Sign in

Vasco Brazão 🏳️‍🌈

@vascobrazao.bsky.social
377 followers 1K following 366 posts

Recovering behavioral scientist, posing as a statistical consultant. Applied stats #RStats, neurodiversity, learning Ukrainian, writing things

PostsRepliesMedia
Reposted by Vasco Brazão 🏳️‍🌈
Elle Cordova @ellecordova.bsky.social · 03/09/2026
Rough Draft vs Final Draft
254116853320
Reposted by Vasco Brazão 🏳️‍🌈
Iris van Rooij 💭 @irisvanrooij.bsky.social · 29/08/2026
The conclusion is NOT "we need to debias LLMs" Besides that it's impossible, it also detracts from the point, which is: We shouldn't use machines to offload of the very cognition that is constitutive of the work to be done & supercharge existing inequities and injustices in the process. 🧵
544191
Reposted by Vasco Brazão 🏳️‍🌈
Iris van Rooij 💭 @irisvanrooij.bsky.social · 29/08/2026
In a "collaboration" (which we left), someone used an LLM to make a "summary" on topics for the project (without our consent) in which published research ideas from @olivia.science & me understandably popped up, but were credited to men, while our names showed up nowhere in the "summary". 🧵
227682
Reposted by Vasco Brazão 🏳️‍🌈
Benjamin Riley @benjaminjriley.bsky.social · 21/07/2026
"To wall off 'contemporary political issues' from our recent history is to make a choice, and here, Anthropic’s algorithm, programmed by Anthropic’s software engineers, is making it very clear that teachers should not connect the past to the present. Shudder."
buildcognitiveresonance.substack.com
Why is Claude for Teachers?
Anthropic bumbles its way into education
043
Reposted by Vasco Brazão 🏳️‍🌈
Rob Calver @robcalver.bsky.social · 02/07/2026
I am giving away a free copy of the #BayesianWorkflow book by @statmodeling.bsky.social, @avehtari.bsky.social, @rmcelreath.bsky.social et al!! Please like, repost, and follow to enter. I will randomly draw the winner next Weds July 8. 😀 routledge.com/9780367490140 #RStats #DataScience #Bayesian
36947
Reposted by Vasco Brazão 🏳️‍🌈
Пангея 🇺🇦 @pangeaua.bsky.social · 26/06/2026
Hi, it's been a while, since I've asked for your help. I'm currently helping to fundraise money for Starlinks mini for 28th Mechanized Brigade who is fighting at Kostiantynivka. Any help is really needed. send.monobank.ua/jar/9TZcdbbMCt #Ukrainianview
03524
Reposted by Vasco Brazão 🏳️‍🌈
Kirsti Miller @kirstimiller30.bsky.social · 12/06/2026
There are NO clear & verifiable biological demarcations between the sexes, research from genetics, physiology,& others shows that nature is much more imprecise, rich (& to some degree impenetrable) from its diversity, multiplicity & variability. theconversation.com/world-athlet...
theconversation.com
World Athletics’ mandatory genetic test for women athletes is misguided. I should know – I discovered the relevant gene in 1990
World Athletics says its genetic test for women ensures ‘the integrity of women’s sport’ – but science does not support this overly simplistic idea.
0145
Vasco Brazão 🏳️‍🌈 @vascobrazao.bsky.social · 12/06/2026
it was surprisingly easy to turn off the stupid generative ai button and banner on adobe acrobat. I don't believe I'll ever want an "ai summary" of my train tickets, thanks
010
Reposted by Vasco Brazão 🏳️‍🌈
Adam Kucharski @adamjkucharski.bsky.social · 08/06/2026
If you're involved in training/teaching, or a student, this is worth a read – "The Anatomy of a Learning Stall": tagide.com/blog/llm/the...
111831
Reposted by Vasco Brazão 🏳️‍🌈
Richard McElreath 🐈‍⬛ @rmcelreath.bsky.social · 27/05/2026
An odd etymology I learned recently: English kaput is (obviously) from German kaputt which is from French être capot (to be a bonnet) which meant to lose all tricks in a card game The machine is kaput = the machine is bonnet
3211
Reposted by Vasco Brazão 🏳️‍🌈
Chise @sailorrooscout.bsky.social · 25/05/2026
GOOD NEWS! Researchers at Stanford University have developed a UNIVERSAL vaccine known as GLA-3M-052-LS+OVA, that protects against a wide range of respiratory viruses, bacteria AND even allergens. The vaccine is delivered intranasally AND provides broad protection in the lungs for several MONTHS.
ecoticias.com
Goodbye to seasonal viruses as we know them: the first universal vaccine against respiratory infections and allergies is now closer than ever
Scientists are getting closer to a universal nasal vaccine that could protect against respiratory viruses, bacteria and even allergies
345147344601
Vasco Brazão 🏳️‍🌈 @vascobrazao.bsky.social · 18/05/2026
I suppose the airplane ride on my way to discussions on safely deploying LLMs in an edtech context is as good a time as any to read this!!
