoning, plus-minus information units, etc.

Nik: You can tell me if I'm reaching here, but I'm wondering because Stephen Krashen talks a lot about the Natural Order Hypothesis, and I know there's evidence for this like the errors that people make seem to follow patterns if they're learning a particular language. Is there any overlap between an ideal kind of arrangement of tasks, from simple to difficult, and natural order? Or is that picking grammatical forms too much?

Rod: I think the answer is no. I'm not even certain that the notion of natural order is very useful anymore. I think if we're looking at how learners develop a second language, you need to take a much broader perspective.

For example, when learners first start to learn a language, they tend to focus on comprehension, trying to comprehend input rather than trying to produce it. And then, when they do start to produce and of course, the natural order is actually based on production, so it ignores everything that goes before people start to produce, but I think when learners start to produce, they tend to produce highly indexicalized sentences, truncated sentences, sentences that may lack a verb, etc. They simply string together nouns and adjectives, and then gradually they progress to a stage where they begin to introduce verbs. When they begin to introduce verbs, yes, these are typically not marked for tense, and then gradually they learn how to mark the verbs for tense, etc.

So, I think you need to take a much broader-brush idea about how learners learn. The idea that there is a list of grammatical structures that are learned in a certain order, which, in any case, many researchers now dispute, is not a very helpful way to start thinking about task-based language teaching. The view of language learning that I think is much more useful is the one that has been advanced by Klein when he talks about the "basic style," and then he elaborates how that basic style develops over time in language learners. I feel this is a much more useful way of thinking about how task-based language teaching can feed into second language learning.

Nik: I'm wondering because in these other examples, like Natural Order and Monitor Theory and stuff, there's kind of this prior assumption that we believe in Chomsky's Universal Grammar. I'm wondering, does TBLT make any assumptions about linguistic knowledge or how implicit knowledge is represented in the brain?

Rod: Well, I don't think task-based language teaching draws on Universal Grammar, and to be absolutely honest, I think that's one of its strengths. I don't think Universal Grammar has anything to offer language teachers whatsoever. To be absolutely honest, the idea that there's some kind of Universal Grammar that dictates our ability and how we master a particular language is heavily disputed these days.

My namesake, Nick Ellis, has argued that learning a language is actually very similar to learning any other particular skill. It involves a set of procedures by which one develops implicit knowledge, and also the role that consciousness can play in the development of implicit knowledge. So the idea that there's something special about learning a language is perhaps not true. I personally would not support it. I think Universal Grammar has definitely seen its heyday. If you look at books on second language acquisition, they probably still mention it, but it gets a pretty small mention.

Nik: The time of the Bayesians has arrived!

Rod: What I think is Klein's view about how people learn a language is obviously entirely compatible with the view that Nick Ellis has: that we have implicit learning procedures and also we have the ability to consciously reflect on language. We need to explain how language learning takes place in terms of the interface between these implicit learning processes and the explicit knowledge that we can get about language by reflecting on it.

Nik: I just want to follow up based on that. I mean, there are certain public figures who characterize, like Bill VanPatten will say things like, "Language is not what's in the textbook; it's not a subject matter, it's something different than say, learning history or math." I'm wondering because what you said kind of cuts against that a little bit. Would you consider language how much is it like other subjects?

Rod: Other school subjects really rely on explicit learning. It's a very conscious process about acquiring knowledge and using knowledge to perform particular tasks. That's what learning history is about, or learning science is about. You have to learn the facts of science, and you have to perhaps also consciously understand the principles that underlie facts.

But where language learning is concerned, there's always the possibility of learning them implicitly. Even older people still have some ability to learn implicitly. I would argue that implicit learning is much more important in our life than explicit learning. Most of the things that govern how we live our lives on a day-to-day basis have involved implicit learning. You don't learn to drive a car through explicit learning; you don't learn how to ride a bicycle or how to tie a shoe through explicit learning. All of these things are picked up by observing and perhaps being given some kind of help and guidance in performing them.

