Przemek Grabowicz ๐Ÿ‡ต๐Ÿ‡ฑ๐Ÿ‡ช๐Ÿ‡ธ๐Ÿ‡ฉ๐Ÿ‡ช๐Ÿ‡บ๐Ÿ‡ธ๐Ÿ‡ฎ๐Ÿ‡ช's Avatar

Przemek Grabowicz ๐Ÿ‡ต๐Ÿ‡ฑ๐Ÿ‡ช๐Ÿ‡ธ๐Ÿ‡ฉ๐Ÿ‡ช๐Ÿ‡บ๐Ÿ‡ธ๐Ÿ‡ฎ๐Ÿ‡ช

@przemyslslaw.bsky.social

CS faculty at University College Dublin and the University of Massachusetts Amherst: responsible AI, social media, news media, computational social science, applied NLP, open-world AI RTโ‰ endorsement, RT=interesting read More: https://przemyslslaw.github.io

6,186 Followers  |  287 Following  |  217 Posts  |  Joined: 07.03.2024  |  2.9579

Latest posts by przemyslslaw.bsky.social on Bluesky

Our first NeurIPS paper on fair representation learning with controllable guarantees. Fantastic research, and a result of extremely dedicated work by @jamesluoyh.bsky.social and our team! Well done! ๐ŸŽŠ

26.11.2025 21:50 โ€” ๐Ÿ‘ 4    ๐Ÿ” 0    ๐Ÿ’ฌ 0    ๐Ÿ“Œ 0

The worst happened. We were DOGEโ€™d. Our NSF funding is gone.

So now thereโ€™s nothing stopping me from sharing Expert Voices Together, a crisis response system for US-based researchers and journalists facing harassment.

It's a true passion project. ๐Ÿงต 1/

expertvoicestogether.org

13.05.2025 16:22 โ€” ๐Ÿ‘ 1696    ๐Ÿ” 778    ๐Ÿ’ฌ 22    ๐Ÿ“Œ 25
Polish flag overlaid with X logo with charts

Polish flag overlaid with X logo with charts

X's Algorithm Amplifies Political Extremes
Prof. @przemyslslaw.bsky.social studied #X during Poland's 2025 election:
๐Ÿ“Š Far-right & far-left posts stay in feeds longer
๐Ÿ“‰ Ruling party content down-ranked 4x
๐Ÿ“ˆ 80% feed content now from outside your network
zenodo.org/records/1751...
#AlgorithmicBias

13.11.2025 12:18 โ€” ๐Ÿ‘ 4    ๐Ÿ” 1    ๐Ÿ’ฌ 3    ๐Ÿ“Œ 0

In recent months, I've seen videos showing Israeli settlers harassing Palestinian women harvesting olives, killing baby goats of Palestinian families, and destroying olive trees on Palestinian lands. Why is this happening? Aren't there policemen and courts prosecuting these crimes?

12.11.2025 17:55 โ€” ๐Ÿ‘ 3    ๐Ÿ” 0    ๐Ÿ’ฌ 0    ๐Ÿ“Œ 0

A small correction: the percentage from outside the network went up from 50% to 80%.

05.11.2025 09:02 โ€” ๐Ÿ‘ 1    ๐Ÿ” 0    ๐Ÿ’ฌ 0    ๐Ÿ“Œ 0

We believe that the only way to achieve neutrality โš–๏ธ and free speech ๐Ÿ’ฌ in social media is through transparency. Right-wing and left-wing policymakers need to talk with each other and online platform representatives to figure out how independent researchers can safely study social media. ๐Ÿงต7/7

03.11.2025 12:58 โ€” ๐Ÿ‘ 0    ๐Ÿ” 0    ๐Ÿ’ฌ 0    ๐Ÿ“Œ 0
Post image

This research subsumes our previous study of the 2025 German federal election (attached figure and paper), before which Elon Musk supported and campaigned for the far-right AfD. We conclude that social media platforms can develop in different ways. ๐Ÿงต6/7
zenodo.org/records/1489...

03.11.2025 12:58 โ€” ๐Ÿ‘ 1    ๐Ÿ” 0    ๐Ÿ’ฌ 1    ๐Ÿ“Œ 0
Preview
Elon Musk on X: "A major update to the recommendation algorithm is rolling out over the next few days. This will help surface smaller accounts and posts outside of your friend-follows network. As always, this will be made open source and undergo continuous improvement." / X A major update to the recommendation algorithm is rolling out over the next few days. This will help surface smaller accounts and posts outside of your friend-follows network. As always, this will be made open source and undergo continuous improvement.

We find that the fraction of posts from outside the network went from 20% to 50% following the 2023 change of the algorithm announced by X owner Elon Musk (attached tweet). It's possible that the algorithm also started to amplify political extremes around that time. ๐Ÿงต5/7
x.com/elonmusk/sta...