020
Reposted by Vasco Brazão 🏳️‍🌈
Dr Abeba Birhane @abeba.blacksky.app · 18/05/2026
Not just lobbying! our new paper maps 27 mechanisms of regulatory capture used by Big AI + 11 narrative framings that rationalise capture. “Big AI’s Regulatory Capture: Mapping Industry Interference and Government Complicity” will be presented at #FAccT2026 next month arxiv.org/abs/2605.068... 1/
Title: Big AI’s Regulatory Capture: Mapping Industry Interference and Government Complicity

abstract: Over the past decade, the AI industry has come to exert an unprecedented economic, political and societal power and influence. The well-functioning of regulatory and oversight structures and processes that govern the industry thus have paramount ramifications for everything from fostering public trust in systems marketed as AI, the credibility of scientific knowledge, educational and healthcare services and products, information ecosystems, the environment, rule of law and integrity of democratic process. In this paper, we first develop a taxonomy of mechanisms enabling capture to provide a comprehensive understanding of the problem. Grounded in design science research (DSR) methodologies and extensive scoping review of existing literature and media reports, our taxonomy of capture consists of 27 mechanisms across five categories. We then develop an annotation template incorporating our taxonomy, and manually annotate and analyse 100 news articles. The purpose behind this analysis is twofold: validate our taxonomy and provide a novel quantification of capture mechanisms and dominant narratives. Our analysis identifies 249 instances of capture mechanisms, often co-occurring with narratives that rationalise such capture. We find that the most recurring categories of mechanisms are Discourse & Epistemic Influence, concerning narrative framing, and Elusion of law, related to violations and contentious interpretations of antitrust, privacy, copyright and labour laws. We further find that Regulation stifles innovation, Red tape and National Interest are the most frequently invoked narratives used to rationalise capture. We emphasize the extent and breadth of regulatory capture by coalescing forces — Big AI and governments — as something policy makers and the public ought to treat as an emergency.
7377221
Reposted by Vasco Brazão 🏳️‍🌈
Berna Devezer @devezer.bsky.social · 17/05/2026
true scholarship is spending a whole career chasing one perfect citation
0141
Reposted by Vasco Brazão 🏳️‍🌈
Mark @aguasonic.com · 17/05/2026
' the "eliza effect" on steroids + meth + ketamine.
011
Reposted by Vasco Brazão 🏳️‍🌈
Olivia Guest · Ολίβια Γκεστ @olivia.science · 16/05/2026
One might say your work on the topic is... intractably unendingly fruitful 🍍🍈🥝🥭🍇
141
Reposted by Vasco Brazão 🏳️‍🌈
🇵🇸Tim Henke (tɪm 'ɦɛŋ.kə) @timhenke.bsky.social · 16/05/2026
Summarising requires figuring out what is important LLMs can certainly produce a response containing items that *sound* like they *would be* important, but in the end importance is a pretty subjective judgment that's hard to delegate to someone unless your personal ideas of importance overlap
0143
Reposted by Vasco Brazão 🏳️‍🌈
Wiep Hamstra @wiep.bsky.social · 16/05/2026
You or a machine can easily make it look like a summary by picking first sentences of a paragraph but words need a context to give meaning. Machines cannot do that. Nuances are lost. For me it’s hard to grasp that people even think an LLM has something to do with meaning or cognition.
131
Reposted by Vasco Brazão 🏳️‍🌈
Mirjam Sp @teunemans.bsky.social · 16/05/2026
Interessant gesprek (met wetenschappelijke argumenten) over waarom LLM’s niet kunnen samenvatten.
032
Reposted by Vasco Brazão 🏳️‍🌈
Iris van Rooij 💭 @irisvanrooij.bsky.social · 15/05/2026
2. Climbing towards NLU: On meaning, form, and understanding in the age of data by Bender & Koller (2020) aclanthology.org/2020.acl-mai... 3/3
aclanthology.org
Climbing towards NLU: On Meaning, Form, and Understanding in the Age of Data
Emily M. Bender, Alexander Koller. Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics. 2020.
