We're sorry but this page doesn't work properly without JavaScript enabled. Please enable it to continue.
Feedback

Spotlight Lecture Series 6: I'm Going DeepL Underground - AI-Powered EFL Learning

00:00

Formale Metadaten

Titel
Spotlight Lecture Series 6: I'm Going DeepL Underground - AI-Powered EFL Learning
Serientitel
Anzahl der Teile
13
Autor
Mitwirkende
Lizenz
CC-Namensnennung - Weitergabe unter gleichen Bedingungen 4.0 International:
Sie dürfen das Werk bzw. den Inhalt zu jedem legalen Zweck nutzen, verändern und in unveränderter oder veränderter Form vervielfältigen, verbreiten und öffentlich zugänglich machen, sofern Sie den Namen des Autors/Rechteinhabers in der von ihm festgelegten Weise nennen und das Werk bzw. diesen Inhalt auch in veränderter Form nur unter den Bedingungen dieser Lizenz weitergeben.
Identifikatoren
Herausgeber
Erscheinungsjahr
Sprache

Inhaltliche Metadaten

Fachgebiet
Genre
Abstract
In this lecture the two speakers gave an overview of the genesis of AI technology in general and the influences it has had already and will inevitably have on language learning and language teaching in the future. Apart from that a light was shed on the rather philosophical question of how the improvement of AI will reduce or enhance the importance of human professions such as the English language teacher.
Schlagwörter
Deutsch
Deutsch
Englisch
Englisch
ComputeranimationVorlesung/KonferenzBesprechung/Interview
Vorlesung/KonferenzBesprechung/Interview
ComputeranimationVorlesung/KonferenzFlussdiagramm
Vorlesung/KonferenzComputeranimation
Technische ZeichnungComputeranimation
ComputeranimationVorlesung/KonferenzFlussdiagramm
ComputeranimationDiagramm
Computeranimation
Computeranimation
Vorlesung/KonferenzComputeranimation
Computeranimation
ComputeranimationVorlesung/Konferenz
Computeranimation
Transkript: Englisch(automatisch erzeugt)
In this chapter, we're going to talk about the basics and practical usage scenarios of artificial intelligence. Artificial intelligence is a branch of computer science that deals with intelligent machines
and algorithms. In general, artificial intelligence is used for data mining, speech recognition and also facial recognition. We're going to talk about these aspects later on.
AI is also used in autonomous robotics to make it seem as if robots have human-like thoughts and emotions. That is probably the most controversial facet of artificial intelligence. So to say, that's a bit scary for people and that's why we have made this reference
to Jamiroquai, an artist you probably don't know. AI is also an area that tackles societal issues, also education and language learning.
The Swiss Federal Institute of Technology in Zurich published a paper saying that experts already predicted a certain kind of tech society 30 years ago.
Furthermore, this article mentions that Steve Jobs predicted drastic improvements in digital communication and also mentions Google's Larry Page that he believes that within the next decades there would be a superior use of technology.
Well these people couldn't be more wrong if we consider and have a look at the meme and TikTok culture. Have you ever wondered why, for example, when you log into Google or your mail account
you're sometimes asked a question like, are you a robot? And then you have to click on pictures and you have to say, which of these squares shows parts of a traffic light, for example, or parts of a street sign? Well, you have to do that because computers cannot really do that.
They cannot predict or they cannot say this is a street sign unless they learn what a street sign looks like. Humans can do that. We can say this is what a typical traffic light looks like, but of course we have to
train computers. So millions of people around the world click on these pictures and by doing so they train the artificial intelligence. And of course a technology like this is necessary, for example, for cameras in autonomous driving in cars so that the cars can stop in front of a red traffic light.
So let's have a closer look at artificial intelligence and the mechanics behind artificial intelligence. AI is by no means a new technology.
For approximately 70 years, computer scientists have tried to build certain systems that can at least partly replace certain kinds of human behavior and do things that only humans can do. In most cases, we're talking about narrow AIs, which means that these systems built can
do one thing very well, for example, playing chess and, for example, win against the world champion of chess. This overview shows that over the last decades, many different systems have been
developed. You might all be familiar with speech recognition systems, systems like Siri and Alexa, for example, Siri launched in 2011, of course, the system that's supposed to help you to assist you in your daily life. But of course, there were other systems developed, for example, exclusively for
playing chess, like the Deep Blue computer, for example, in 1997. So this is a big aim of computer science to develop these kinds of systems. But again, the best AI in the world can never do different things at the same
time. In most cases, they are just designed to exclusively do one thing perfectly. So AI is a fluid concept nowadays because it carries the human dream of creating the
perfect machine. And of course, we live in times where data gets more and more important. Data is supposed to be the new oil and data is everywhere. We're creating data with everything we do, using our smartphones, logging into the computer, doing things, shopping, whatever we do creates data.
