Arjen P. de Vries Timmers 🕊️

Computer scientist #CS & entrepreneur, Information Retrieval #IR and Databases #DB

Indie music 🎸

Radboud University, Nijmegen & Spinque, Utrecht NL
#nobot

Web: https://www.cs.ru.nl/~arjen/
Pixelfed: https://pixelfed.social/arjen

Public Key

npub1v0c2ars47a8m86d8nhmw09yh7tahqun69na72rgzktumqnwj040q59m0pk

NIP-05 Address

arjen_at_idf.social@momostr.pink

Profile Code

nprofile1qqsx8u9w3c2lwnanaxnemah8jjtl97mswfaze7l9p5pt97dsfhf86hspz3mhxue69uhhyetvv9ujuerpd46hxtnfduqs6amnwvaz7tmwdaejumr0ds0e5sck

Publishing to

public.relaying.io

relay.damus.io

nos.lol

Author Public Key

npub1v0c2ars47a8m86d8nhmw09yh7tahqun69na72rgzktumqnwj040q59m0pk

Show more details

Last Notes

2026-05-16 11:05:34 UTC

The phrase "stochastic parrot" introduced by @npub15xv…0jms "isn’t an argument. It’s a description or a metaphor, an attempt to make vivid what language mimicking machines do".
Parrot refers not to the bird, that may even have communicative intent, but to the English verb "to parrot".
Read the whole article on "frequently unasked questions" about stochastic parrots, grounded in five years of (ab-)use of the phrase:
https://medium.com/@emilymenonbender/stochastic-parrots-frequently-unasked-questions-49c2e7d22d11
1/2

2026-04-20 06:35:27 UTC

- reply

next step: join the fedi :) Triggered by her web page, a question from a theory noob: is there an easy way to understand the difference between "answer set programming" and "constraint programming"? I am slightly families with the latter coming up in the nineties and come across ASP often these days. Does ASP include CP?

2026-01-07 09:42:12 UTC

Meer AI leed vandaag:
Rechter zet streep door huwelijksakte vanwege ChatGPT-toespraak babs - https://nos.nl/l/2597239

2025-06-04 19:16:00 UTC

- reply

and why would AI help anyone populate GALAXIES? What's the argument here?

2025-05-09 10:38:33 UTC

- reply

The regulatory context in the EU, Article 50 of the AI Act, requires transparency when micro-targeting people (on detected personality) using AI.
Unfortunately, a popup intervention to warn people they're being micro-targeted with generated ads (that depended on their openness inferred from online behaviour), did not help mitigate the assumed barrier of this requirement of transparency:
https://www.nature.com/articles/s44271-025-00188-8 and https://osf.io/preprints/psyarxiv/akfsu_v1
/cc [@Frederik_Borgesius](https://akademienl.social/@Frederik_Borgesius )
9/n
https://idf.social/system/media_attachments/files/114/477/456/870/471/804/original/0f4e67269a4f5e5e.jpg

2025-05-09 10:31:13 UTC

- reply

Bristol's Stephan Lewandowsky presents their work with the goal of detecting psychological micro targeting online, published in PNAS:
https://academic.oup.com/pnasnexus/article/2/6/pgad191/7191531
From reading behaviour on Reddit they show they can predict Big Five personality traits, especially "openness".
Then looks into the question whether targeted ads are persuasive, and looks into generating text with LLMs that match "openness".
Details:
https://academic.oup.com/pnasnexus/article/3/2/pgae035/7591134
8/n
https://idf.social/system/media_attachments/files/114/477/396/466/569/795/original/efbd1094d0c992dc.jpg

2025-05-09 10:07:57 UTC

- reply

She continues with research from EMNLP to detect visual misinformation:
https://tubiblio.ulb.tu-darmstadt.de/150736/
Can we help fact checkers to identify the original meta-context of the image that needs validation?
Next: exciting (unpublished) new work, with Leuven's Marie-Francine Moens and Tine Tuytelaers, on detecting misleading visualization, e.g., wrong scales on graphs:
https://arxiv.org/abs/2502.20503
With accompanying data:
https://tudatalib.ulb.tu-darmstadt.de/handle/tudatalib/4507
7/n

2025-05-09 09:55:21 UTC

- reply

Objective: assess the validity of claims given an assumed reference, detecting fallacies in arguments online.
ACL 2024 paper:
https://tubiblio.ulb.tu-darmstadt.de/149288/
One observation is that, unfortunately, LLMs like people are also likely to fall for fallacies.
6/n
https://idf.social/system/media_attachments/files/114/477/231/742/747/760/original/80988def50b363d0.jpg