241
Reposted by Vasco Brazão 🏳️‍🌈
Iris van Rooij 💭 @irisvanrooij.bsky.social · 15/05/2026
1. On the dangers of stochastic parrots: can Language Models be too big? 🦜 by Bender, Gebru, McKoller-Major, & Shmitchell (2021) dl.acm.org/doi/abs/10.1... 2/3
dl.acm.org
241
Reposted by Vasco Brazão 🏳️‍🌈
Iris van Rooij 💭 @irisvanrooij.bsky.social · 15/05/2026
Here an illustration that I cannot resist sharing. It has journal, date, and authors wrong (deleted a.o. 3 women co-authors and replaced with a man with whom I often co-author). ‘summary’ slop, while abstract was right there. Not using the LLM would have been less work. bsky.app/profile/iris...
32213
Reposted by Vasco Brazão 🏳️‍🌈
Mark @aguasonic.com · 15/05/2026
details, details. grey-goo for the internet. write code for a living - and the "ai" examples are practically, without exception, garbage. They _look_ like they might work. Half the time they can't even be _parsed_ - forget whether the logic is good or not.... People, mostly, just have no idea....
032
Reposted by Vasco Brazão 🏳️‍🌈
bachlvrdwb.bsky.social @bachlvrdwb.bsky.social · 15/05/2026
My primary thing is trust. I know the creature hallucinates. But for me, pretty much every AI is backed/from/for a very right wing and or technocratic nut who needs to define things only in their terms. It would be in their interest to shape responses for certain things, in other words.
011
Reposted by Vasco Brazão 🏳️‍🌈
Denny Carter @dennycarter.bsky.social · 15/05/2026
really insightful and horrifying stuff from @thefarce.org
electorate. In 2024, advisers to Musk ran a $45 million false-flag campaign so precise in its cynicism that it deserves to be described slowly and disdainfully. Muslim voters in Michigan received pro-Israel ads designed to look like Harris campaign materials. Jewish voters in Pennsylvania received the opposite message from the same shop. Young liberals got videos about how Harris had betrayed the progressive movement. Working-class white men in the Midwest got warnings that Harris would institute racial quotas and take away their Zyn pouches. Four brand names, zero common origin, one coordinating strategy.
They called it "false positives" internally. The goal was not persuasion. It was subtraction
— push enough Democratic-leaning voters into confusion, disgust, or exhausted abstention. Harris dropped eight million votes from Biden's 2020 total. Trump gained fewer than two million. The election was decided, as the architects
401996951
Reposted by Vasco Brazão 🏳️‍🌈
Iris van Rooij 💭 @irisvanrooij.bsky.social · 14/05/2026
“The latter demonstrates nicely the core flaw of LLMs, i.e., they cannot read and summarize, because they have no understanding or knowledge.” irisvanrooijcogsci.com/2025/08/12/a...
irisvanrooijcogsci.com
AI slop and the destruction of knowledge
Cite as: van Rooij, I. (2025) AI slop and the destruction of knowledge. This week I was looking for info on what cognitive scientists mean when they speak of ‘domain-general’ cognition. I was curio…
13220
Reposted by Vasco Brazão 🏳️‍🌈
Nate Angell @xolotl.org · 28/02/2026
Great post! Wondering if you had the chance to see the paper I wrote with @annamillsoer.bsky.social where we argue for the term AI "mirage" over "hallucination" or other terms: papers.ssrn.com/sol3/papers....
papers.ssrn.com
131
Reposted by Vasco Brazão 🏳️‍🌈
Maarten Paulusse @maartenp.bsky.social · 24/02/2026
Tired of AI hype posts? You might like my sober assessment of whether AI-generated summaries are suitable for studying and research. Spoiler alert, they are not. The text is primarily aimed at students and researchers, but has much broader relevance. So share freely!: www.tue.nl/en/our-unive...
tue.nl
Are AI-generated summaries suitable for studying and research?
Despite didactic, ethical, and environmental concerns, the use of GenAI is on the rise in academia. For most applications, the jury is still out on whether and how they will benefit education and rese...
12388171
Reposted by Vasco Brazão 🏳️‍🌈
Iris van Rooij 💭 @irisvanrooij.bsky.social · 14/05/2026
Oh, one more recommendation: bsky.app/profile/iris...
