How CHIBE affiliates use AI (and what they don’t use AI for)
AI talk is inescapable these days, and it’s no different at Penn’s Center for Health Incentives and Behavioral Economics (CHIBE).
This fall, our Center is launching an AI taskforce to discuss how artificial intelligence tools can save us time, teach new skills, or inspire ideas. In the meantime, our faculty and staff are already experimenting with AI models with their work. A recent survey we conducted with our affiliates found that over 60% of respondents use AI at least several times each week,with ChatGPT being the most used model and Claude being the second most popular.
CHIBE spoke with some of our affiliates to hear about their uses, thoughts, and fears around AI.
How he uses AI: I’ve been using AI as a learning tool for unfamiliar statistical methods or when I need a quick refresher. I love how it breaks down complicated concepts when you feed it a real-world data problem, usually serving up code snippets you can grab and build into your initial programs. You absolutely need your own statistical knowledge to vet the quality of what the AI tells you. Because the clarity of explanations varies by model, I would cross-reference multiple AI models as needed.
How he uses AI: I primarily use it for writing letters of support, rapidly organizing thoughts into prose, literature searching, and reviewing grants prior to final edits from the research team.
How she uses AI: I use it for transcription of sticky notes from a design or brainstorming session.
How he uses AI: AI has been really useful for me to conduct literature review of topics that I would like to explore more of.
What he doesn’t use AI for: I deliberately do not use AI to write manuscripts or papers.
How she uses AI: I use AI to help me code. It’s particularly great for debugging, restructuring data, adding details to figures, automating repetitive tasks, and checking code for conceptual errors. I let AI see statistical code, but I never let it see data.
How she uses AI: I use AI for reviewing and synthesizing the literature (prior to human review).
What she doesn’t use AI for: I do not use it for research ideation. I feel it limits creativity.
How he uses AI: I use it as a brainstorming partner to stress-test ideas for research projects.
How he uses AI: I’m using AI for doing rapid literature reviews as well as critiquing and editing my papers and grants for flaws and contradictions in logic/chronology/etc.
How he uses AI: I use it as an image/graphics generator for early-stage behavioral intervention prototypes.
How he uses AI: I use AI to surface unanswered questions in an area I am curious about and for summaries of work in various areas.
How she uses AI: I use AI to check grants for responsiveness to guidelines and to help cut them down to word count requirements.
How she uses AI: I use the AI companion in Zoom to provide meeting summaries.
How she uses AI: I find it helpful to create a first draft of plain language summaries when I’m sharing research findings back to study participants.
What she doesn’t use AI for: I sometimes use AI to help me brainstorm how to design an infographic, but I find that ChatGPT often introduces errors (and stylistically I don’t always like what it produces) so I typically use Canva instead.
How he uses AI: I use AI often to get a background on the literature in new (and old) areas I’m doing research in. I also often use it for writing — though only to improve single sentences, one sentence at a time. I also sometimes use it for inspiration; like, I am working on a book on emotion in economics, and I asked it for a good opening story. I asked 4 different AIs this same question, and one (Claude) gave me a totally brilliant suggestion.
What he doesn’t use AI for: I don’t use AI for writing from whole cloth; I like to maintain my own writing style. I also don’t use it to review/referee papers.
AI fears: I am in awe of it, which makes me very very afraid of its effects. I am convinced that it will pretty quickly be able to take over most jobs, creating unfathomable economic damage/upheavals. I’m super-concerned that it’s going to make new generations stupid, in the same way that Google Maps atrophied people’s navigation skills — their sense of geography/direction. I’m worried that it’s going to become everyone’s source of information, which is especially scary given that it is going to continue to be controlled by sinister reactionary billionaires and trillionaires. And I’m worried about its use in warfare. I think we need massive worldwide regulation, but, judging from our success in combating climate change, and ongoing wars, and our current president, I see no prospect of that happening.