How close are we to automating education?
With the public release of AI programs such as ChatGPT, many are asking what impact AI will have on the future of education. Much of this hype is justified. While still imperfect, ChatGPT is well on its way to being able to write college-level essays on any given topic. Yet, in other ways, the hype surrounding AI technology is somewhat premature. For example, claims that robots will soon replace educators or that AI systems can compete with 1:1 tutoring remain squarely in the realm of science fiction. I am not an engineer and do not claim to be an expert in AI. However, I’d like to share my thoughts on what it will take for AI to truly disrupt and improve education based on my experience as an educator and studying the process of teaching and learning.
Many think about teaching and learning as the one-way transmission of knowledge from an expert to a learner. Oxford language describes teaching as “showing or explaining to (someone) how to do something.” However, I believe this is incorrect. Teaching and learning is a fundamentally two-way process. Students do not learn well from having their teachers simply talk at them. Learning occurs when teachers and students engage in back-and-forth discourse. Students must respond to teachers, and teachers must respond to students.
To illustrate this process, let’s look at this example of real student work, focusing on question 2 (which the student did not answer correctly) and explore what a teacher would do to promote learning.
A teacher’s first job is to deduce why the student answered the question incorrectly — to uncover where the student got off track. Based on the scratch work here, we can see the error was caused by having too many Chocolate Chip Cookie Dough tubs. For some reason, the student calculated the total for five tubs instead of four. As educators, the next question we ask is, “why”? Was this a careless mistake, or is the student struggling to understand the difference between “groups” and “quantity per group”? To figure this out, we’d likely need to ask the student to explain the thinking. We might ask the student, “How did you decide what to add?” or “How did you know how many fives to add?”. Based on their responses, we’d probe their thinking and challenge them to defend their ideas. Over a short period of time, we’d coach the student to realize their mistake and correct their thinking. This type of interaction is what teaching and learning is all about. This back-and-forth interaction teaches students how to solve this problem and develops their listening skills (they have to engage and respond to what you say) and their communication skills (they must verbalize and explain their thinking).
The question I have been pondering is: What will it take for AI to be able to replicate the interaction described above? First, we need to develop the capability for machines to recognize and parse student work. It’s easy for us humans to spot that the above student went off track by having too many 5s. However, this is a complicated process for a computer. I have yet to see any programs capable of making these conclusions. Secondly, AI must be able to engage in two-way dialogue with the student to coach them toward the correct answer. Generating this type of language is distinct from what ChatGPT is capable of. Today, ChatGPT can answer many questions but is not very good at asking questions and listening. When I try to converse with ChatGPT, it doesn’t let me speak much. It gives me answers but is not a good asker. After a few prompts, the conversation ended, with ChatGPT not asking for more input.
I believe that both of these challenges are solvable with time and effort. There is no reason to think that a computer cannot parse student scratch work or that generative AI cannot be adapted to ask more open-ended questions. I am optimistic that these challenges will be tackled and that AI can increase access to high-quality teaching and learning experiences. However, we are not there yet. I’m excited to see what the future holds! What are your thoughts — how do you see AI shaping education in the next 5–10 years?
More content at PlainEnglish.io. Sign up for our free weekly newsletter. Follow us on Twitter, LinkedIn, YouTube, and Discord.
Interested in scaling your software startup? Check out Circuit.