In order to create a very efficient artificial intelligence, we need to get the data first. That's the first step of developing a good system. The second step is always the cleaning process, preparing and manipulating the
data. Then, based on that, the computer scientists have to train a certain model. For example, the system has to learn how to play chess, or the system has to learn to recognize faces. Then, the data has to be tested.
That's the fourth step. And finally, the entire system has to be improved. And the system can, again, go back maybe to step two, again, step one, new data come in, and more data come in, and the system becomes better and better. The model gets trained, and finally, the system can be improved step by step.
So let's have a look at one particular example. Face recognition. This is, of course, something that's very challenging. For human beings, that's something we learn. We're trained to recognize faces.
The more faces we see, the more we experience, the more we just interact with other human beings, the better we can recognize faces, and the more we get familiar with this very challenging task of recognizing faces. Face recognition, on the other hand, done by computers, is, well, one might say the Champions League, maybe, of artificial intelligence, or at least
it's something that's not trivial. It's something that, well, has to be done very carefully. But again, there are certain steps that have to be taken in order to build a system that can recognize faces. One key technology or mechanic or approach here is the so-called deep
learning, or so-called neural network. So very much like in the human brain, computer scientists try to build systems that have certain layers, certain hierarchical features, and that think about connections between different elements and different features
of, for example, a face. So first of all, we would need a database full of faces, all kinds of faces, pictures, photographs. This would be the so-called input layer. That's what's fed into the system. But then we need to train the system, we need to manipulate the data,
we need to clean the data, and we have to tell the system, look, a face has certain components. And of course, first of all, the system needs to be trained in a way that it is able to understand, well, whenever there is a black dot or a white dot or a gray dot in a picture, it might carry certain
meaning whenever these structures are combined in a certain way so that they, well, shape something like a nose or an eye or an eyebrow. So step by step, the system gets trained and it gets more and more specific. And then finally, the system might be able to say, well, let's
look at what it really is. And that's the so-called output layer. So by combining these different pieces of information, the system can finally say, well, this is probably a face. But again, the system might be wrong. So it might say it's a face, but like you can see here
on the right hand side, you might have typical features of a face, like two eyes, a nose, an eyebrow, two ears. But if it's not in the correct position, you have to train the system and say, no, that's not a face, although it has all the features, but it has to be positioned in a particular way so that it is a real face.
And that's what we can do as humans perfectly. But for computers, it's very difficult to do that. AI is also relevant for teachers.
We've created a little video for you that shows certain applications that teachers already use nowadays.
In this video, we had a look at so-called weak slash narrow AIs. These are all artificial intelligence applications that are extremely good at doing one thing, especially in the EFL context.
Speaking of the EFL slash ELT context, there is a phenomenon which is also extremely relevant and fascinating, which is natural language processing. What is natural language processing?
Put into very simple words, NLP makes human language and interaction understandable for computers, machines and algorithms. This, of course, has an enormous potential for language learning and language teaching scenarios.
NLP is a branch of artificial intelligence that filters through a large amount of content and data and extract keywords. So the main task is to find so-called semantic connections
also during the teaching and learning process. In general, it's the goal of NLP and AI to improve tasks like translation or speech analysis. Think of Alexa and Siri. They help you with your spoken input.
There is so-called speech synthesis, speech processing and voice recognition, which might be helpful for your learning processes when you learn a language. We just had a look at these applications as well.
It's used in various fields like manufacturing industry, but also in the military or in the field of school and university and language teaching. Global voice recognition and natural language processing
more or less are projected or forecasted to reach about 32 billion US dollar of market value by the end of 2025. So especially for us as language teachers, we have to realize that AI and NLP have to be included in a methodological way
in our teaching and learning scenarios as well. We have just heard what natural language processing can do for you,
especially in the foreign language teaching and learning context. But artificial intelligence is a matter of everyday life as well. We have to understand that artificial intelligence enters our everyday life. So artificial intelligence technologies have to be considered
a certain kind of holistic approach in our everyday life. A lot of companies have already invested a lot of brainpower in order to improve these technologies. Especially in the field of image recognition, a lot of money is being invested right now.