075
Reposted by Vasco Brazão 🏳️‍🌈
Alex Brown 🇵🇸 (they/them) @bookjockeyalex.bsky.social · 14/05/2026
Great resources in the replies. For those at the lower and secondary school levels, I have some additional thoughts. 🧵
12916
Reposted by Vasco Brazão 🏳️‍🌈
Olivia Guest · Ολίβια Γκεστ @olivia.science · 14/05/2026
This by @lucyavraamidou.bsky.social might be useful doi.org/10.1007/s111...
doi.org
Demystifying AI: The urgency of a critical stance on the use of AI systems in education - PROSPECTS
The education field has generally embraced artificial intelligence (AI) systems for their (undocumented) potential to reduce teacher workload, address inequalities, and enhance learning through personalization. However, issues such as algorithmic bias, monoculturalism, discrimination, use of private data, and access to AI systems raise concerns about exacerbating learning inequalities, especially for minority students. We argue that researchers in education have largely failed to critically address the limitations of AI systems, and that a critical stance framed within critical AI literacies is urgently needed to ensure responsible use, if any, of AI systems in the classroom. We first examine how AI systems have been conceptualized for educational purposes. Through engagement with related literature, we discuss what has been overlooked by the mystifying of AI and the failure to engage more thoroughly and critically with existing literature on the social nature of learning. Following that, we provide an overview of contemporary AI literacy frameworks through a comparative approach and discuss their potential and limitations. We conclude with a set of recommendations for future research, teacher education, and curriculum design, framed within a critical stance to foster fair, ethical, and responsible use, where appropriate, of AI systems in education.
085
Reposted by Vasco Brazão 🏳️‍🌈
Alex Brown 🇵🇸 (they/them) @bookjockeyalex.bsky.social · 14/05/2026
We want kids to build the skills of reading longer documents effectively (such as navigating methodology, abstract, conclusion). LLMs shortcut this skill-building. They learn to rely on a computer, not to use their brains to pick apart a text.
1264
Reposted by Vasco Brazão 🏳️‍🌈
ve, the ghost of info past @vortexegg.com · 14/05/2026
Summarization involves describing the “aboutness” of a resource, which is a) not necessarily derived from a statistical description of its observable features but also it’s meaning and grounded truth claims, and b) is related to the context of origin of the resource and context of use of the summary
0114
Reposted by Vasco Brazão 🏳️‍🌈
Iris van Rooij 💭 @irisvanrooij.bsky.social · 24/04/2026
It is really this easy. Take any *cognitive capacity* X. For instance, X = reasoning, learning, communicating, summarizing, reading, writing, etc. If any one asks: Can an LLM do X? You can answer: “No, LLMs cannot do X, because X is a cognitive capacity and LLMs have none” LLMs are decoys.
621372
Reposted by Vasco Brazão 🏳️‍🌈
EssayWells @essaywells.bsky.social · 14/05/2026
Iris has great references in the thread below. A very short rebuttal is that a summary should include the important facts and ideas, and LLMs have access to neither, they only have access to text. If asked for a summary they will give you a shorter text which probably includes frequently used words.
0112
Reposted by Vasco Brazão 🏳️‍🌈
Language Learning Matters @llmllc.bsky.social · 14/05/2026
Controlling for language level. Not only did I have to fix everything I tried because whoever programmed had no understanding of language development or what was relevant for the lessons, the last document I tried it with came out with a summary so racist, I never used it again.
241
Reposted by Vasco Brazão 🏳️‍🌈
Olivia Guest · Ολίβια Γκεστ @olivia.science · 14/05/2026
Published with some minor changes! Please share widely. 😊 Guest, O., Suarez, M., ... & van Rooij, I. (2026). Against the Uncritical Adoption of 'AI' Technologies in Academia. Digital Culture & Education, 16(2), 85–118. doi.org/10.5281/zeno... Journal: www.digitalcultureandeducation.com/volume-162
digitalcultureandeducation.com
Volume 16.2 — Digital Culture & Education (ISSN: 1836-8301)
2208116
Vasco Brazão 🏳️‍🌈 @vascobrazao.bsky.social · 14/05/2026
how would you argue that LLMs do not have the capacity to summarise? I'd say that a summary has communicative intent and an author with the capacity to distinguish relevant from irrelevant information, but would love a deeper treatise on this. Any papers from cognitive science that could be useful?
