About

I'm Mike. I study how to tell whether a language model is actually doing what we want it to — particularly in the learning contexts I know best. Right now that means measuring how close synthetic tutoring conversations come to real ones, and showing where off-the-shelf similarity metrics can systematically mislead.

I'm a Postdoctoral Research Fellow at the University of Michigan School of Information, working with Kevyn Collins-Thompson. My PhD was at Michigan too, with Deborah Ball. I came to AI evaluation through education research — measurement theory and discourse analysis — rather than the other way around, and my work now sits at that intersection.

Before the PhD I did a BS and MS in mathematics at Cal Poly, taught math enrichment courses for advanced middle and high school students, and served as a Peace Corps volunteer in Botswana.

I live in San Luis Obispo with my wife and our goldendoodle. Outside of research I play guitar and ukulele, make music in the evenings, hike whenever the chance comes up, and play Super Smash Bros. Melee.

Get in touch

I take on a handful of consulting and research-collaboration projects each year. Best entry point is email: mikeion@umich.edu.