There are different industries that benefit from AI technologies, but also people benefit from AI image recognition technologies. For example, we do have certain companies that try to analyze
and develop certain visual patterns in order to make our streets and traffic safer. Another very controversial part of AI and image recognition, of course, is surveillance, CCTV and image recognition.
So a lot of data from police departments is used in order to recognize people in the crowds, in order to try to identify certain criminals, for example, in concerts or at sports events. This, of course, has to be discussed also on a very basic level.
What else can AI do for you? We were talking about everyday life situations, and here they are again. AI devices like smartwatches can predict events.
They can actually predict your health problems. If you use your smartwatch or a wristband, it can measure your mood, it can measure your performance, and it can even understand your behavior and your daily routines.
That means artificial intelligence can be used in the medical context, because artificial intelligences can actually warn you right before a heart attack using all your medical and personal data. Of course, this has to be discussed in a controversial way as well.
What is even more promising with AI is that AIs make your bodies better than ever, especially in the field of sports. They can analyze, evaluate all the required data in order to prepare well for your next marathon, for example.
But again, the big philosophical question here is, do we really want to optimize our bodies all the time based on algorithms and artificial intelligence performative data?
AI can also predict who will win elections or not. There are many cases where the algorithms are perfectly alright and work properly, but of course we know from the past that sometimes the AI epically fails when it comes to the prediction of election.
But why is that? The reason is that sometimes certain human events cannot be solely analyzed based on sheer data, numbers and figures. And that's why probably the role of the human slash language teacher
might be of great importance here. What else can AI do for you? Let's talk about a classic, which is speech recognition. Speech recognition is important in everyday life again.
Most people won't be able to type in a text into their smartphone because they're busy driving or cooking. We can use speech recognition in order to dictate certain texts, which of course is quite useful for various usage scenarios.
You can use these AI powered audio recognition tools like Siri or Alexa or even chatbots in order to use these for your language performance. For example, in order to understand your accent
or in order to formulate coherent sentences while speaking. You can use Siri or Alexa by asking a question. For example, how do I fix a broken part from my car? Or how do I get to Paris? And the system immediately provides an answer.
This of course is a big advantage because dictation is faster than typing. Speaking of efficiency, it's also very relevant for the context of digital inclusion
because audio recognition can of course support people with disabilities in order to produce texts. But of course, we also have to talk about the limitations and pitfalls of artificial intelligence technologies.
So let's do a little recap. In general, AIs are used and often created to make things more transparent, effective and fair. But sometimes this is not the case. And when this happens, we often talk about machine bias.
So what does that actually mean? Sometimes algorithms are not perfect. Recent studies show that certain facial recognition programs show a certain bias against people of color.
That means that certain insurance companies do not offer the same products at the same prices for these people. And this of course is extremely problematic. AI creates systems that make decisions without human input.
And that of course seems to be the biggest challenges for machine bias. Many insurers, AIs, found it's more likely to charge someone a higher price or subscription model with a lower socioeconomic background,
which of course is, politically speaking, wrong. So how do we tackle these problems? And here, again, it needs the human input, a human supervision, in order to really get rid of these machine-biased performances
uttered by artificial intelligence mechanisms. So let's now focus on the role of practice in the foreign language classroom and the role of AI in practice phases of foreign language learning contexts.
So back to the roots, some theoretical considerations about practice in foreign language learning and in task-based language learning and teaching in particular. In this example, you can see an overview
of a typical task-based language teaching or learning unit. So we have a complex learning task as an outcome of this entire teaching unit. So based on this TBLT planning scheme by Gerlach et al., the product of the entire teaching unit is the simulation of a world climate summit.
Of course, that's a very complex task because it contains many different elements, many different communicative elements, and of course, lots of elements and things have to be practiced in advance. So in order to get to the point where students can finally do this world climate simulation,
they have to practice and they have to go through different exercises. So in terms of a backwards planning design, one has to think about what would be pre-communicative activities, what would be things that the students have to learn in the classroom,
what are language forms that they have to practice in order to finally successfully complete the target task as a challenge. So in terms of backward planning, practice plays a very important role. So we have to think about what makes practice efficient,
what makes exercises and practice leading to the target task in the end. So in this design scheme, exercises and practice activities are critical elements of good task-based language learning activities. And these activities, focusing on forms, being pre-communicative or communicative,
or focusing on methodology or learning strategies, providing the support that learners need to complete a task in the end, a complex task, focusing on linguistic skills, all competencies might be involved. So practice is very important.