It seems to me that this is one of the problems with putting foreign languages into the school curriculum: ideally, they require you to reorientate your thinking about how learning takes place in a classroom. It's not going to be primarily by explicit learning; it needs to be, with a language, primarily implicit learning. If you don't engage in these implicit learning processes, you'll never really develop any meaningful communicative ability in a language. You might be able to do tests and get high scores on tests, but you will not be able to communicate.

This is what goes wrong, and has gone wrong in many Asian countries for a long time—although it's changing now a lot—and that is that by and large, languages have been taught in much the same way as math, science, history, etc., as a set of facts and maybe the principles that underlie the facts. There has been very little opportunity for implicit learning. So you have situations in China, Japan, Vietnam, etc., where students may have gone through six years of learning a language in school and then you meet them and you can't even really have the simplest conversation with them. They simply have no implicit knowledge of the language, and without that implicit knowledge, you are communicatively incompetent.

Nik: That definitely matches with my public school education in Spanish.

Rod: Yeah, I mean, I learned French and German very badly at school following a traditional approach: "What are the rules of French? What are the rules of grammar?" memorizing vocabulary, etc., and very little opportunity to ever use that knowledge in a communicative type way. My Spanish is a bit better, but still very limited. No puedo hablar español bueno.

Nik: Well, I got that one!

Linh: I got that too! [laughs]

Rod: But I can actually communicate a bit in Spanish, and the reason was I learned it in cafes and streets in Spain in the 1960s. And it stuck with me in a way that French and German has not really stuck with me.

Linh: That’s why we are passionate about task-based language teaching because of the potential of teaching students the ability to use the language and to communicate in the language. My take is that you mentioned the challenge in grading tasks, but you also say that it’s a strength in some way. I feel the same way: if we provide students with rich and comprehensible input in various situations through various tasks, and if we have a lot of different types of tasks so that students can encounter language in many different situations for different purposes, then perhaps the acquisition takes care of itself in some way.

Usually, those are the two things that I feel I want to pursue: diversity of input and diversity of tasks. Even when you give a task to students like a very simple one, "spot the difference" the very basic students can perhaps complete it with their very basic language, but a more advanced learner or a speaker like myself can be more elaborate in the language. I actually like that openness of tasks in that way as well. 

However, I do understand the need to also push learners forward by maybe pushing them to take risks with the language to produce more complex language. That’s where learning happens. We want them to use language differently or in a more complex manner, and perhaps that’s why task complexity is important in that aspect. Do you feel like I understand it correctly?

Rod: If you go back and listen to what you’ve just said, your starting point is that tasks mean learner production. They need not be.

Linh: I also mentioned comprehensible input, so that is the input aspect that I think is very important as well, and I fully believe in input-based tasks.

Rod: If they are beginners, then you can’t have any production; your tasks have to all be input-based tasks. If you have complete beginners, you probably have to carry on with input-based tasks for weeks before you start putting them in a position of having to actually try to produce the language. We know that basically in any type of language learning—first, second, third, foreign—comprehension precedes production.

That is a very important principle in task-based language teaching when thinking about how to design a program. But you’re right: certain types of production tasks can actually work well with learners of very different levels of proficiency in the target language. Some students can do a task using very simple language, while others can do the same task using more complex language.

In fact, there’s a really interesting paper by someone called Lenzing, a German researcher. She used a very simple production-based task which she called the "Martian Task." The Martian Task had students working in pairs: one student had to pretend they came from Mars, and the other student had to pretend they were an Earthling. They had to simply find out as much as they could about the life of the other person. She did the same task with learners at a pretty low level of proficiency—probably moving from A1 into A2—and she also did it with students who were much more C1 in terms of the European way of classifying students' proficiency.