03.11.2025 12:58 โ€” ๐Ÿ‘ 4    ๐Ÿ” 0    ๐Ÿ’ฌ 2    ๐Ÿ“Œ 0
Post image

A prior study by @inference.vc of the 2016 Twitter's feed algorithm did not find evidence of amplification of political extremes in seven countries, including Germany (attached figure), so the feed algorithm of X must have changed considerably since then. In fact, it did, as we show next. ๐Ÿงต4/7

03.11.2025 12:58 โ€” ๐Ÿ‘ 0    ๐Ÿ” 0    ๐Ÿ’ฌ 1    ๐Ÿ“Œ 0
Post image

We test whether the feed algorithm favors political extremes, by regressing the number of feed appearances against user engagements received by each post and political party affiliation of its author. We find that each post from political extremes appeared about 20 minutes more in the feed! ๐Ÿงต3/7

03.11.2025 12:58 โ€” ๐Ÿ‘ 0    ๐Ÿ” 0    ๐Ÿ’ฌ 1    ๐Ÿ“Œ 0
Post image

The study (attached) was conducted during the culminating month of Polish presidential election. We tracked news feeds of four sock-puppet accounts and analyzed public posts of 485 Polish politicians and related users. ๐Ÿงต2/7
doi.org/10.5281/zeno...

03.11.2025 12:58 โ€” ๐Ÿ‘ 0    ๐Ÿ” 0    ๐Ÿ’ฌ 1    ๐Ÿ“Œ 0
Post image

Our latest study of "Algorithmic Biases on X before the 2025 Polish Presidential Election" reveals that the default algorithmic feed of X showed mostly posts from right-wing parties and amplified political extremes, most likely due to the 2023 updates of X following its takeover (read on). ๐Ÿงต1/7

03.11.2025 12:58 โ€” ๐Ÿ‘ 13    ๐Ÿ” 0    ๐Ÿ’ฌ 1    ๐Ÿ“Œ 0
Preview
White House proposal would have FCC and FTC police alleged social media censorship | CNN Business A draft executive order from the White House could put the Federal Communications Commission in charge of shaping how Facebook, Twitter and other large tech companies curate what appears on their webs...

This may sound like an impossible task, given how this year Trump and Republicans started to oppose online platform transparency [1]. However, in 2019 Trump and Republicans drafted a platform neutrality legislation [2]. ๐Ÿงต2/2
[1] www.politico.eu/article/us-c...
[2] edition.cnn.com/2019/08/09/t...

02.11.2025 17:23 โ€” ๐Ÿ‘ 0    ๐Ÿ” 0    ๐Ÿ’ฌ 0    ๐Ÿ“Œ 0

We want neutrality โš–๏ธ and free speech ๐Ÿ’ฌ in social media. But can we achieve that without transparency? I don't think so. So right-wing and left-wing policymakers need to talk and figure out how independent researchers can study platforms. An excellent piece for @science.org, with my inputs. ๐Ÿงต1/2

02.11.2025 17:23 โ€” ๐Ÿ‘ 4    ๐Ÿ” 0    ๐Ÿ’ฌ 1    ๐Ÿ“Œ 1
Preview
Meta and TikTok are obstructing researchersโ€™ access to data, European Commission rules Data are needed to study how social media spreads misinformation and influences elections, scientists say

"Access to data by researchers is a crucial element of ensuring transparency and democratic accountability,โ€ @lewan.bsky.social says. โ€œThe public is entitled to know what platforms and their algorithms are doing to the information landscape.โ€
I wrote for @science.org about Meta, TikTok and the DSA ๐Ÿงช

29.10.2025 19:24 โ€” ๐Ÿ‘ 67    ๐Ÿ” 41    ๐Ÿ’ฌ 3    ๐Ÿ“Œ 4

What if it's interpretable AI and users are in charge?

27.10.2025 17:16 โ€” ๐Ÿ‘ 0    ๐Ÿ” 0    ๐Ÿ’ฌ 1    ๐Ÿ“Œ 0
Preview
Introducing GreenEarth We're building advanced open source algorithms for social media

GreenEarth is creating open source AI-driven recommender infrastructure for BlueSky. Type a prompt, see your feed change. We are here for the users, the builders, the dreamers. Join us.
greenearthsocial.substack.com/p/introducin...

23.10.2025 03:08 โ€” ๐Ÿ‘ 83    ๐Ÿ” 24    ๐Ÿ’ฌ 8    ๐Ÿ“Œ 5

I'm feeling inspired and thankful for having so many insightful conversations with extremely bright people! ๐Ÿ˜Š๐Ÿ™

27.09.2025 13:08 โ€” ๐Ÿ‘ 2    ๐Ÿ” 0    ๐Ÿ’ฌ 0    ๐Ÿ“Œ 0
Post image Post image

Last days every morning and evening I was passing by a church that Berliners nicknamed "the hollow tooth", badly damaged in 1943 and never rebuilt as a memorial of WWII, on my way to an excellent and timely workshop about platform transparency organized in Berlin by @dsa40collaboratory.bsky.social.