And of course, intelligent practice is even more important. So let's have a look at the theoretical background. Why do we practice in the foreign language classroom and what makes efficient practice? We have to look at some theories in order to understand what makes practice more efficient and what elements of practice are necessary.
First of all, of course, the input hypothesis, you might all be familiar with it, is very important. So comprehensible input is essential. Somewhat more complex input than the learners into language.
But of course, input alone leads only to basic variety. So you have to think about how to design the input and how to make it comprehensible, and how to make it slightly more complex than the learners into language. Then of course, output is very important.
Noticing, metalinguistic reflections, all these elements are crucial for language learning, but we have to create opportunities where students can produce language in meaningful contexts, in challenging contexts. Then of course, we know from Vygotsky's theory, or Lanthöven Appel,
that learning results from interaction in so-called zones of proximal development, where you get support from more capable peers and by the teacher. So, others, peers, help you to become better, to understand the things better, and to really, well, develop in terms of learning.
The teachability hypothesis is also very important here. Foreign language learning follows predictable stages. And we have to keep these stages in mind. So we can only teach something once we can guarantee that this is a logical next step in the learning process.
And we have to refer to these predictable stages. And foreign language learning, of course, also in communicative foreign language learning, needs an explicit focus on form in order to provide opportunities for students to overcome partially existing or missing competences.
And of course, we have to keep the declarative knowledge in mind. Declarative knowledge is an important starting point, but practice is the key ingredient of successful language learning. And we know, for example, from the Hedy study, that feedback is very important.
High quality feedback and scaffolding as two of the most conducive factors of learning. We need to support our learners. We need to help them. We need to provide meaningful, helpful feedback in order to guarantee learning and in order to make learning more effective.
So this is kind of a theoretical, of course, slightly simplified framework for practice in the foreign language classroom. So practice is super relevant. It's one of the most important ingredients of successful language learning.
But practice has to be repetitive. It has to be purposeful. It has to be goal oriented. And of course, we need to think about what are possible functional, transferable, and contextualized, receptive, and productive forms of practice.
And of course, the linguistic phenomenon that we're focusing on is very important here. So linguistic knowledge must be consolidated through practice and then transformed into the linguistic ability in a communicative setting. So the overall goal of language teaching is, of course,
preparing our students for communication, communicative skills. And frequent repetition is important. Automation is important. Awareness raising is important. So we have to show our students and make our students deal with those phenomena.
But practice is also an unloved and often, well, necessary evil. It depends on how we practice. Of course, we all know these drill and kill activities. Very behavioristic exercises with lots of repetitions. But of course, there are better ways of practicing.
Communicative activities. Activities where students practice the language phenomenon in various ways, and in very creative ways, where higher order thinking skills are activated in the exercises and through the exercises.
So if we think about the how of practice in EFL learning and intelligent practice, we think that intelligent practice mainly focuses on the needs of the learner in order to communicate, to communicate fluently and freely. Intelligent practice has to be authentic or at least very realistic.
Intelligent practice also cultivates autonomy. It has to be scaffolded. So think about various learners with their various problems in terms of language learning or different levels or interests or, well, the kinds of practice, the circumstances, the context of the classroom.
All this plays a very important role. So practice has to be individualized and differentiated depending on the learner's needs. So you have to think about a support system and a scaffolding system and a feedback system that ideally supports learners individually in the practice phase.
So adaptation is the key word here. Adaptive language practice and goal orientation. Well, you have to think about the target task, about the communicative goal, about the goal of the entire teaching unit and beyond. What do you practice and why do you practice these things?
So using the workbook should not be the goal of a language lesson, but using intelligent practice opportunities and exercises in order to provide learning opportunities for your students to communicate in various settings in everyday life.
That should be the goal of language teaching. So teachers play a crucial role here in order to design these exercises, in order to support the students in the learning process, in order to provide scaffolding and feedback. And that's where technology comes into play.
If you look at this example here for a second, you'll realize that this is a typical workbook activity. So imagine a setting where 25 students fill out this activity as a homework exercise. So you would have 25 students with all kinds of problems filling it out.
