How to embrace AI chatbots in schools
Appearance of ChatGPT has raised concerns about the potential abuse of chatbots in schools. Unsurprisingly, teachers are more concerned with endangerment of traditional assessment practice than with chatbot’s devastating potential of blurring the line between real and fictional. Worried teachers are mostly afraid that chatbots would destroy the practical value of school essays. Truth is that essays are not endangered, but only grading of essays. I’m wondering shouldn’t we be more amazed with opportunities for learning than with inconveniences for traditional school grading?
A big problem with education in today’s hectic world is that students don’t get important and meaningful feedback to what they said, wrote, or did. The school grade, especially summative, is poor feedback. What students need is someone to tell them where they go wrong, what to pay attention to, what they wrote well, or what could be even better. Someone or something, whatever. Teachers, generally, do not have time for deeper analysis and giving well-thought-out feedback to each student. They should, of course, but the reality is a bit different. Chatbots, however, are there for each student, just to interact one-on-one. Suddenly, we have something that pays attention to what students express or ask, an artificial learning companion who has more time and patience than teachers. Also, it is available to all (internet connected) students, so the education becomes more equitable. It’s great! If we don’t maximally exploit this feature, I don’t know why do we need artificial intelligence at all.
Chatbots and school grading
Every teacher has his own way of grading. Although there are standards, regulations and unwritten rules, there is no domain in which teacher autonomy is greater. They have their own habits and rules of what is valued and how it is valued. These habits are hard to change.
ChatGPT, as a precursor to similar and more advanced GPT tools that inevitably will come, seems to be a disruptor of assessments based on essays and composition writing, especially when written away from the teachers’ sight. Should we ban the use of this tool and hope that the dam will hold until we come up with a better solution? It seems that in some schools they have already started with bans, while universities have started to warn lecturers to review their assignments.
Teachers fight their battle against plagiarism and ghost-written essays and term papers for a long time. But artificially intelligent chatbots threaten to make this problem much harder. Generally, we have two types of written assignments that are challenged with unethical use of AI chatbots: one that requires hard work and needs to be graded (like term papers or theses) and the other that shouldn’t be graded at all (e.g., homework). For the first type, solution is much more difficult. You have to prevent plagiarism there, but it will be increasingly difficult. However, there are already some new solutions for detecting AI created text, like GPTZero tool created by a student. It’s a good sign. Similarity can be detected. On the other hand, service providers also think about opportunities to label AI generated text. Researchers from OpenAI recently revealed that they work on “watermark” for GPT-generated text. Probably, there will be ways to tell apart what is genuinely human writing from the text created by machines, but the cost of reviewing won’t be insignificant.
On the ground of everyday homework assignments, solution might be much simpler: just don’t grade essays assigned for homework! The purpose of homework is to deepen and reinforce learning, not to be part of a summative assessment. If we don’t want to invest time and effort in detecting traces of plagiarism instead of assessing learning itself, we shouldn’t put unnecessary burden to students’ homework. It’s not the point.
What could we do?
1. Use chatbots as writing tools. Chatbot is a kind of very advanced auto-predict. It can help students to overcome initial writer’s blockade or to provide feedback on their writing. The chatbot should always be patient and supportive. At least, it can help students to find main talking points and to create a storyline.
2. Use chatbots to practice chatting in foreign language. Since chatbots knows many different languages, it can be used to help students practice their conversational skills in a foreign language. It is very likely that chatbot will use authentic constructions and phrases. Chatbots learned language from a huge corpus of such material. Just pay attention and learn from these unusually eloquent apps.
3. Show students how chatbots work. Give them gentle introduction to terms like NLP, LLM and GPT. Demystify them. Invite them to discover what it is and to tell the others what they learnt. Try to motivate them to create their own tools. For instance, students can tokenize some newspaper article and summarize it as a word cloud. It’s the beginning.
4. Use chatbots as learning tutors. Chatbots really don’t understand anything about any topic, but they can create plausible sentences like if they do. Even if chatbot know less than students, it still can encourage their learning. Just like famous Sugata Mitra’s “grannies” once did. Students generally feel uncomfortable when they need to ask something they didn’t understand. In conversation with a machine, students shouldn’t be shy and hesitant to ask. It is easier for students to ask “stupid questions” to a machine than to a human.
5. Show students how dangerous could be misuse of chatbots. Today’s students know very well how many trolls there are and what they do. They also know how dangerous it is for someone to steal their virtual identity. What they probably don’t know yet is how effectively chatbots can be misused for such purposes. Adequate examples surely already exist on YouTube. Just show them. Teach them how to be smart users of AI and why is uncritical usage dangerous. Talk about it. It’s extremely important.
6. Use chatbots to inspire students to write fiction. It takes skill, inspiration and courage to write fiction. Not to mention poetry. Chatbots do it with ease. Since the chatbots, especially ChatGPT, don’t care much about reality, they can create nice and stimulating environment for imaginary conversations between a student and a machine. It doesn’t take courage.
7. Teach students why is it important and how to use references. Chatbot models don’t cite sources for their ideas and claims, but we should. Extraordinary statements always deserve explanation or a reference. Teach your students how to refer to the original and trustworthy sources and how to cite intellectual contributions of other people. When they start using the references properly, their writing will be more credible than text produced by chatbots.
8. Use chatbots to teach students how to ask good question. How many opportunities do students have in school to practice asking questions? Not many. Usually, they practice answering. If we intend to teach them how to think critically, questioning is the most important. Why don’t we ask students to make a research interview with a chatbot in order to find out and present something. Why not, tell them to try to corner the chatbot and find out which question it can’t credibly and consistently answer. It would be real triumph of human intelligence.
Whatever we do with chatbots in school, clearly communicate to students the expectations and guidelines for their assignments, including the importance of citing sources and avoiding plagiarism. In the world they will live in, it will be more important to discern and fight unethical use of AI. GPT provides them with excellent platform to develop such competences.
Finally, keep in mind that chatbots can be only a learning tool, not a substitute for human instruction. Only teachers can do what is the most important — to provide learning with the context.