She was able to document that you got very different types of language from the same task. So, perhaps one way of solving the complexity problem is, as I said, to focus much more on types of tasks. Start with input before you start to use output-based tasks. When you start to use output-based tasks, try to use tasks that give students the freedom to be able to use whatever language they have in order to do it, and not worry too much. In the Martian Task, the notion of task complexity seems redundant because, basically, how complex it was depended not on the task, but on the learner.

Nik: It’s a bit like the movie Arrival.

Linh: Yeah, so we want to ask the next question because we have a lot more. I guess that you have addressed this, but I want to ask it anyway: is TBLT alone sufficient for language acquisition? This is just me and Nik thinking about whether students also have to listen a lot and read a lot, even without performing a task. That’s what learners do. For example, Stephen Krashen would say that the most useful thing that you can do is self-selected reading for pleasure. So I just want to ask: do you think that TBLT alone is sufficient?

Rod: I think reading for pleasure is a task.

Linh: Hmm, okay.

Rod: I think it's a task

Linh: Although you don't have to really achieve any non-linguistic outcome.

Rod: There is your non-linguistic outcome: that you understood what you've read, right? And maybe enjoyment; you enjoyed what you read. There's still a non-linguistic outcome. In any case, outcomes needn't all be non-linguistic; sometimes they can be linguistic.

I mean, probably Krashen would not want to have a task linked to reading that involved production. Would he be interested in a task that perhaps involved non-production, such as reading something and drawing a picture about what you've read? My guess is that probably Krashen wouldn't mind that. What Krashen would probably be against is the idea of trying to add a production element to pleasure reading, but he might be interested in some type of task that has a very clear outcome, such as drawing a picture or filling in a map.

In my mind, your examples of things that weren't tasks are actually tasks. If you're reading something, you can argue the focus is on meaning and there's an information gap because you don't know what's in the text until you read it. You're using your own linguistic resources in order to be able to understand, and there is probably some kind of outcome in terms of whether you enjoyed what you read or if you have a clear picture of what happened in the story.

Linh: Right, so there has to be a real purpose there, and that gives learners the reason to engage in language use, in language learning or those activities anyway. They care about it.

Rod: Freud talks about the "pleasure principle." That is the essential principle: that you are interested and enjoy what you read. So the outcome is the pleasure or the enjoyment that you get from reading something. Silent reading is still a task.

Nik: How does this apply when the L2 has a different script than the one you're used to? I was thinking about this because I lived in Ghana for three years as a teacher in the Peace Corps. That's kind of what got me into language technologies in general, because of the number of languages and the possibilities for building tools for them like speech recognizers or translations.

But there are a lot of languages, particularly in West Africa, where people learn them to fluency without ever encountering the written script. Some of them don't really have a standardized orthography, apart from something that may have been written down by missionaries or a colonial authority. If the script is unfamiliar or scripts aren't really used in practical day-to-day use, how do you approach early tasks or pleasure reading in those situations?

Rod: Well, first of all, you worked in Ghana. I started off my professional life as a language teacher in a very "bush" school in Zambia way back in 1967. I worked there for three and a half years in this very remote school. You're right—Africans are generally amazingly successful linguists in the sense that they often know more than one language. Zambia is a bit like Ghana; there are multiple languages, right? There is probably one that they've grown up with, but they've probably been exposed to many others as well, and they are able to move from one language to another quite easily.

One thing is absolutely certain: this ability that they have did not take place through explicit instruction and explicit learning. It took place through the natural process of implicit learning through exposure to the language, mainly oral exposure, and also the ability to interact with people speaking the language. So interaction plays a very important role. I think your example of people in Ghana, the reason why they have that kind of linguistic ability, is because they have this natural capacity for implicit language learning, and that's what they've engaged in.

Nik: I was thinking of the design of a task-based program of instruction. If you are trying to teach or learn a language that has an unfamiliar script or is primarily oral in practice, how do you approach it?

Linh: Then you just have to teach it orally, I guess.

Rod: We're really talking about two things that I think need to be kept separate here: developing communicative competence in the language—primarily orally—as opposed to developing literacy skills, including being able to identify particular writing symbols.