Of course, using a computer system could really help you as a teacher understanding what kinds of mistakes your students make, what kinds of problems they have, what kind of feedback they need. And of course, the system, if it's built in a very smart way,
could also find the exercises that ideally suit the needs of the learner. So instead of a workbook-oriented approach with a one-size-fits-all approach to learning and practicing, intelligent systems and AI could make a great contribution to individualize
learning and practice phases in the foreign language classroom. So when we look at modern language learning software, the question is, is it smart already?
Can it provide this kind of intelligent practice that we want to have? Let's look at some examples of programs or apps that are widely spread and ask ourselves the question, is it already smart and can we call it intelligent practice?
So you might all be familiar with apps like Duolingo where you have to, of course, go through many units. You have many fill-in-the-blank activities. You have activities where you learn writing, vocabulary, grammar. The question again is, how intelligent is it?
Can this software provide intelligent feedback? Can this software provide scaffolding? Is it an individualized learning path or is it very linear? So that's something that we as teachers and teacher trainers always have to ask ourselves, how good is it already?
And of course, when we look at one study that we did in Lunabook a couple of years ago, we had a look at 50 different programs, 50 different language learning apps, and we analyzed these different apps. We looked at the background and the context, so who's the developer,
when was it developed, by whom, for which languages. We also did a pedagogical analysis, so what's the language theory behind it? And that's not always that easy to judge or to say. We also looked at the game design because many of these apps have lots of gamification elements in it or they are gamified in a way that learning is supposed to become easier
and they have this so-called chocolate-covered-procoli effect where, you know, grammar or vocabulary activities are just, you know, hidden underneath a layer of something that's a little bit more gamified and then maybe more fun. We also looked at the usability, at the multimedia design,
and we also looked at different studies published. Well, again, the question is how intelligent is the EFL software? We found out that amongst those 50 programs, 42 focus on vocabulary, 27 on grammar, 31 on listening,
but they very rarely provide activities focusing on more productive skills like speaking or mediation or language learning awareness activities. So we have to say that these programs are mostly designed in a way that
some of the workbook activity kind of activities are transferred into the digital world. And again, we have this chocolate cover, but it's not really something that we can call intelligent or something that really uses the possibilities of an intelligent system in a good
way. In addition to the relative lack of meaningful feedback, for example, and scaffolding or individualization of the exercises provided in these programs, we found that the majority
of these applications provide little or almost no individualization and a very linear learning path. So while digitization is often linked to the promise of individualized learning support, current tools in foreign language learning mostly fall short of delivering on this promise.
We have just heard so many essential theories behind the actual use of artificial intelligence. Now let's talk about practical implications of so-called narrow slash or question mark
intelligent AI tools. Let's start with chapter 4.1. Let's talk about AI and translation.
Teachers love these tools and teachers hate these tools. It's more or less the Marmite of AI technology. Let's get down to business and now let's talk about Deep L and its underground question mark. Deep L is the superstar among AI powered technologies.
So what Deep L does is actually it tries to translate language. But that's not everything. The AI behind it tries to provide solid semantically coherent
translations for the language learner as well. So it's not only a plain word by word translation, it also tries to analyse grammar structure and it tries to translate
idiomatic expressions. And this is what it's all about in language learning, not only a plain word by word translation procedure. One big advantage of Deep L is its diversity of lexical resources and corpora. So there is a massive linguistic corpora layered behind the whole
system. And this is what it's all about. Computer linguists work with this tool in order to provide solid and pragmatically coherent translations. However, sometimes it
can be a bit difficult to differentiate between similar meaning passages due to pragmatic inaccuracy the system offers. So that means sometimes the system is not 100% perfect. But who is actually? So sometimes the system slash the machine has problems with
jokes or idioms or culturally specific expressions. And that's why the role of the interculturally aware teacher, language teacher comes into play massively. So the
big question actually is, is Deep L a tool for the underground or for the modern ELT classroom? In the following video, you will see the benefits, potentials of this very
technology. Let's talk about another AI powered tool which can be of great use for your
teaching for the EFL classroom. The tool is called My Simple Show. My Simple Show is an application that automatically generates explainer videos without the tech hassle. So what does that actually mean? You as a language learner or teacher, you type in a
specific text into the machine, into the software and My Simple Show automatically generates an explainer video with animated visuals and movements. We all know from
language learning theory that of course that multi-channel language learning that means the use of visual prompts of course supports the whole process. And the cool thing about this tool is you don't have to fool around with tech and so on and so forth. The thing is the system does the work for you. You as a language teacher or learner, you have to provide the
lexical linguistic, the language input. And this is what language learning is all about when it comes to AI powered tools. You actively create and produce language. The machine just provides a certain kind of output which might be helpful for your ways
of understanding a certain topic. So you can use My Simple Show in order to practice pros and cons concerning a certain topic. For example, the pros and cons of life in the city are typical, are classic in the EFL classroom. And what the students can do is they can actually
use the prompts and phrases and lines of argumentation they learned in the EFL classroom. They actively use it and type it in into My Simple Show and My Simple Show creates a video which nicely shows the pros and cons of life in the city using very appealing images
that are based on a Creative Commons license. So My Simple Show offers an automated dubbing system. So the learner types in the text, the system provides and produces the voices.
