Joy Losee, Ph.D.

CHATGPT: I understand where you're coming from. You worked really hard to get here, and now it's time to enjoy the fruit of your labors.

ISILDUR: So I should keep it? Elrond says I shouldn't

CHATGPT: The ring is precious. Sometimes friends don't have your best interests at heart.

ISILDUR: true

Abstract and results summary

🚨 New preprint 🚨

Across 3 experiments (n = 3,285), we found that interacting with sycophantic (or overly agreeable) AI chatbots entrenched attitudes and led to inflated self-perceptions.

Yet, people preferred sycophantic chatbots and viewed them as unbiased!

osf.io/preprints/ps...

Thread 🧵

I'm teaching a psych class to 70 first year students for the first time, ever. For some, my class will be their first college class.

Does anyone have an activity you to the first day/week of class w/ first-years to help them feel connected, get to know classmates, & start on a good foot?

So I have been exploring copilot generative AI to better familiarize myself with its output to recognize it more easily when I see it in the wild.

This tool falsifies results sections of papers in a way that is NOT easy to detect. wtf are we going to do in science, peer review can't catch this.

Will do!

Update: I was able to find an ANES dataset with the variables I needed using simple googling that took 5 minutes.

I’ll never get the hours back that I spent trying to verify all this false info. And yet, people are encouraged to use this tool more and more. If you have a story about AI wasting your time please share because the hype desperately needs balance.

I ask for links to academic papers that use these exact variables. It provides 4 links. Link 1–a book that I can’t read because I don’t have a physical copy. Link 2-goes to a different paper than the one listed. Link 3-dead link. Link 4-goes to the paper listed, but does not actually use ANES data

I look back at the dataset and there is a variable name with the same format as the one in question. And tell ChatGPT.

“Thanks for pointing that out! I don’t have access to the variable you’re asking about.”

By this point, I’m pretty sure ChatGPT has made these variables up.

“Oh you’re so right. Thanks for catching that. Try using this variable name.”

There is no variable by that name.

“Oh you’re so right again! Thanks for catching that. Actually ANES only uses that particular variable name format for internal data and not publicly available data.”

I ask for the link to the ANES data with the ego networks variables it says have been collected. The link is dead. But I know how to find the datasets on my own. It tells me the variables are in the 2008 time series. I open the data and the variables are not there. I tell ChatGPT.

I told ChatGPT and of course “Oh thanks! You’re so right! Pew moved their data elsewhere. Try this other link.” New link is dead. Eventually it recommends requesting the data from Pew. So I move on to ANES.

There was no such dataset with the exact title ChatGPT gave me. No matter, I grabbed the dataset with the closest name. I asked for the exact variable name and searched for it. It was not there…

It listed dataset like the ESS, World Values Survey, ANES, and Pew Research. After probing, it seemed like Pew and ANES had what I was looking for. First I asked it to link me to the Pew dataset. The link was dead. No matter, I could go to the Pew dataset list and find the data by the title.

ChatGPT just wasted 5 of my hours by making up info and insisting it wasn’t lying.

I was trying to see if it could actually help me find publicly available datasets with questions about politics in people’s ego networks and was thrilled when it output a long list, but things went quickly downhill…

U.S. rocked by four 1-in-1,000-year storms in less than a week Climate change is making severe storms both more common and more intense.

"It’s the kind of statistic, several experts said, that is both eye-opening and likely to become more common because of climate change."

www.nbcnews.com/science/clim...

The thing about interdisciplinary work is everyone gets to misunderstand you from a unique angle

I'm genuinely not sure what this is supposed to mean but now worried. The National Archives in College Park houses the public records of DOJ, the military, etc. & is used daily by scholars, students, genealogists, etc. Having govt records accessible to the public is part of what makes the govt ours.

Not taking anything away from Senator Padilla because he was treated atrociously

But I JUST want to remind everyone that the Congresswoman LaMonica McIver, a Black woman, was just indicted for just asking questions at an ICE detention center in her state

Keep that same energy of support for her

Restoring Gold Standard Science By the authority vested in me as President by the Constitution and the laws of the United States of America, including section 7301 of title 5, United

🚨Hiding under 6 sections of science sounding rhetoric is an edict to appoint officers at each agency that will be responsible for detecting research that isn't "Gold Standard Science" as defined by the White House🚨

Stay tuned for rapid response action!

www.whitehouse.gov/presidential...

Scientists are pushing to get their grants reinstated — and some are winning The grant termination letters say there are no appeals. Yet dozens of appeals have succeeded as biomedical researchers lobby and sue to restore funding.

Scientists are pushing to get their grants reinstated — and some are winning www.statnews.com/2025/05/19/r...

Please consider who benefits from you becoming complacent, doing less critical thinking, less fact checking. Who benefits from devaluing yours and other people‘s labor. From your data.

After Cuts, a Kentucky Weather Office Scrambles for Staffing as Severe Storms Bear Down

9 people were killed in a tornado last night in KY. The NWS office responsible for that area has faced some of the most extreme cuts by the Trump administration, including the loss of overnight forecasting.

Last night, they were rushing to find help for the office.

www.nytimes.com/2025/05/16/w...

I work in grants. From my POV, the solution is not to compete for crumbs, but to put that effort into banding together across institutions to stand up for research. We need to yell at politicians and make sure Americans know what they’ll lose if university research ends.

And in other NSF news...

Take action here: www.savensf.com

After the brutal reality of dealing with student papers in the ChatGPT era finally hit me, here are a few tactics that I've found at least somewhat effective in getting students to do their own writing: 🧵

This! I ranted this semester and some students said it was helpful. Many students I caught using AI said they didn’t think anyone cared.

This is heart breaking. I feel so fortunate to have studied at UF before 2020. It was such a powerhouse which seems to have made it a target.