I think that the teaching of writing and the learning of writing is quite different from the development of basic linguistic communicative competence. I think it has to involve a very high level of consciousness. Most ways of teaching writing involve very deliberately exposing learners to symbols and their oral representation. So, we really need to distinguish the development of a sort of basic linguistic communicative competence and the development of literacy skills, which would include the learning of new writing systems such as Chinese, Japanese, or Korean.

Obviously, if you are a language teacher teaching those languages, it is quite important to address those. I probably wouldn't start off by teaching the writing systems from the word go. I would want to focus on oral input and getting them to process oral input, moving gradually to trying to say a few things in the language. But somewhere along the line, you've got to fairly explicitly teach them the writing system. The process of mastering a writing system is not entirely explicit, but I think it is much more explicit, particularly in the initial stages, than developing oral competence.

Linh: Yeah, wow. That is actually an interesting realization because sometimes I have been talking about task-based language teaching so much, but then I also teach reading and writing and I tend to be much more explicit in teaching those skills. It makes me realize that, yes, learning those skills requires perhaps more explicit knowledge.

Rod: Yeah, I mean, that explicit knowledge does not exist in a vacuum; there also has to be opportunities for implicit learning. How do people learn to spell? They're not taught the spelling of each individual word; they don't have to consciously memorize the spelling of each individual word. They pick up spelling, as Krashen has pointed out, largely by reading. The implicit processes of matching the shape of a word to the meaning of the word take place.

Even with oral development, one of the misunderstandings about task-based language teaching is that it's always associated with oral communicative competence, interaction, and speaking. But it's not. You can have tasks that are completely writing tasks or completely listening or reading tasks. It is completely wrong to see speaking and oral interaction as absolutely fundamental to TBLT. It's important, but it's not fundamental.

What's fundamental is that you have tasks that meet the criteria that I've said, and these are implemented in ways that are going to facilitate the process of language learning. Sometimes in writing, you need some explicit help. One of the things that goes on in technology is the role of glossing in reading texts. Now, a gloss is an attempt to make explicit the meaning of a particular word. You can ignore the glosses and just read on, engaging in a very implicit process, or you can stop and look at a gloss, which gives you some explicit understanding of a particular word. Long would say that glossing is a type of "focus on form"—attracting learner's attention to linguistic form while they are primarily focused on meaning.

Linh: Yeah, and that is exactly what we are doing with the Eduling platform as well. When the students read a text, we also let them tap on some language perhaps to check the meaning of a word or just a quick explanation of a phrase or a sentence. In my mind, it's a kind of incidental focus on form. But talking about technologies, we would really love to ask you about your take on technology.

I will ask a few questions together. We have a question about AI technologies and whether you have used ChatGPT or seen students use it to practice speaking. What do you think about the application of AI technologies in language learning? 

Next, what do you think about the potential of AI in performing tasks with language learners? Nik and I want the AI agent to be able to perform a task with a learner, for example, it has a secret animal and the learners have to ask questions to find out what it is, or we describe a secret city and the AI has to guess what it is.

Finally, how important is having opportunities to engage in naturalistic dialogue in language learning, and what are the features of human communication that you think chatbots or AI systems are noticeably missing?

Rod: As I’ve said, I am not an expert in technology in general and AI, but I have used it. I’ve used it a bit in research, and I’m thinking a little bit more about how you can use it in teaching.

I think the starting point for thinking about the role of AI in task-based language teaching is to, once again, take the distinction that I made—which is true of all approaches—between content selection and syllabus design on the one hand, and the implementation of a particular language learning activity or language teaching activity on the other. I think the way to think about AI is to consider: How can I use AI in terms of task selection and task grading? What role is it going to play in developing the content of my task-based language program? And then secondly, how can we use AI to actually help the process of implementing a particular task?