You have automated English voice overs which sound quite native-like. But of course, one has to be fair, such tools also have certain kind of limitations when it comes to the
are not 100% native-like in other languages. Which of course brings up a very interesting philosophical question. How much native-like should a machine be like? This is what you can probably discuss in class. Furthermore, sometimes the system chooses a certain kind of
biased image. Which means if you type in teacher classroom, for example, My Simple Show actually provides an image of a female teacher wearing a short skirt with glasses on,
having a showing stick standing in front of the blackboard. Which is actually not a contemporary approach of how to use images using technologies. But this limitation of course is a big challenge for your language learning and teaching scenarios. Because now you can discuss with
your students why this image the machine provided is possibly problematic. And here again, we can kill two birds with one stone. We talk about the limitations using the L2, using certain lines and phrases, presenting and underlining your argumentation and point of view.
Let's talk about AI and speaking. We all know and agree that speaking is an extremely important
skill in the EFL classroom. There is a tool which possibly helps your learners improve their speaking performance, their pronunciation. Otter.ai is an AI powered transcription website that uses speech recognition. Otter.ai can be used for ELT purposes, transcribing audio
inputs in order to allow students to rehearse their presentations, their dialogues, their dialogues and so on. Otter.ai can also be used to practice important situations before having
an actual conversation in the English classroom. The cool thing about Otter.ai is that it provides feedback after each recording by tapping the comment button next to the transcription
of the audio. Furthermore, you can practice additional dialogues in the EFL classroom where you take two students who practice a dialogue and Otter.ai records these students. And after this recording, you as a teacher together with the class and the peers can
analyze this dialogue, whether it's coherent, whether the turn takings sequences and scenes are appropriate and so on and so forth. A disadvantage of this tool is that sometimes Otter.ai does not fully recognize accents or dialects or mispronounced words. Sometimes
these words are not transcribed properly. But on the other hand, this is of course a very cool chance for your language learners that you actively discuss these mispronounced
or misspelled words together with the teacher in order to find the right solution. So again, the language learning process is collaborative. Sometimes the system does not detect or always recognize culturally sensitive contexts, which means that sometimes certain jokes,
idioms and passages cannot be embedded properly in turn taking sequences or dialogues, which you can see in the video. Let's talk about AI and chatbots. So first of all,
chatbots are not robots talking to you. But chatbots are very popular because they allow the language learner to practice speaking with an algorithm or a software. So that means
this very learner can interact with a virtual chat partner without the fear of being embarrassed for example. And this is a great potential, especially for the shy or inhibited students,
if they want to practice speaking not always in front of the class but with a virtual chat partner. So you should not underestimate the potentials here. A chatbot can also more or less review
certain grammatical structures and lexical items you have taught before in the classroom. So actually it's a very good way to practice certain grammatical features, certain lexical prompts and certain dialogic sequences. It's a very good tool to do a bit of remedial
drill practice in a very interactive context. But of course there is also a downside of chatbots. The problem sometimes is that chatbots are often not very effective in answering
complicated questions or input. Sometimes the machine is still too stupid to respond coherently to a very complex question for example. Just click on this link and you will
is sophisticated smart or not. Another disadvantage of so-called chatbots is of course that sometimes privacy issues arise since the data that is sent to an external repository
is being used by the company for marketing purposes. And this of course has to be reflected in a very controversial and critical way. One area where AI tools can support learners is
writing. Supporting learners in the writing process from writing short texts in the first or second year of learning a foreign language, to writing more complex texts, to let's say writing essays at university level, or term papers, bachelor theses, master theses.
So for all these areas, AI systems that can support learners have been developed. And all these tools support different areas. So here's a selection of some AI tools in the
world. There are tools which support functions such as creating outline suggestions for your texts or introductory texts. And there are even tools which support plagiarism detection.
