@vedransekara
Assoc professor @ITUkbh. Previously @UNICEFinnovate & @sony Working on AI bias, stupid algorithms, and computational social science More silly things at 👉 https://vedransekara.github.io/ #TechWillNotSaveUs 🍉
Der er nu debat om de unoterede aktier i debatten om formueskat (se hvem der ejer dem herunder).
En lidt af en gammel traver er blevet hevet af stalden, nemlig at man ikke kan beskatte ikke-realiserede værdier, der er bundet i selskaber, fordi det ikke er 'rigtige penge'. Men det er ikke korrekt.
Hvor robuste er de danske valgtest? Ikke synderligt, viser ny forskning fra ITU 👇
#dkpol #fv26
According to the latest reports, at least 83 children have been killed and 254 wounded since 2 March in Lebanon, as hostilities have intensified. On average, more than 10 children have been killed and approximately 36 children injured every day over the past week. UNICEF urges immediate efforts to de-escalate the situation and prevent further harm to children.
“As military strikes continue across the country, children are being killed and injured at a horrifying rate, families are fleeing their homes in fear, and thousands of children are now sleeping in cold and overcrowded shelters.” - UNICEF's Edouard Beigbeder
Learn more: https://unicef.link/4b0z3Ho
it turns out that the answer to the headline here is: like humans. listen. don't judge, but don't endorse the beliefs, either.
as my friend wisely said: “There's value in having friendships where it's like, ‘I love you, but also, you're full of shit.’” www.404media.co/ai-psychosis...
Our results demonstrate that even minor tweaks to the algo can produce entirely different outcomes, potentially affecting outcomes of political elections.
Based on our findings we provide recommendations to VAA devs and policy makers, on how to monitor and eval the robustness of these systems
7/7
+ we study how dropping questions, i.e. calculating match similarity based on fewer questions (than the original 20), affects which candidates are recommended.
We quantify changes on candidate and political party-level. I.e. how often are you recommended a new candidate, or new political party
6/7
These systems have been studied in the past (mostly how humans interact with the), but we were unaware of anybody testing their robustness.
As such, we investigate how minor modifications to the algorithm’s weights can change the outcome of which political candidates are recommended
5/7
Through a freedom of information request we got access to the algorithm used to match users and political candidates. The system functions by users filling out a questionnaire which is compared to a database of answers from candidates. User are then presented candidates w. highest agreement
4/7
…and 45% reported they followed the recommendation of a VAA when voting. This means these voting tools can have a large influence on election results.
We look at a popular VAA ('Kandidattesten'), developed by a large media house, Politiken, and wonder, how robust are its recommendations?
3/7
Do you know who you want to vote for? (For US readers, Denmark is a multi-party democracy, where we can choose between 12 different parties)
In the last election 62% of eligible voters used some form of voting assistance app (VAA, or 'Valgtest' in danish) one month prior to the election …
2/7
NEW Preprint out 🚨
Given the upcoming general elections in Denmark, it’s a perfect time to post about our new preprint (w. @robysinatra.bsky.social & Giovanni Astante) --> arxiv.org/abs/2603.03532
(this work was deemed important and highly relevant for FAccT, but also rejected 🫠)
🧵 1/7
Interesting interview study about AI in the labor process. Best possible atmosphere (well-paid, autonomous) still results in more work, not less, because AI helps you multitask, work during nonwork time, extend your capability into other people's turf (experts then need to clean up vibejobs)
Killing people should not be something we do quicker than the speed of thought
Spanish Prime Minister Pedro Sanchez doubled down on his opposition to the attack on Iran by the US and Israel, warning that the conflict risked playing 'Russian roulette' with the lives of millions
The same technological pinpoint accuracy that was used to slaughter children in Gaza now being used to slaughter children in Iran.
A lot of people say AI isn't very good for the world and its outputs are routinely unreliable, but these haters fail to see how it's revolutionizing how we incinerate schoolchildren.
Jeg er rørt over den omsorg, som medierne viser for virksomhedsejeres følelser når det handler om formueskat.
Hvorfor hører vi aldrig om hvordan specialarbejdere føler sig bekymrede over at de skal betale 50% i indkomsskat eller skolelæreres følelser om moms på deres indkøb.
84% of the time, “ChatGPT Health” killed a customer using their paid service as advertised
This is basically legalized insider trading
This weekend’s military escalation in the Middle East marks a dangerous moment for millions of children in the region. UNICEF is deeply concerned by reports of strikes in Iran and across the region, illustrating the grave danger for children. There are reports of schools being struck in Iran, including a girls’ school in Minab in Hormozgan province in southern Iran. Scores of students are reportedly killed and many others injured.
UNICEF echoes the call of the Secretary-General for an immediate cessation of hostilities and de-escalation. UNICEF also urges all parties to exercise maximum restraint and uphold their obligations under international humanitarian and human rights law, including the protection of civilians and the essential services children rely on to survive. Targeting civilians and civilian objects, including schools, is a violation of international law.
UNICEF, together with other UN agencies and humanitarian partners, is actively assessing the situation and stands ready to scale up support to affected children and their families as needed and requested.
This weekend’s military escalation in the Middle East marks a dangerous moment for millions of children in the region.
Full statement from UNICEF: https://unicef.link/473y9Hu
The Onion nails it
Love it when an Onion article has been relevant my entire adult life
Many are appropriately outraged by Altman’s comments here implying that raising a human child is akin to “training” an AI model.
This is part of a broader pattern where AI industry leaders use language that collapses the boundary between human and machine.
🧵/
Great opportunity 👇 especially if you are interested in working at the intersection of AI and behavioral science
Gotta love #FAccaT reviews (il assuming this is about them).
My experience this year is from three reviewers:
- important work
- interesting results
- very relevant to the community
- good methodology w. some limitations
- trivial results 😅
- reject 🙃
This new Lancet study would indicate that the death toll in Gaza is actually around 120,000 at minimum, if the undercount rate is similar since January 2025.
New article in Nature describes a field experiment on X which shows that assigning users to an algorithmic feed pushes conservative content and demotes mainstream news.
Sounds obvious but important in light of prior claims algorithms don't cause attitudinal changes.
www.nature.com/articles/s41...
Can feed algorithms shape what people think about politics? Our paper "The Political Effects of X's Feed Algorithm" is out today in Nature and answers "Yes."
www.nature.com/articles/s41...
*hver gang*