I really think it is useful to make that distinction and talk about AI in both. I’m much more comfortable with the idea of using AI in terms of task design and task selection. I think it’s an amazing tool. We know from early studies in task-based language teaching that one of the problems teachers often face is that their textbooks don’t have enough tasks, or don’t have enough interesting tasks. Therefore, one of the problems is how they can do task-based language teaching when they don’t have the actual material resources, the tasks. Obviously, AI can solve that.

I think what’s involved here is thinking about how AI can be used to design different types of tasks. What prompt do we need to give to an AI system in order to come up with an input-based task that would be suitable for, say, beginners or intermediate-level learners? The design of the prompt is an absolutely crucial factor in the use of AI. The same applies to a focused task; if we want to design a task that would provide opportunities for learners to use the hypothetical conditional, what prompt do we give it?

Remember, the prompt that we give must tell the AI system that it has to develop a task, not an explicit exercise. So the AI system has to know what a task is. It does, but if you just put "task," it probably won’t do it because the word "task" tends to be used ubiquitously for any type of language learning activity, including multiple-choice and exercises. It needs to know what you mean by task. I think I’m probably more interested in how we use AI to design tasks suitable for learners of different stages of development.

With regard to implementation, my experience of bots, when I go into systems where I go online and you can get online assistance is that it is painful.

Nik: It’s like pulling teeth, right?

Rod: It’s painful and very frustrating. Again, with regard to the use of a bot, I think one probably has to think about how you use it in two ways. One might be: Can we program a bot to interact with you in order to perform a particular task? Can we tell the bot how it needs to interact with you? I think that is what a lot of stuff does; it tries to program the bot to respond to what you are doing.

Recently, I started to get involved in a new AI-developed language program in China. I’ve now opted out because I found it so frustrating, actually. It was where you could arrive at an airport, and then they devised a task where you could go to the information desk and ask about hotels. I started to use that and it just didn’t tell me what I wanted to know. For example, what I really wanted to know was: "What is the hotel that is close to the airport and gives me free shuttle service to it?" It couldn’t cope with that. It couldn’t do it because it hadn’t been programmed to deal with that particular question about "how far"; it was programmed for "Can you suggest a hotel?"

So, I’m very, very dubious about the use of AI to program a learner's performance of target tasks, like how to find a hotel near the airport that gives you a free shuttle. I’m very dubious about that because I think you’ve got to consider that there are potentially so many different questions you could ask about the hotel, and the bot is not going to be able to cope. It’s only going to be able to cope with the ones that it’s been programmed to cope with.

What happens if you’re interacting with a bot in a much more open task where the bot has not been programmed—the only thing that the bot knows is the task materials and the task work plan? You tell them what the task work plan is; you specify what it is they have to do. If it’s an information gap, you provide the bot with the information that it will have.

Linh: Yeah, that’s what we do. It interacts with a person who’s got different information.

Rod: I don’t know what it’s like when it does that. You tell me. Can it do that? 

Nik: More or less

Linh: Yes, so I'll give you an example. We are still fine-tuning it to make it as engaging to learners as possible, but we think that it’s an opportunity for learners to practice speaking.

Again, students in many different countries don’t have these opportunities, and maybe they have had the experience with chatbots that you have had. It’s kind of very open-ended, it is kind of boring, and it keeps saying the same thing. We know some of the limitations of those chatbots. But if you create an information-gap task and you give it some information, and you tell it that maybe it doesn't know the information that the learner has, then the learner has to solve that problem or complete a task with the chatbot.

I personally find it much more engaging than any open-ended conversation that you can have with a generic chatbot out there. In terms of implementation, Nik knows more, that we are fine-tuning it so that, say, it can give a recast or it can negotiate meaning, or maybe at some point it can give explicit correction or some kind of metalinguistic explanation later on. If you want to try it, we don’t have it exactly out yet, but if you want to try it, that’s what we are trying to do.