So that these tools can change a text that you've written or that you've illegally copied from someone else. And they can turn this text into a text that cannot be found or identified as a
plagiarized text. Of course this is something that we carefully have to look at and be aware of as teachers and teacher trainers. Because of course if the tasks are designed in a way that students can combine for example all these different tools, take a text from a website from
someone else, turn this text automatically into a new text, use an AI tool to create a new structure and to transfer this text that I haven't written into a text that has the typical features of a good academic paper or term paper. And then finally checks that whatever I have produced
cannot be identified as a plagiarized text. Then of course this is something we have to be aware of when it comes to task design, when it comes to maybe integrating and using these tools effectively in the text production process. Because otherwise if these tools are used
and abused in a way that students are supposed to practice their writing skills, but what they actually do is just use the technology to you know find a way of producing texts without showing what they can really do but just you know copying text from someone else. This is of course
something that's not very effective in terms of language learning, it's just something that we have to be aware of. Computer linguists have also developed AI-based technologies or services for
teachers, foreign language teachers or students. So whoever produces a text or assesses the quality of a text in terms of correction or in terms of an analysis of embedded language structures can use a tool like Flare developed by the colleagues at the University of Tübingen. On this website you can just enter search words
like on a typical search website or search engine like Google or Yahoo. Then you can well click on certain language functions or text characteristics or constructions you're looking for
and you can select a certain level of language. So you could say you're looking for a text on climate change, you're looking for a text on climate change on level B1 according to the Common European Framework of Reference for languages and you're looking for certain language structures like auxiliary verbs. And then based on the selection this search engine could then
finally provide a list of texts from the internet but all with the language structures and with the typical features you're looking for and on the language level you've been looking for.
And this is of course very useful for teachers when they want to design activities in a very differentiated way and for example find text suiting the language skills of their learners and if they want to find a text that has the typical language phenomena and linguistic
structures in it that they want to teach. On the other hand a system like Flare can also be used by teachers for correction purposes for text assessment because you can upload either your own text. So for learners it can also be very useful but also teachers can upload the texts of their learners and do an in-depth linguistic analysis of the typical language structures embedded in the
text produced by the students. And this is of course a very helpful resource for teachers who want to understand how their students produce texts and not just based on their you know
impressions based on certain corrections of individual papers but based on a correction done by the system done by the computer and based on a correction of all the papers all the essays or all the written products by all my students. So the data used for this analysis
is of course much broader and can provide lots of information for the teacher who can then use these information to turn that into next steps and good teaching and exercises and follow-up activities. Let's go on to chapter 5, AI and EFL research. We would now like to provide some
insights into some projects we're currently working on, research projects and development projects. One example is the Interact for School project funded by the German Ministry of Education
a project that's been running since March 2022. And this project is an interdisciplinary project because not just different universities are involved but also researchers with different backgrounds, educational scientists, psychologists, computer linguists and of course people with
a language pedagogy and teaching English as a foreign language background. In this project we designed four new complete task cycles. So task orientation plays a very important role here but we also designed in terms of backward planning all the exercises
and pre-communicative activities leading to the final tasks. And what we did in this project is we replaced the typical workbook based exercises into computer-based exercises using a system called FeedBook. So all the exercises provide feedback, scaffolding,
very detailed analyses of the learner's mistakes. We have teacher dashboards, learner dashboards, so teachers and learners get an overview of their learning progress,
what they can already do and they also get an idea of what they have to work on. So in terms of transparency, in terms of making the learners understand what structures they are working on, why they need these structures and what they can already do well and which aspects they have to focus on is of course something that we are very interested in. So in this research project
and approximately 40 different teachers in three different federal states are involved in this project with their classes and the students, we're interested in learning efficacy. So what are the effects of using such a system that can individualize the practice phases and the
exercises? What are the effects in terms of learning outcomes? And one important research question here is as well, can this system and can these digitally supported exercises contribute something to...
to a task-based language classroom in terms of preparing our students better and more effectively for the communicative goal of a teaching unit. Another research and development project is called CHAT clause. This AI system focuses on the development
of the students' speaking skills. Based on different AI mechanics and the analyses of learner products, of recordings, of their recorded answers, the system can analyze how they speak, can provide feedback, provide scaffolding,
and of course we also have elements like peer feedback and the teacher can give feedback. So the entire system is designed in a way that it wants to increase the speaking time and provide many opportunities for language production in meaningful tasks and in meaningful speaking activities.
