Sep 30, 2026 · 32 min · 8 segments
Your child wants a new game, and the app store says it's educational. That label is checked for age, not for learning. In this episode I sit down with Dr. Erik Harpstead, Associate Research Professor…
Erik HarpsteadGuest
Patricia Cangas RumeuHostWhat do you study, and what do you study specifically on educational games?

So my background is in human-computer interaction, uh, broadly, but specifically within that understanding, uh, game-based learning and how people learn from play and learn within play.

So how could we, using log data from players interacting with game environments, understand that there is evidence that they are learning something or maybe they're not, not learning something? So, um, and, and my interest has, has generally been in the space of helping designers of educational games understand essentially the things they're making, and is it actually doing what they want it to, and can we highlight places where it's not so that they know how to update or change or improve things.

So, um, one of the techniques that I've worked on and published on is actually kinda importing a technique from the, the field of educational data mining, which is, which is more broadly, like, how do we understand learner data in, in all kinds of technologies, where we look at what we call learning curve analysis.

So there are ways of taking player data and essentially annotating within a game or within a, a other kind of instructional environment what kinds of skills, what kinds of knowledge do we think pieces of the game leverage.

And then looking at a trace of a player going through that environment and seeing, it's like, well, if these are the knowledge involved and this is how they're improving or not within the game, then they're probably learning this set of skills, for example, and you can plot that over time.

So essentially, over your opportunities to practice something, are they reducing an error or improving in success? It's kind of the same thing.

And you can actually empirically plot the curve of learning against a particular set of skills.

But then there's some interesting nuances in how the statistical models work, where if you notice a distinction between what the curve should look like and what players actually seem to be doing, that helps you highlight places where, oh, maybe the skills that we're understanding about our game are not correct or they're different, and that students maybe are learning something else.

So in a really concrete example of that in the particular game that I'm thinking about here is there's this balance beam game for roughly third-ish grade, and we had a few different...

Essentially what would happen is bugs would fall on one side of a balance beam in, in a certain number or in certain positions, and then you had to grow flowers on the other side, and the goal was to get the beam to balance.

And we had a whole different range of challenge levels there, where, like, maybe the design was bugs fell on, like, on position two, and then you had to grow a flower in position two.

Or maybe you had to do some of the algebra, essentially, um, of do I count things up and figure out positions, and then that tells me what to do on the other side, where I have to come up with something that looks different from the other side.

And we found that when we plotted those out, things like mirroring are much easier.
What do you study, and what do you study specifically on educational games?

So my background is in human-computer interaction, uh, broadly, but specifically within that understanding, uh, game-based learning and how people learn from play and learn within play.

So how could we, using log data from players interacting with game environments, understand that there is evidence that they are learning something or maybe they're not, not learning something? So, um, and, and my interest has, has generally been in the space of helping designers of educational games understand essentially the things they're making, and is it actually doing what they want it to, and can we highlight places where it's not so that they know how to update or change or improve things.

So, um, one of the techniques that I've worked on and published on is actually kinda importing a technique from the, the field of educational data mining, which is, which is more broadly, like, how do we understand learner data in, in all kinds of technologies, where we look at what we call learning curve analysis.

So there are ways of taking player data and essentially annotating within a game or within a, a other kind of instructional environment what kinds of skills, what kinds of knowledge do we think pieces of the game leverage.

And then looking at a trace of a player going through that environment and seeing, it's like, well, if these are the knowledge involved and this is how they're improving or not within the game, then they're probably learning this set of skills, for example, and you can plot that over time.

So essentially, over your opportunities to practice something, are they reducing an error or improving in success? It's kind of the same thing.

And you can actually empirically plot the curve of learning against a particular set of skills.

But then there's some interesting nuances in how the statistical models work, where if you notice a distinction between what the curve should look like and what players actually seem to be doing, that helps you highlight places where, oh, maybe the skills that we're understanding about our game are not correct or they're different, and that students maybe are learning something else.

So in a really concrete example of that in the particular game that I'm thinking about here is there's this balance beam game for roughly third-ish grade, and we had a few different...

Essentially what would happen is bugs would fall on one side of a balance beam in, in a certain number or in certain positions, and then you had to grow flowers on the other side, and the goal was to get the beam to balance.

And we had a whole different range of challenge levels there, where, like, maybe the design was bugs fell on, like, on position two, and then you had to grow a flower in position two.

Or maybe you had to do some of the algebra, essentially, um, of do I count things up and figure out positions, and then that tells me what to do on the other side, where I have to come up with something that looks different from the other side.

And we found that when we plotted those out, things like mirroring are much easier.
The rest of this transcript — segmented and speaker-labeled, so you land on the exact moment something was said
Search every transcript — by keyword, by phrase, or by meaning, across every show Radar indexes
Trends — what is surging across podcasts, measured against its own baseline
Alerts — when a name you follow appears in a newly indexed episode
No account is needed to search Radar.