Rod: I think that is much more useful. I mean, you know, there’s a lot of literature on the dangers of AI. If you’re able to develop bots that can interact and negotiate meaning as effectively as a competent speaker of the language, then you don’t need the competent speaker of the language anymore. You don’t need the teacher. AI replaces the teacher.

This is what people are saying. You know what people are saying: "AI is going to change the world" because it’s going to make a lot of the things that people now do to earn money and live their lives, their work, redundant. They don't need to do it anymore; they can hire AI to do it. That concerns me. That really concerns me. And obviously, I suppose ultimately you can develop an AI system that will interact in the way that you’ve just described, Linh, just as effectively as a competent speaker would. The AI becomes the competent speaker. Do I look forward to a world where that happens? No, but I won’t be around.

Linh: But my feeling is that students still want talk to people; the reason why they are learning a language is that they want to talk with speakers of that language. So, it just gives them the opportunities to learn the language, and then they will seek more opportunities to listen to authentic materials, YouTube videos, and try to interact with speakers of that language because to me, that's still more interesting. Maybe you have a take on that, to see if at one point students may just want to talk with the AI instead of real human beings.

Rod: Well, you know, another difficulty with AI is that it removes the need to learn a language at all. 

Nik: You mean for translation? 

Rod: Yeah, with instant cellphone translation. Just say what you want and it comes out in English. I've had very successful communication with people in China where they do that all the time. I'm not doing it, but they are; I say something and they record it, they work out what it is in Chinese, and then they say a reply in Chinese and it comes to me in English.

It's a bit of a tedious process, but most people don't need to engage in very rich interaction and rich communication in a foreign language. They probably need a foreign language in a quite limited way. It completely removes the need for any sort of implicit learning because it's a very explicit process: say something, translate it, and then 

Nik: Yeah, yeah

Rod: So, I really do fear what AI is doing, both in terms of producing bots that can communicate effectively with you doing an information-gap task though I think it might be much more difficult to get a bot to really do an opinion-gap task effectively. Because an opinion-gap task is much more open, whereas an information-gap task is normally quite closed. A task that involves developing argumentative positions is much more open, and I wonder to what extent you can develop bots that will effectively do that.

Linh: Yeah, people try to debate with it and it seems like it's capable, but maybe its ideas are still somewhat banal and not as interesting as hearing from someone with real experience expressing their own ideas and telling their own stories. To me, it's still kind of boring to debate with the bot.

Rod: But let me ask you another question about bots, alright, do bots have a sense of humor?

Nik: No.

Rod: Well, sad.

Linh: Okay, so it cannot replace us then!

Rod: Yeah, I mean...

Nik: But that is the fear you express, and I think that’s definitely well-taken. That is like one potential outcome of the current generation of these technologies. But regarding the practical use of what's there now, even for something like voice translation, like the example you gave, there is a wide, wide gap between the "public relations" of these technologies versus the actual, day-to-day, on-the-ground experience of using them.

We've been thinking a bit about how we can get it to, for instance, do something like a closed-form task. So, if there's a known answer, but you still have to weave and talk through the series of steps to get to that outcome.

Rod: That's where I think bots will work best. They will work best with closed tasks.

Nik: I mean, the open form... well, with open tasks...

Rod: Let's go back to the humor thing. What would be one of the characteristics of a person you would consider to be a good teacher? A sense of humor. Someone who can make you laugh, who can joke, right? And if bots can't do that, then you know, what a shadow of what a real teacher should be. 

Linh: Which is fine. That's why we are thinking of creating engaging opportunities for students to talk with the chatbot to solve these kinds of information-gap tasks, but then it cannot replace teachers in the sense that when you talk with teachers, you talk with a real person.

Rod: It will... you know what drives the world? What drives the world is money. I'm sad to say what drives the world is money, particularly in the United States these days. And if people can get bots to do tasks very effectively, then why hire teachers? Just buy a program with lots of closed tasks where they can interact, etc.

