Thinking with Colleagues: AI in Education

• 10/09/23

Second, I think there is one particularly beneficial result we'll see as we have more and more AI in our lives. AI is going to give more people opportunities to conduct very complex transactions by guiding them through the processes, procedures, and barriers that may have limited their participation in the past. In other words, making more "smart tools" available to more people is going to help level playing fields.

Third, AI is likely going to require new rules of engagement between and among institutions, colleges, departments, and students in the environments where it is being used. Here's a perfect example of a generative AI problem that WCET found when they surveyed their members this past summer: Many of their member institutions did not have consistent policies on their campuses when it came to whether using ChatGPT and similar tools was the same as cheating. Even in places where institutions did encourage ChatGPT use, some departments would overrule the central administration. Institutional inconsistencies like that have the potential of creating some real challenges for students to navigate.

And my fourth pick from a long list of things many people in higher education cite as their top-of-mind AI considerations: We are going to need to revisit many of the ways that student assessment is being conducted. I have been really excited by many of the ideas I am seeing from people like Ethan Mollick and Ryan Baker from UPenn, and David Wiley from Lumen Learning, who are showing us ways to learn more, explore more, and try some of these innovations for ourselves. We can't keep holding on to traditional assessments. I have to say, I can't stop wondering what is going to happen to the term paper!

Grush: Will the thinking about generative AI tools ever be defined by formal research results? Is it science?

Wagner: This is all driven by science! And a lot of scientific work over the long term. People have been working with neurolinguistic programming and large language models for more than 30 years. Similarly, consider the long course of development of GPS systems. And recommendation engines. And robotics. And much more.

This is all driven by science! And a lot of scientific work over the long term.

A "fun" example in the robotics realm is the YouTube videos everyone loves of robots dancing in unison. You should take a look at some of the early dancing robot videos — the robots would take a single step and fall over. But the scientists and engineers kept at it, and kept at it, until you can now see something like this demonstration video.

I think that one of the most exciting things about the new generative AI tools is that we mere mortals now have the opportunity to experience some of what our best scientists have been doing for decades.

We are in the very early days of seeing how AI is going to affect education. Some of us are going to need to stay focused on the basic research to test hypotheses. Others are going to dive into laboratory "sandboxes" to see if we can build some new applications and tools for ourselves. Still others will continue to scan newsletters like ProductHunt every day to see what kinds of things people are working on. It's going to be hard to keep up, to filter out the noise on our own. That's one reason why thinking with colleagues is so very important.