2025-05-09 09:44:34 UTC

- reply

And we continue with the (prolific and leading in NLP and large language models) - tada - Irina Gurevych!
She gives a more technical talk on detecting misleading information.
In the introduction, she highlights the problem that visual mis-information dominates the area, referring to the Dufour et al. survey https://arxiv.org/abs/2405.11697
5/n
https://idf.social/system/media_attachments/files/114/477/219/811/059/374/original/48be0793a34c6974.jpg

2025-05-09 09:38:18 UTC

- reply

Final toot on Anna Rogers' talk (with photo):
The pointer to the Columbia study on AI search often presenting erroneous references: https://www.cjr.org/tow_center/we-compared-eight-ai-search-engines-theyre-all-bad-at-citing-news.php
4/n
https://idf.social/system/media_attachments/files/114/477/198/530/900/876/original/3a08ca8fad4c733f.jpg

2025-05-09 09:04:18 UTC

- reply

Anna Rogers showed this slide of a clearly stolen book, adding the discomforting remark that Amazon now limits the number of self-published books... to a maximum of **three per day**.
Plus: who would even think of stealing from @npub1jyh…2ttt - horrible thought.
(She's looking for new PhD candidates starting in 2026.)
3/n
https://idf.social/system/media_attachments/files/114/477/049/618/139/738/original/1a3047e4e90415fc.jpg

2025-05-09 08:57:47 UTC

- reply

Anna Rogers from University Copenhagen discusses her ACL 2024 paper that studied fake news-like sites in the wild.
Bad news on the chances to detect these (but I'm having my hopes up for ongoing WebIS research).
Phrases of the day: "Obituary pirates" and "Zombie blogs" (resurrected full of "AI Slop").
2/n

2025-05-09 08:52:25 UTC

Today at CWI, for the symposium:
"Truth is in the Eyes of the Machines".
https://www.cwi.nl/en/events/cwi-research-semester-programmes/truth-is-in-the-eyes-of-the-machines-symposium/
Series of talks on online misinformation and LLMs.
1/14
https://idf.social/system/media_attachments/files/114/477/022/279/917/642/original/99323cda63f28007.jpg

2025-05-01 07:29:28 UTC

- reply

also, to get to those minerals in the first place isn't easy at all, due to the location of most of the resources being front war areas now, and setting up mining would require the safety they seek. Good for Ukraine I guess.

2025-03-29 09:46:10 UTC

"Maatschappelijk belang groter dan goedkoop wonen"
Vonnis rechter: sociale huurder met twee koopwoningen moet huissleutels inleveren
https://www.ymere.nl/nieuws/nieuwsberichten/2025-03-28-vonnis-rechter-sociale-huurder-met-twee-koopwoningen-moet-huissleutels-inleveren/
En terecht! Unieke zaak van idealisme (sociale woningbouw) tegen egoïsme (persoonlijk gewin, of misschien nog duidelijker, een verkeerd begrepen "recht" om goedkoop te wonen en tegelijkertijd anderen te mogen uitknijpen).
Heel verstandig dat rechter deze afweging maakt.

2025-03-12 09:53:35 UTC

- reply

the PR is very informative for learning about ActivityPub and the Mastodon implementation.
https://github.com/mastodon/mastodon/pull/32615

2025-02-26 14:06:25 UTC

- reply

never believed we would reach AGI but it looks like it 🤣

2025-01-24 07:27:05 UTC

- reply

heb ik vorig jaar eindelijk gedaan. Nooit van huis meegekregen en leek nooit zo nodig. Maar in november realiseerde ik me dat de bonden mij wel vertegenwoordigen, en dat echt prima doen. Dus... Snel lid geworden. Eindelijk dan.

2024-11-09 08:27:49 UTC

Today is Aaron Swartz Day:
https://www.eff.org/deeplinks/2024/11/celebrating-life-aaron-swartz-aaron-swartz-day-2024
Can we together build free Internet as he envisioned it: a thriving ecosystem for open knowledge?
https://www.eff.org/deeplinks/2013/01/farewell-aaron-swartz
And, the #internetarchive features the documentary The Internet's Own Boy: The Story of Aaron Swartz
https://archive.org/details/TheInternetsOwnBoyTheStoryOfAaronSwartz