I’ve been exploring the world of chatbots and tinkering with AI-powered chatbots for language learners for some time now. The first chatbot I made was designed as a Language Teacher Assistant, tasked with providing feedback to learners and transforming the checking process into a learning experience. It was incorporated into the Question Charts. I’ve also built a few chatbots to assist teachers in generating assessment rubrics, role-play scenarios, and project ideas (though they’re still in the process of fine-tuning; alas, time is a scarce resource). In this post, however, I’ll be focusing on chatbots specifically tailored for learners, along with tips and shortcuts that can help you experiment and determine if they might have a place in your learners’ journeys.
So, what’s a chatbot?
A chatbot is a digital buddy that ‘understands’ user messages and responds in a way that feels just like talking to another person. But would my learners need a digital buddy?
Step 1: Why should I make a chatbot?
And that’s exactly the question you should answer before anything else. If you really want to explore the opportunities this technology offers and see if it can add value to your language classes, the first step is pinpointing a specific need your learners have. Consider a practical challenge that you can solve by crafting your own chatbot. This approach will help you select the right platform with the appropriate functionality and set up the bot to benefit your learners.
The whys can be various: some learners may need extra practice to develop a particular skill, while others might just need a confidence boost. And then there are those who could really use some feedback on their writing or speaking.
Today, I’d like to introduce three AI bots I designed to address such particular needs of my learners.
Meet ThinkChat Buddy, Reading Coach and WrAIting Buddy. (Click on the images to preview the bots)
These three bots have different objectives and target learners, which define their behaviour and language.
ThinkChat Buddy is designed for upper-intermediate students – young adults who need to practise their conversation skills and be able to act in various situations. I planned it as a ‘chatting’ scaffold, meaning it should guide students through the conversation by thinking out loud, explaining why it says what it says, and clarifying what it expects to hear.
If you’ve ever used conversation scripts with your students, then it’s essentially the same but in a ‘chatting’ format.
After the role-play, ThinkChat Buddy will provide feedback to the student.
Since it is designed for conversational practice for young adults, the bot’s language employs spoken grammar.
Reading Coach is designed for adult learners with a very low level of English. Its main objective is to help them read in English. The learner will paste a text, and then the Coach will select challenging words to practise reading them before the learner reads the entire text. Afterward, the bot will guide the learner through the text and provide, as a follow-up, a new text similar to the uploaded one.
Since it is intended for use by low-level learners, I programmed it to be very brief and to use simple language (getting a chatbot to be concise can be quite challenging!). Additionally, when asked by the user, the bot will also translate its instructions or words into the learner’s language.
WrAIting Assistant, as you may guess by its name, is designed to help students (young adults at university level) write any piece of writing they need. Its objective is to guide students through the writing process, acting as a ‘thinking’ partner.
The bot is based on the Flipped Interaction Pattern, which is commonly used in machine learning. Usually, you ask questions to the language model, and it spits out answers. But with the Flipped Interaction Pattern, it’s the other way around. The chatbot will ask questions one by one, prompting the student to provide thoughtful responses. Once it’s got enough info, it will summarise it into the writing piece needed. As part of the feedback session, the chatbot will also spot and list out any language bits that need some work and generate a follow-up exercise for the student.
This approach helps structure students’ thinking and encourages them to see things from different angles and consider different viewpoints, while the language model gets all the info it needs to give a spot-on answer.
Integrating a chatbot into a lesson flow
A chatbot can be used as a standalone activity or as part of a particular lesson. For example, in this Tumbleweed Invasion lesson plan, students are supposed to chat with Alex the chatbot to find out the details of the story as narrated by Alex. What I particularly like is the structure of the conversation; students are offered a chart to fill in while conversing with the chatbot.
Source: Tumbleweed Invasion
What I’ve observed in student-chatbot conversations is that even higher-level students often struggle to ask and answer extended questions if the task is simply to ‘go and chat’. Therefore, consider scaffolding students’ conversation with a chatbot by either prompting them to think about specific questions or instructing your chatbot to guide students through the conversation.
Now, when you have a clear objective for your chatbot, you can move on to
Step 2. Choose the chatbot building platform
You have quite a number of options depending on the level of your tech skills.
For teachers who are not planning to join the ed-tech developers’ army, I’d suggest beginning with a no-code platform, especially one made with educators in mind. Just make sure the platform checks off those first three crucial points when you’re picking a third-party tool – Keeps Student Info Safe and Secure; Fits Your Pedagogical Approach; and Compatibility and Sustainability (see the decision tree in SHOULD YOU BRING AI INTO YOUR LANGUAGE CLASS? A TEACHER’S DECISION TREE). I’ve chosen Mizou for these three bots. This is a no-code platform which is currently free and designed specifically for educators. So, you can put your focus on teaching rather than getting bogged down in the tech stuff.
For me, what really mattered was also how much control I have over my bots and the ability to share them with my students without them having to set up accounts or log in. Well, the text-to-speech functionality is, no doubt, an added perk.
Setting up a bot is super easy. The platform is intuitive, and it also offers three tutorials on setting up your first bot, starting a new session, and reviewing grading.
Step 3. Craft a prompt
Setting a clear objective for your chatbot will help you tailor its behaviour and language in the corresponding prompt. If you have never used LLMs or AI-powered chatbots (e.g., ChatGPT), start with my easy guide for teachers
IF YOU CAN TEACH TEENAGERS, YOU KNOW HOW TO TALK TO AI
Make sure that your prompt reflects the process flow for your chatbot. Certain things are already included in the system prompt, like being polite and respectful, so you don’t need to spell that out. But you might need to tweak the language to fit the level and style you’re aiming for.
I’ve noticed that setting the Grade may affect the bot’s language. The lower the grade, the simpler the language becomes. Getting it to be less verbose is probably more challenging than otherwise. You can also try a few prompt snippets that I shared in ChatGPT PROMPTS FOR LANGUAGE TEACHERS, or get in touch with me and I’ll try to help.
Step 4. Test drive and tweak, tweak, tweak the prompt
Play with your chatbot and see if it follows the prompt. Tweak the prompt if needed. A good idea is to have fellow teachers or friends test drive the bot before introducing it to your learners. And then tweak the prompt again.
Once the bot is good to go for a trial run, share it with your learners and see if it affects how they learn.
If you don’t notice any difference in learning, take a moment to think about why that might be. Did you maybe make the task too broad and need to narrow it down? Or could there be another reason behind it?
AI is like a toolbox. It’s handy to have, but not every tool is needed for every job. Just because you have a hammer doesn’t mean you need to use it every day, does it? But there may come a time when it’s a better option compared to an old shoe.
Happy experimenting!
Here you can read more about chatbots and their applications:
Creating Language Chatbots to Safely Share with Students from Cristina Cabal
I CREATED ANOTHER CHATBOT FOR ELL TEACHERS & IT SEEMS TO BE RELATIVELY DECENT? from Larry Ferlazzo
