Building the Future of AIET IN STEM: A Working Group Discussion

Participant: "It seems to me that this group has the potential to do something bigger. If it's not developed yet, why couldn't this be a working group to put out that product for a global basis? Here's a framework, here's some rules, here's some guidelines about how we should be thinking about and using AI in education."

Knut Neumann: "Don't we already have enough journals? There's AI&Ed, Journal of Learning Analytics, JRST, and all these journals. The other thing that typically is underestimated is how long it takes to establish a journal, partly because it takes you forever to get into the social scientific citation index."

Chris Quintana: "I wouldn't mind thinking about some sort of dissemination vehicle for things - use cases from teachers, practical things that show here's what people are trying to do with AI. Point counterpoint discussions. Something different than yet another journal."

Yasemin Copur-Gencturk: "I really like the idea of we call it in NCTM 'linking research to practice.' It's a three-day conference, the third day always there is a connection to the practice. Teachers, practitioners join, and there are learning opportunities for teachers. I don't think there is a conference for STEM education."

Janice Gobert: "Companies are not thinking about that. They have no ethics, they have no IRB. Their AI is unsupervised - it could take in zip code to determine edge cases, which in the United States is a proxy for socioeconomic status."

Participant: "Maybe that's exactly why it would be good to bring the companies in - to hear those struggles and see what the risks are."

Yasemin Copur-Gencturk: "If we are selective, then it could be privileged for them to be part of this. It signals something. If we create a very good vetting system - who can come, who can present - it creates value. The standing out part: we are data-driven, we are theory-driven."

Janice Gobert: "I go to AI&Ed pretty regularly. It's extremely dominated by Carnegie Mellon. I wouldn't say it's theoretically impoverished, but it's not theoretically rich. They have a very coarse-grained idea about learning. They think learning science is like learning math is like learning geography is like learning French. Context isn't that important to them. That's where there could be room for bringing together rich theoretical, practical, and computational work."

Umesh Ramnarain: "We are not looking for handouts. We want to be part of the process. We want to contribute. The unfortunate reputation is that Africans are excellent consumers but not developers. We're trying to change that narrative. We work with a very young team of researchers - if there are opportunities to be pulled into development teams and mentored by senior developers, that would make an incredible contribution."

Knut Neumann: "The Global South has something to share that the Global North can learn from. Germany could learn from South Africa about establishing equity with AI. And regarding representation - it's utterly important. If we're thinking about this conference being global, it can't just be a Northern Hemisphere conference."

Jiliang Tang: "Before 2015, the facial recognition performance difference for white men and black females was 33%. It's because black females joined the efforts to change the situation. This problem is partly from training data imbalance, but at the algorithm level, we can solve this. That's why it's very important to join the global efforts - if you're left behind, nobody would even find this problem."