Haroon

@MaazBinMusa Did your fine-tuning crash OpenAI?

Damn, it happened 10 minutes after I deployed a model for fine-tuning. Thought I was in for a lecture from my advisor XD

🚨🎉 Exciting news! Our paper on algorithmic vs. user nudges for increasing news consumption on YouTube was just published at @PNASNexus w/ @imxudong, @mwojcieszak, @erickaakcire

@manoelribeiro @PrincetonCS @PrincetonCITP Congrats Manoel! 🎉

I'm thrilled to announce that I'll join @PrincetonCS/@PrincetonCITP as an assistant professor in Spring 2025 — can't wait to join this amazing group of researchers. I will continue to work on online platforms / the societal impact of AI. If you’d like to collaborate, reach out!

We asked you to submit prototype social media algorithms to be tested, and you did! We had 21 strong entries, and our judges have chosen nine finalists. These include adding credible news, bridging-based ranking, downranking toxicity, and more. rankingchallenge.substack.com/p/announcing-t…

New Preprint 🚨: Read our new paper: arxiv.org/abs/2403.13362 where we try to answer the question: Can we leverage LLMs and realistic bot accounts to incentivize users to consume more news content? More below👇[1/6]

@dggoldst @jakehofman @DavMicRot Interested!

🚨Microsoft PhD internship alert🚨 PhD students: The Microsoft Research NYC Computational Social Science group is really good. Apply to do a 12 week project of mutual interest with @jakehofman @davmicrot & @dggoldst Winter / spring starts encouraged! jobs.careers.microsoft.com/us/en/job/1660…

YouTube, the Great Radicalizer? There's an ongoing debate as to whether YouTube has a tendency to make problematic recommendations (think extreme, radicalizing, etc). We share our findings in our new @PNASNews paper pnas.org/doi/10.1073/pn… [w/ @mharoon_ @mwojcieszak et al]

Interested in algorithms, rabbit holes, radicalization? 👇Auditing YouTube's Recommendation System for Ideologically Congenial, Extreme, and Problematic Recommendations is finally out in @PNASNews pnas.org/doi/10.1073/pn… Long hrs over 2+ yrs w/@mharoon_ @zubair_shafiq @prasantm

🎉My paper on #auditing #YouTube for ideologically congenial, extreme, and problematic #recommendations has now been published and can be accessed at @PNASNews. Long time coming.🙂 pnas.org/doi/10.1073/pn…

Chatted with @levanhieu about his research on automated filter rule generation for ad-blocking on this week's episode of @__NothingToHide. Interesting stuff!

My first PhD paper: "The Inventory is Dark and Full of Misinformation: Understanding the Abuse of Ad Inventory Pooling in the Ad-Tech Supply Chain" (@zubair_shafiq, @rishabn_) got accepted (shepherding) to @IEEESSP 🎉. Stay tuned for details. More works to unravel in this space.

Interesting tool/website developed by @mharoon_ and his team for visualizing how YouTube's algorithm reinforces political bias. I've been playing with it & the "left" side STILL promotes Fox News videos, something I observed firsthand. youtubeaudit.com/explore/explor…

🎉 Excited to share that my first-authored paper on Ideological Congeniality in YouTube recommendations has been accepted @PNASNews. It's been an exciting journey with excellent reviewer feedback. w/ @nshuman_chhabra, @mwojcieszak, @zubair_shafiq, @xinliuucdavise1, @prasantm

This week in a special episode with @atmarkopoulou , @zubair_shafiq , @proffnes , and @nlaout from @proper_data Frontier Project, we unpack the current state of user privacy on the internet and explore the path forward! Tune in to listen: nothingtohide.online

The great @imxudong presenting #ICA23: unobtrusive nudges to #YouTube #algorithm increase recommendations *and* exposure to news, *and* diversifies recommendations and exposure for Strong Conservatives. Nudges to users do nothing. with @mharoon_ @erickaakcire

@imxudong @UNDCommDept Congratulations!

Super thrilled to announce that I'll be joining @UNDCommDept as an Assistant Professor this fall! I'm beyond excited to embark on this journey, and I'm grateful for the incredible support that's helped me reach this milestone!

New preprint 🚨 We show that recommender system audits are of limited utility in determining the prevalence of phenomena like radicalization, rabbit holes, etc., if they do not model how users interact with algorithms (which no one really does!) 📰: arxiv.org/abs/2302.11225

💯

News coverage of my group on our research focuses of good ai vs bad ai by @UCDavisCOE engineering.ucdavis.edu/news/good-vs-b…

We are very excited to launch Nothing to hide? A data privacy podcast! nothingtohide.online