Nik: I would say one ray of hope there is because we mentioned two different kinds of modalities: there's the actual implementation of a task, and then the design or production of the materials.

Rod: And with the design and production materials, I'm a big supporter of AI.

Nik: But the experience that I've had with them is that, maybe ironically, that's the thing that they're worst at! If you ask it to produce a task and you have a prompt out there that has the four criteria for a task and maybe some other parameters like open form, closed form, or what type of information gap, you could supply all of that and then ask it to produce something, but it's going to give you five tasks and then it's just going to start regurgitating different versions of them.

We find that, at least, the creativity that goes into some of the expert-curated content, like a "spot the difference" task if that task exists already, your AI system might be able to regurgitate it just from the training data, because it's basically been built to memorize more or less the entire internet. So, if it exists already, then it can likely produce that or something adjacent to it. But new tasks that are fun and interesting, or funny, maybe the premise is funny, you don't find a lot of that creativity from things that are AI-generated.

Rod: I'll send you something that I asked DeepSeek to do. I use DeepSeek; it's free and it does most of the things I want it to do for research purposes. I'll send you what it produced, including the prompt that I put in so you can see it. I think it's come up with quite an interesting, well-done task. It's an opinion-gap task. I plan to be experimenting with trying to get it to produce interesting input-based tasks. The other thing is that you're right about creativity: it will tend to follow a particular pattern once it's got it.

Nik: Right. Although you can find ways of breaking the pattern by the prompt that you put in by telling it not to do something and to do something a bit different. You can do that.

Rod: So, yeah. I think I'll tell you something I heard recently. I was talking to someone and we were talking about publishing EFL/ESL coursebooks. This person was telling me that they've heard that one publisher is no longer signing contracts with experienced teachers to write coursebooks. What they do is hire one teacher simply to review materials that were developed by AI. So the whole coursebook is now developed by AI, and the only person they're hiring is one expert who will look at it in order to suggest fine-tuning. That's the future of EFL coursebooks. It's nightmare-ish, but it's happening. It's happening.

Nik: There's a good take from a comedian—I think he was on The Daily Show—talking about AI taking jobs away. Particularly in the field of language technologies, like if you're a translator or transcriber. Basically, the idea was that the framing of the statement "AI is taking jobs" is a little bit backwards. Because what's really happening is somebody gave your job to AI.

There are choices being made by people who have the power to move resources around, thinking, "I'd rather hire one person to edit the output of AI as opposed to a number of people that are going to come together and write new content for a textbook." And those cost-cutting measures, the end result in a lot of cases is vastly reduced quality. Maybe the powers-that-be don't place that high of a market value on those kinds of things, but in general, that is the trend.

We're essentially trying to build something that would do what a trained instructor would do, and I suppose if it was good enough, the argument could be made that it essentially replaces the teacher. But the model doesn't necessarily have to be that you get rid of teachers, because in our experience, the curation of the content and the involvement of humans in the loop to test, grade, and fine-tune the outputs is a lot better of an experience than just letting the AI do it. There are quite a few apps already that are just putting it in AI's hands and letting that be the driver of the entire thing. So, yeah, it's a little bit bleak. It's kind of ironic to talk about these things while being in a company that's doing it.

Rod: I mean, the thing that changed the world was obviously the World Wide Web, and this is still part of AI as well; the AI couldn't function without the World Wide Web. So, that was the first big change. That brought huge advantages, like email, so that you can communicate with people very quickly. But also disadvantages.

For example, in 1967 to 1970, when I was in Zambia, I used to write very long letters to my parents in England and receive letters back from them. That was very important. I still kept those letters, by the way, the letters I got back, and I still read them sometimes. They remind me of my father; my father wrote them, so they remind me, right? Did you do that in Ghana with email? 

Nik: Yeah

Rod: Long ones?

Nik: I mean, I had a blog that I tried to write at one point and I wrote some fairly longish entries in that. 

Rod: Would you say you're typical? I mean, what I'm getting at is that probably most people would communicate with their parents by email, but you know, one-paragraph things, not kind of 500 words.

