When a research participant referred to his AI companion as a “diet girlfriend,” Research Assistant @merylye.bsky.social saw a useful metaphor: chatbots lack social nutrition; they “offer comfort in the moment, but are rarely filling on their own." datasociety.net/points/socia...

AI Sycophancy Expert Survey Sign-Up We’re conducting a 10-minute expert survey to better understand how researchers define and evaluate AI sycophancy. If you’re interested in participating, please sign up using this form. We’ll email yo...

Studying AI sycophancy?
We’re running a 10-min expert survey on how researchers define & evaluate it.

If you’ve authored ≥1 paper on sycophancy or related areas (alignment, personalization, preference learning), you’re eligible to participate.

Sign up here: forms.gle/dY5L79cAxb95...

Oddly accurate

Comment to the FDA on Generative AI-Enabled Digital Mental Health Medical Devices

In our comment to the FDA, we draw on our ongoing research to focus on what people’s actual, everyday use of chatbots for mental and emotional support means for the FDA’s approach to generative AI-enabled digital mental health medical devices. datasociety.net/announcement...

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 🧵

laurenkuehne.github.io/grfpChanges/

Assessing Collective Reasoning in Multi-Agent LLMs via Hidden Profile Tasks Multi-agent systems built on large language models (LLMs) promise enhanced problem-solving through distributed information integration, but also risk replicating collective reasoning failures observed...

Huge thanks to Aoi(naito-aoi.github.io) and Hiro(shirado.net)!

👉 arxiv.org/abs/2505.11556

Please boost this thread if you find it interesting!
[n/n]

All the books I ordered last week just arrived! Can't wait to pour myself a drink, relax, be too tired to read and play around on my phone all night

I would love to never see the phrase "little is known about" in a paper ever again.
(1) Most of the time the authors are far more confident about that statement than they should be.
(2) There are lots of things we know little about because no one cares. Just not knowing isn't a good motivation.

I’m actually waiting for another Boston tea party

This is one of the worst violations of research ethics I've ever seen. Manipulating people in online communities using deception, without consent, is not "low risk" and, as evidenced by the discourse in this Reddit post, resulted in harm.

Great thread from Sarah, and I have additional thoughts. 🧵

Successfully defended my thesis today! Call me Doctor now :D

Big thank you to my advisors @vhellendoorn.bsky.social and @clegoues.bsky.social! :)

At this point, “breaking news” should just be called “broken news”.

We're leaving Twitter behind and moving to Bluesky! Here, we'll keep tackling the toughest challenges at the intersection of technology, ethics, and society—without all the noise. Follow us to join thoughtful conversations, responsible innovation, and real-world impact. #SolveBig #ResponsibleTech

Found an old notepad from elementary school. Not exactly sure what I was up to…but it seems like I’ve had a longstanding interest in social media users?
Hopefully now I know better than to use USA Today as a reference.

Actions Speak Louder than Words: Agent Decisions Reveal Implicit Biases in Language Models While advances in fairness and alignment have helped mitigate overt biases exhibited by large language models (LLMs) when explicitly prompted, we hypothesize that these models may still exhibit implic...

Huge thanks to my advisors Hiro(www.shirado.net) and Sauvik @sauvik.me!
👉 arxiv.org/abs/2501.17420
Please boost this thread if you find it interesting!
[n/n]

I think the negative aspects overshadow the positive ones once the size of our networks/amount of content exposure pass a certain threshold.
Everything in moderation…right?

Hi Jaz, I was wondering if you have time stamps for edge creations, or other ways that would allow for historical snapshots of the Bluesky network structure. I would be interested in studying the network growth.

Differences in misinformation sharing can lead to politically asymmetric sanctions - Nature We find that conservatives tend to share more low-quality news through social media than liberals, and so even if technology companies enact politically neutral anti-misinformation policies, political...

🚨In Nature🚨
Meta is dropping fact-checking to avoid anti-conservative bias- but is there actually evidence of bias?
We this test empirically & find that conservatives
* ARE suspended more
* BUT share more misinfo
So suspension isn't necessarily evidence of bias www.nature.com/articles/s41...

The examples you give are much more direct, but this kind of question makes me think about the butterfly effect and chaos theory.

This fall, I was TA for Network Analysis, taught by @vasilescu.bsky.social and @patpark.bsky.social .
To make it fun(ner), we played a song for each lecture. Check out our website and playlist! I would love suggestions of other network-related songs for next year.
open.spotify.com/playlist/468...

Posted this at the old place, but thought folks on @bsky.app might like to know I finally met my "mother".

I am making a Societal Computing starter pack for folks that (broadly) look at the impact of computer science on society. So far, mostly people affiliated with the eponymous program in the school of computer science at CMU, but pls respond to this thread if you want to be added.

go.bsky.app/2avpExM

Congrats Sanjay! Awesome news!
Since you’re not at the helm anymore, what’s going to happen to Reddit for Researchers?