10469
Reposted by Vasco Brazão 🏳️‍🌈
Rachit Dubey @rachitdubey.bsky.social · 07/04/2026
🚨New preprint and our results are rather concerning.. We find the "boiling frog" equivalent of AI use. Using large-scale RCTs, we provide *casual* evidence that AI assistance reduces persistence and hurts independent performance. And these effects emerge after just 10–15 minutes of AI use! 1/
261533684
Reposted by Vasco Brazão 🏳️‍🌈
Richard McElreath 🐈‍⬛ @rmcelreath.bsky.social · 07/04/2026
I would have preferred to have the "draw the rest of the owl" meme on the cover, but this will do. Seems like it is on schedule, and we'll leave some typos so you know we didn't write it with AI.
cover of the book "Bayesian Workflow" by Gelman, Vehtari, et al. Coming out later this year, in the summer probably.
1237956
Reposted by Vasco Brazão 🏳️‍🌈
Iris van Rooij 💭 @irisvanrooij.bsky.social · 22/11/2025
AGI is a scam www.ru.nl/en/research/...
ru.nl
Don’t believe the hype: AGI is far from inevitable | Radboud University
Will AI soon surpass the human brain? If you ask employees at OpenAI, Google DeepMind and other large tech companies, it is inevitable. However, researchers at Radboud University and other institutes ...
36721
Reposted by Vasco Brazão 🏳️‍🌈
tante @tante.cc · 19/02/2026
I am a luddite and to me there is a lot of joy in technology. In technology that we deploy for all our wellbeing. Solar panels and vaccines, high-speed rail and wikipedia. But they joy stems from the feeling of being able to be part of humanity in an embedded, meaningful way. Not from buying shit.
572413357
Vasco Brazão 🏳️‍🌈 @vascobrazao.bsky.social · 19/02/2026
!!!! is this what I think it is !!!!
and orchid with the tiniest bud coming out in between two leaves. I think it will be flowers!!!what looks like a flower stem coming out of an orchid. what if it is true!!!
020
Vasco Brazão 🏳️‍🌈 @vascobrazao.bsky.social · 18/02/2026
"The price (and gift) of authenticity is closer proximity to queerness. Cis men can take the risks inherent in rejecting the system that gives them their power, or continue to conform to its hegemony, but they cannot do both." 🥹
0268
Reposted by Vasco Brazão 🏳️‍🌈
Meredith Whittaker @meredithmeredith.bsky.social · 25/12/2025
Merry Christmas! May you enjoy the cinnamon nutmeg season as much as I do, with all the peace, major chord progressions, warm drinks, butterfat and sparkling this and that, whatever you believe, however work it ❤️🎄 🎁 here's a great album to get your day going www.youtube.com/watch?v=4Mv4...
youtube.com
Oscar Peterson Trio 1964 Tokyo Concert
YouTube video by Blue Born
1226120
Reposted by Vasco Brazão 🏳️‍🌈
Sonja Drimmer @sonjadrimmer.bsky.social · 26/12/2025
Hi, hello! My book _Impressive Politics: Print before the Press in Late Medieval England_ is forthcoming from University of Pennsylvania Press. I think about print all the time but in ways you might not expect. sonjadrimmer.com/blog-1/2025/...
sonjadrimmer.com
"The Printing Press Democratized Knowledge": When Slogans Masquerade as History — Sonja Drimmer
The phrase is said so frequently it seems, like the mechanism it celebrates, to mechanically replicate itself.  It's become a favorite catchphrase among tech boosters of any sort (see my post on...
513027
Reposted by Vasco Brazão 🏳️‍🌈
Timnit Gebru @timnitgebru.blacksky.app · 17/02/2026
I helped build the FAccT academic conference, and white men like Seth have to come along and make sure that the people we tried to get away from, the TESCREAL eugenicist ghouls, have every academic space in addition to the billions they're drowning in.
I'm not sure if it's worth adding this, but in case anyone's wondering, 
@AmandaAskell
 is a brilliant philosopher doing by far the most important philosophical research anyone in the world is doing right now (in terms of both impact and philosophical interest). I've had the privilege of seeing her at many stages along this journey, from a chance meeting in the hallways of Coombs (the building where ANU Philosophy used to be) back in about maybe 2015, to a visit with her at OpenAI's offices on the day, more or less, that she and others decided to leave OpenAI to form Anthropic, to a number of workshops and other things since, where she showed an incredible generosity of spirit to the folks both in philosophy and in the FAccT community who most needed to learn from her work.
521060