Yeah, I'm just watching OpenAI's newly released ChatGPT Agent power through my Qualtrics survey. I have thoughts about the future of human subject research.

A majority of Americans support greater government investment in research: 57% favor increased medical research funding and 42% support increased scientific research funding. Relatively few want funding cuts: only 10% for medical research and 16% for science.

Average approval of science-related government actions is highest among Republicans (42%), men (28%), graduate degree holders (30%), and high-income respondents (31%). Disapproval is strongest among Democrats (74%), African Americans (56%), women (53%), and those aged 65 and older (55%).

Friends, the 2025 ACM Web Science conference is coming up at #Rutgers this May. We have fantastic presenters and keynote speakers lined up. The schedule and registration just went live on www.websci25.org. Do come and join us! cc @websciencetrust.bsky.social @shugars.bsky.social

How it's going.

Many of us have had a very rough few weeks. But I am told, in no uncertain terms, that nobody has suffered more than this good boy who had to endure a haircut today. #dogsky #dogsofbluesky

Trust in AI is highest among those with moderate knowledge (57%) but lower among those with very high or very low knowledge (38%).
Novices and experts are skeptical, while those with some knowledge might be overconfident. (🧵6 of 6)

We tested Americans on basic AI knowledge with 8 questions. The results:
- Only 23% answered more than half of the questions correctly.
- Less than 1% got all answers right.(🧵5 of 6)

While 62% of Americans trust news from journalists, 48% trust AI-generated content. (🧵4 of 6)

We see an emerging AI divide in the US. People with higher income and education are more likely not only to use AI, but also to trust it, as well as to have higher objective and perceived knowledge about it. (🧵3 of 6)

- 47% of Americans trust AI to act in the public interest
- 50% trust companies to use AI responsibly
Trust in AI is higher than trust in Congress (42%) and in social media (39%) (🧵2 of 6)

We just released a new report on AI trust and knowledge in America: naiom.net/reports/
We examine trust in AI, the companies that use it, and the content it generates. We also measure how much people know about it. (🧵1 of 6)

Validity, efficacy, and bias in computational mental health models: Munmun De Choudhury gives a fantastic talk in the Rutgers Computational Social Science Lab (csslab.rutgers.edu) speaker series #CSS #mentalhealth

Great talk by the awesome @shugars.bsky.social on computational methods and natural language processing at OCMC