“…I genuinely don't think that take-home work is a reliable way of assessing students anymore. I think it just isn't.” Teddy Svoronos · Mac Power Users
September 2026
Asked what he would tell teachers struggling with AI, Svoronos led with the thing he expected people to dislike hearing. His reasoning is that process and output have come apart, so a paper written at home no longer reveals how the student thought. He is not calling for the end of homework, he wants it treated as preparation for assessment that happens in the room.
Transcript
I do. So, I mean, a lot of it is what you just said, which is that once you have an initial idea and research plan, getting feedback on it can be really helpful. But I think it's important to treat that as distinctive things. Like you come up with the plan, you come up with your idea, and then it's like a teacher or a friend giving you feedback as opposed to co-creating it, quote unquote, with AI. I think that's really important. Similarly, with statistical analysis, like a real challenge that students have is I think the two big things that are going to matter is domain knowledge and like judgment and skills with AI. Those are like the two things that are really going to matter going forward. And the problem is one kind of comes at the cost of the other. The more you use AI, I use judgment less than I used to. Like I think it like starts to kind of erode at the other thing. And so in my courses, I'm treating them as like very kind of like stonewalled. There's the first three quarters of the semester that's really focused on developing statistical judgment. And then the last part is focused on, okay, now that you have this judgment, let's see what it looks like to apply it when an agent can do a zillion things for you at once. And if you don't separate it that way, I think it really kind of gets in the way of the habitual development that we get as experts over time.
Well, Teddy, we've got a lot of professors and K through 12 teachers in the audience, a lot of educators in our audience. What's some good general advice you would give people listening that are struggling with this right now in education?
So a view that I have, and I think this is not a great thing to hear, and people might disagree, is I. I genuinely don't think that take-home work is a reliable way of assessing students anymore. I think it just isn't. I think the idea that you go home and write a paper and then I grade that paper is, and therefore I have a sense of like what your thought process was isn't true anymore because the process and the output are now decoupled. It's as easy to make an output that was sort of devoid of process as it is to make one that has process really baked into it. I don't think that means that take-home work is over. I think it means that take-home work is basically preparation for in-person stuff now. So, for example, I still do have a research project in my course, but half the grade is a thing where I sit with the group that wrote that paper and I just kind of ask them questions about their analysis. And I go one by one. Any person can be asked a question. Their collective grade is how well all of them respond to those questions.
Back to Socrates.
Right. So the idea is like, sure, you could like basically do whatever you want to make the report, but if you can't stand behind it, then it doesn't, you're not going to get a good grade. And so I think the more we think of homework as like prep work as opposed to assessment work, I think that's an important shift that is going to happen and needs to happen if we want to like keep treating the ability to assess students as something that we value.
And you had mentioned one kind of assignment you do is where you will even have a student talk to like a chat bot and then you grade the transcript. What is that like?
More from Mac Power Users
