It was also such a pleasure to catch up with my coworker @oak_raj at @ICWSM!
We each presented different papers here, but our work on using LLMs to mitigate harmful exposure will be presented at @ACLmeeting soon.
The ACL preprint available here: https://t.co/kMI1eTi9tg
Excited to have presented a poster on ‘MetaHarm: Harmful YouTube Dataset annotated by crowdworkers, a domain expert, and GPT’ at @ICWSM in #Copenhagen!
The dataset and paper are available at https://t.co/xF9QWEst9W
Grateful to have connected with amazing people there.
#icwsm
So ICWSM Main Conference Day 1 has kicked off! Catch me at "Session 1: Multimodal Meme Analysis" from 10:30 AM to 11:00 AM, where I’ll be presenting two papers.
The 19th International AAAI Conference on Web & Social Media (ICWSM) is underway in Copenhagen. We are enjoying seeing everyone at the conference so far! Lots of valuable information exchange and networking.
🔥 The list of accepted papers (full, poster, and dataset) for #ICWSM25 is now live! Check it out: https://t.co/640yzSuVBC
🎟️ Want to attend the talks and connect with the authors? Registration is now open! Don’t miss the early bird rate — deadline is May 9th!
🚨🎉 Exciting news! Our paper on algorithmic vs. user nudges for increasing news consumption on YouTube was just published at @PNASNexus
w/ @imxudong, @mwojcieszak, @erickaakcire
Interested in #harm detection & mitigation? See preprint w/ the *wonderful* @clairewonjo👇 2-in-1: taxonomy of online harm & large-scale annotation using MLLMs, crowdworkers & experts. Core result: GPT outperforms crowdworkers but still a 'silver standard' https://t.co/8o6yJ1eYpn
In one of the first papers using ChatGPT-4 for text analysis, I find that it beats both crowdworkers and experts in a difficult tweet annotation task
This work has convinced me that LLMs will completely transform how we use text in the social sciences 🧵
https://t.co/qqI97wvM3G
Excited to share a preprint with @mwojcieszak, titled "Harmful YouTube Video Detection: A Taxonomy of Online Harm and MLLMs (GPT-4-Turbo) as Alternative Annotators"
Does GPT outperform humans in 'harmful' video annotation as well? Find out at https://t.co/y5uOo08Lgv 🙌
Extremely delayed post, but my first #ica24 hackathon wrapped up as well. I presented how to handle and analyze multimodal aspects from social media content.
Thank you @hackingcommsci for organizing such a great event!
My #ica24 summary!
I was honored to present two papers from my first-year PhD work at @poli_com and @ica_cm. I discussed multimodal analysis using LLMs and redefining cross-cutting exposure with digital trace data.
Thank you to my wonderful advisor and coworker, @mwojcieszak!