27.09.2025 12:54 โ€” ๐Ÿ‘ 5    ๐Ÿ” 0    ๐Ÿ’ฌ 1    ๐Ÿ“Œ 0
Post image

Right-leaning podcaster Joe Rogan has chimed in on the suspension of late night host Jimmy Kimmel amid pressure from the Trump administration, saying the government should not determine what comedians say and that conservatives were โ€œcrazyโ€ to support any such interference. https://cnn.it/4mvp3Ik

24.09.2025 15:01 โ€” ๐Ÿ‘ 226    ๐Ÿ” 40    ๐Ÿ’ฌ 39    ๐Ÿ“Œ 5

I've been pigeonholed by both sides. Whenever it happens it's a problem. That said, this topic is often twisted and exaggerated for partisan gains. What we need is balance, neutrality, facts, rules, and procedures.

19.09.2025 00:59 โ€” ๐Ÿ‘ 0    ๐Ÿ” 0    ๐Ÿ’ฌ 0    ๐Ÿ“Œ 0
Preview
Disney Pulls Jimmy Kimmelโ€™s Show After Kirk Remarks Disney said it is pulling ABC late-night host Jimmy Kimmelโ€™s show in the wake of remarks he made about the killing of conservative activist Charlie Kirk.

Read what Kimmel actually said. He only took issue with the rightโ€™s use of Charlie Kirk for political gain:
www.wsj.com/business/med...

18.09.2025 11:16 โ€” ๐Ÿ‘ 0    ๐Ÿ” 0    ๐Ÿ’ฌ 0    ๐Ÿ“Œ 0

I'm voting for free speech with my wallet, and cancelling my Disney+ subscription.

18.09.2025 10:48 โ€” ๐Ÿ‘ 112    ๐Ÿ” 24    ๐Ÿ’ฌ 7    ๐Ÿ“Œ 0
Post image

The design of social media platforms, including their AI systems and algorithms, likely contributes to the over-representation of extreme views (image @financialtimes.com). What platforms do we want? For example shall we balance diverse perspectives and societal stability?
www.ft.com/content/9251...

09.09.2025 10:29 โ€” ๐Ÿ‘ 7    ๐Ÿ” 1    ๐Ÿ’ฌ 0    ๐Ÿ“Œ 0
We conduct a randomized controlled trial (RCT) to understand how early-2025 AI tools affect the productivity of experienced open-source developers working on their own repositories. Surprisingly, we find that when developers use AI tools, they take 19% longer than withoutโ€”AI makes them slower. We view this result as a snapshot of early-2025 AI capabilities in one relevant setting; as these systems continue to rapidly evolve, we plan on continuing to use this methodology to help estimate AI acceleration from AI R&D automation [1].

We conduct a randomized controlled trial (RCT) to understand how early-2025 AI tools affect the productivity of experienced open-source developers working on their own repositories. Surprisingly, we find that when developers use AI tools, they take 19% longer than withoutโ€”AI makes them slower. We view this result as a snapshot of early-2025 AI capabilities in one relevant setting; as these systems continue to rapidly evolve, we plan on continuing to use this methodology to help estimate AI acceleration from AI R&D automation [1].

Programmers who thought AI would make them 20% faster ended up being 19% slower metr.org/blog/2025-07...

15.07.2025 13:37 โ€” ๐Ÿ‘ 111    ๐Ÿ” 37    ๐Ÿ’ฌ 5    ๐Ÿ“Œ 5
Preview
Scientists Are Flocking to Bluesky Academics once loved Twitterโ€”but in the age of X theyโ€™ve abandoned it in droves.

Academics once loved Twitterโ€”but in the age of X theyโ€™ve abandoned it in droves. www.wired.com/story/bluesk...

28.08.2025 15:33 โ€” ๐Ÿ‘ 2308    ๐Ÿ” 599    ๐Ÿ’ฌ 67    ๐Ÿ“Œ 61
Post image

Brazil offers a lesson in democracy to an America that is becoming more corrupt, protectionist and authoritarian econ.st/4peBqLR

28.08.2025 11:47 โ€” ๐Ÿ‘ 1618    ๐Ÿ” 522    ๐Ÿ’ฌ 81    ๐Ÿ“Œ 276

On the day when Israel killed four more Al Jazeera journalists, it's time we recognize that Israel killed in Gaza more journalists than ever died in any other war, including the World Wars. Source: attached research at Brown University. Let this sink in.
watson.brown.edu/costsofwar/p...

10.08.2025 23:30 โ€” ๐Ÿ‘ 9    ๐Ÿ” 8    ๐Ÿ’ฌ 1    ๐Ÿ“Œ 0
Preview
Trump order aims to politicize decisions on federal science grants Researchers worried by plan giving political appointees more control over where funding goes

Research advocates are expressing alarm over a White House directive on federal grantmaking released yesterday that they say threatens to enhance President Donald Trumpโ€™s control over science agency decisions on what to fund. scim.ag/45q6P4J

08.08.2025 20:28 โ€” ๐Ÿ‘ 72    ๐Ÿ” 49    ๐Ÿ’ฌ 3    ๐Ÿ“Œ 5
Post image

Holy crap every single post is negative www.reddit.com/r/ChatGPT/

08.08.2025 17:11 โ€” ๐Ÿ‘ 274    ๐Ÿ” 29    ๐Ÿ’ฌ 11    ๐Ÿ“Œ 40

@przemyslslaw is following 20 prominent accounts