Today’s Plan:
- Review Business
- Potential Research Projects
- 20 Minutes of Thinking and Talking
- For Next Session
Some Review and Business
This week and next week we will be working on proposals. I will formally introduce the assignment on Wednesday. I wanted to spend today reviewing some previous material and then giving you some time to talk and think.
After what I deemed fair warning time, I deleted the original Further Thinking assignment. If you got a zero, you are welcome to resubmit.
Potential Primary Research Topics
I am asking you to develop a primary research project in which you use AI. Regulation folks, I am still wrestling with how you might use AI. I do think there’s several potential research topics for more traditional research papers. I started working (and wasted 45 minutes today) writing a research paper prompt about this MorePerfectUnion video. It brought me to this Food and Wine article which brought me to this letter from Senators Warren and Casey. And chasing down the inflammatory quote from the MorePerfectUnion video brought me to this NPR article, which offers a pretty balanced view of the issue (and puts that inflammatory quote into a different context).
But I’m not sure there’s an authentic research question here. There’s a question–should we be concerned about EDGE pricing and AI facial recognition in grocery stores. But with an authentic research question, I should be able to see what I can go test or do to help advance the issue. I’m not sure I see that here. I have the Warren & Casey letter and the NPR article. But what do I do next? And how do we avoid the commonplaces?
If anything, I think we can look at the 11 questions Warren and Casey propose and ask if those questions, and the forms of potential regulation they imply, are necessary/wise/productive/dangerous/adjective of choice? Meaning this sets up a potential exploration of forms of regulation we (as a nation) might want to explore. If you want to discuss a particular kind of economic or social regulation, then you need to prove the danger that such a regulation needs to prevent–that’s where research can come in. [And, one last link, the FTC is currently investigating what it terms “surveillance pricing.” I’m sure Noble would have something to say about this.
I’m also thinking about this interview (starts at 56 minutes) with Tristan Harris who gave this (in)famous talk called the AI Dilemma and is the chief officer of the Center for Humane Technology. Which leads to this. Some kind of paper that tries to articulate sensible regulation that also addresses the “roll out” argument (“if we slow down, then we will lose to China [or the terrorists etc etc”). How do we find a path through those commonplace starting points?
Okay, creative and education projects. One dimension of this project is generative: I want you to make something with AI.
How you use it is up to you–do you want to try and create or recreate something? Do you want to see if you can identify its limitations? Do you want to investigate whether humans can identify its use? Do you want to analyze its output? Do you want to use it to create educational materials? Art? To test a software/app?
Whatever you decide to do, you also need a way to analyze or evaluate it. This can be tricky. I think this second component, analysis, is trickier than the first, generation. We will worry about evaluation *after* you’ve pinned down what you want to do/make.
Projects in the works:
Ben, psychology students’ attitudes towards AI integration into therapy (see article)
I was reading Sarah’s paper and she wrote the line: “It is “tempting” because there is no decrease in quality with the use of artificial intelligence, only a dwindling of genuineness and consideration.” I responded:
Hmm. I’m not sure I agree with this. I do think there’s a difference in quality of writing. But I cannot easily disprove this. The evidence for my disagreement comes from reading bad AI generated writing on the internet and in my current research. It lacks voice.
It would be interesting to see how many people share my view. Have, say, 3 humans write a short paragraph about their favorite food or memory. Then have AI generate 3 similar paragraphs. Have research subjects rate them on a few different elements (engagement, feeling, etc). Compare.
The tricky parts here: how do you pick the three humans? We would probably want humans who are great writers or love to write. Who would prompt the AI?
I’ve also got an idea for a project kicking around. Last month as I was prepping for class, I came across a Google Chrome plug-in called Vera: The Truth AI. Let’s take a look. Let’s think about what we could do with this. Let’s think about why this might be dangerous.
25 Minutes of Thinking and Talking
Stages:
- Everyone take 5 minutes to think / write
- Group up and share 10 minutes
- Quick Sweep
For Next Session
A few things:
- Catch up. If you haven’t yet read your article from the library session do that. Write a paragraph or two about it (first paragraph on what the article seeks to accomplish, second paragraph on why it might or might not be useful for you and your research project). You will be writing about research soon.
- Think more about today’s activity and what you might want to do. Maybe spend 20 minutes giving it a test run (with GPT, or internet searching to see if you find something)
If neither of those sound fun and you are struggling to think of an idea, then spend 20 minutes reading from the Try This! textbook. I’ll be asking everyone to read a chapter or two from this book depending on the kind of research they want to do. It might help you think of a project.