This program is accompanying a textbook. So all the activities, all the speaking activities, are linked to the content of the textbook. So again, the question is, is this leading in the direction where we can speak of intelligent software
and intelligent language practice? Intelligence in this case means that the system uses speech AIs, dialogue systems, content AI, assessment AI. So for all these different processes, different AI components are used.
And by providing all these overviews for the learners, analytics, learning analytics, the students and the teacher get an idea of how their learning went, which progresses they made, what they have to work on. And finally, again, the question is, how can such a system be embedded
in a communicative language classroom? Because it's not supposed to replace the teacher or replace all the other great face-to-face activities in the classroom. It's just supposed to provide a learning context that's better than the typical workbook-based,
linear exercise focused process with learners working in a very linear way on certain activities, not getting enough feedback, not getting enough scaffolding. So maybe that could make learning a little bit more efficient.
Another project is iVito. It's a research project carried out by the Center of Learning Technologies and Innovation at the Vienna University College of Teacher Education. And we want to find out to what extent narrow visualizer AI tools can support certain grammatical performance
with EFL learners. So the iVito study tries to find out the impact of these visualizer tools. So we want to find out how AI algorithms providing certain multi-channel visual input
can provide a certain kind of increase concerning the use of grammar. So we want to, in a mixed-method research approach, we want to find out how 11 to 14-year-old kids learn with classic analog media
and how they learn with AI-powered tool. So here's the deal. One group gets the AI-supported tools and the other group works with textbooks and handouts and the teacher. In pre-post and delayed-post tests,
we want to see if there is a difference in certain grammar skills between these two groups using validation strategies such as data triangulation and guided interviews with these teachers. So what we want to do with this project
is to focus on the potentials of these narrow AI multi-channel multimedia tools that use algorithmic intelligence to enable a certain kind of immersive
and ubiquitous language learning experience. So finally, let's talk about the future directions concerning technology-enhanced language learning and teaching, especially with AI-powered tools.
So first of all, it's very important to emphasize that it is not a question of AI replacing the teacher. For us, it's extremely important, especially in the communicative EFL classroom,
that the teacher still acts as an interculturally aware and pragmatically proficient facilitator with a solid expertise. The differentiation between AI tools and the EFL teacher
is rather a matter of a certain kind of division of labor. And on this slide, you can see the various things. The teacher is mainly responsible for a one-on-one communication in the classroom. It's about empathy, it's about human interaction,
it's about planning and designing the lesson, it's about including research into lessons, and it's about creative projects. But of course, AI can also support this language learning and teaching scenario.
AI, as we heard before, delivers adaptive, personalized content. It analyzes the learner's learning data and needs. Furthermore, it gives automatic, scalable feedback prompts.
It does simple auto-correcting, and it does the grading of texts. Combining this multitude of various tasks makes sense that we can coherently and sustainably use AI technology together with the interculturally aware
teacher with solid pragmatic slash linguistic skills. At the end of every presentation, there has to be a conclusion.
We decided to have no conclusion, but present a utopia, the foreign language classroom in the year 2030. In 2030, the foreign language classroom will be an evidence-based, planned learning environment
with bring-your-own-device solutions in all classrooms, high-speed internet connections, as well as standardized content and learning management systems. In 2030, the printed textbook has become an interactive, multimedia, adaptive learning and practice environment that is perfectly adapted
to the needs both of face-to-face teaching and the faces of individual practice and self-directed learning. Experts in the field of foreign language pedagogy, computational linguistics, learning psychology, machine learning, and big data, as well as the best multimedia designers
have worked closely together to create this digital learning environment. The digitally enhanced classroom of the year 2030 helps to diagnose needs and learning progress and provides direct access to differentiated and needs-based support services.
The learning platform deployed in 2030 is thus a digital learning support system, a resource, a tool for students and teachers, which is used precisely to add value to any learning process. The classroom of the future in the year 2030
will thus cleverly combine the advantages of digital learning with proven computer-free methods, content and tasks for face-to-face teaching, which will remain indispensable and highly significant for successful learning. And even in 2030, we will still have and need teachers,
well-trained teachers, data-literate teachers who are competent, critical, reflective, reflective especially in the use of media and technology support, and who use challenging learning tasks and exercise opportunities designed
to support individual language learning. Thank you very much.