Nik: Yeah, I don't... well, I'm writing letters!

Rod: Oh, there you are! I might be atypical in that case.

Nik: Because I do write emails that are too long. You can probably tell from this conversation that I'm too wordy!

Rod: In terms of the trend of what do people in general do...

Nik: Yeah, a lot of it. I mean, I had Facebook; I taught ICT skills, so I actually had a computer lab connected to the internet in Ghana. That carried its own whole different set of circumstances around teaching and access to the internet, and what it was like being in Ghana but connected to the internet at the same time.

Rod: But I think you were quite unique in that respect. Most people writing long emails... yeah, that's... I mean, you see something else that I think has happened as a result, certainly, of technology, and that is that the "long turn" has gone. The long turn is when you produce speech that consists of a number of different sentences strung together, as opposed to communication that consists of a single sentence or maybe two. Obviously, X is destroying the idea of the long turn.

Linh: Right, or texting in particular.

Rod: You know, short turns are very, very dangerous because you don't develop a position; you just sort of put your position there. So, I see the world is losing the long turn. I see the new generations growing up no longer really capable of producing long turns. Many people are not really capable of producing long pieces of writing, or are incapable of producing spoken language where they are putting together a coherent set of thoughts. I see it disappearing.

Linh: The writing part is in danger for sure, now that you can just input some ideas into AI and let it form a paragraph for you.

Rod: Well, one thing you might want to look at regarding the use of AI is to what extent are you providing opportunities for learners to produce long turns? Because my guess is that bots tend to focus on fairly short turns.

Linh: We still have a kind of activity where learners record a two-minute speech, for example. I like it very much.

Rod: I think that's important.

Linh: They have to tell a story, and there has to be the beginning and the end.

Rod: I think that's important, yeah. I think it's an important aspect of language learning that you are not simply engaging in interaction with very short turns, but that you can actually produce a long turn orally as well as in writing.

So, I think there are these things that we need to think about. To my mind, I much prefer the idea of using AI by individual teachers to help them develop tasks that they can use for their particular students. Therefore, what we need is to train teachers to produce prompts that will enable them to get good stuff from AI.

Linh: Yeah, but for them to do it well, they need to know about task-based language teaching. They need to know what a task is, and they need to be able to edit the materials. I do see the advantage that teachers have in terms of knowing their students and knowing the context. If they are able to edit, then the materials are much more interesting than what you and I can write from a distance and then give to them. That's my experience as well: very specific details that are really related to their students' experiences.

Rod: You know, AI is with us, and therefore we've got to work out what ways we can use it. But there is a hell of a lot written about AI that is looking at the dangers. The dangers are what I've been talking about. I don't know how you deal with that because, basically, AI is money. What drives the world is people making a profit out of something.

Nik: I was going to say, ironically, there's very little profit to be made.

Rod: More money by using AI than using real-life human beings, they'll do it. I mean, maybe we should do what I think Finland did, which is everybody gets a minimum amount of money, a certain amount of money paid to them that the government simply gives them. Enough money to basically survive in the culture; not enough money to own beautiful things and have a great lifestyle, but enough money.

The idea is that instead of having welfare, we just give everybody a basic sum which enables them to live. Now, of course, people who want to have a better life or a more interesting life will go and get jobs and do things, but people who don't just have the basic amount and that's it. There are no welfare payments because everybody gets that basic thing. I like that. The basic needs exist for everybody, so we'll provide, and then it's up to the individuals whether they want to add something to that? I really like the idea. I haven't read too much about how successful it's been, but I think it's a fascinating idea.

Linh: Thank you so much for your time and enjoy the rest of the day.

Rod: Thank you very much for the opportunity to talk. It's been very interesting, very entertaining, and next time I'll listen more and learn more and speak less.

Linh: Next time we'll show you our chatbot and see if you like them. We still think that students still want to talk to people.