Tessera
Folgen

Dagmar Fraser

@dagmarfraser.bsky.social

Doctoral Researcher, Senior Centre Technician @thechbh.bsky.social, Birmingham Transformative Humanities Doctoral Fellow, MATLAB SIG Chair & Ambassador https://linktr.ee/dagmarfraser

745 Posts1005 Follower3402 Folgt
Peter Tennant@pwgtennant.bsky.social · last wk.

Peter's Law of Chatbot Accuracy*: "everything you read that's produced by an LLM is absolutely true, except for the rare output about which you happen have expert knowledge". *Remixed from Knoll's Law of Media Accuracy, where the exact same rules apply!

By Cory Doctorow (GPG 0xBF3D9110957E5F4C)@doctorow.pluralistic.net · last wk.

AI is a normal tech. That means there are times when it is useful and times when it is pointless or actively harmful. One rule of thumb for chatbots is that they can only provide useful information to experts who have the discernment to ignore the defective output that LLMs *always* emit. 60/

Barbara Pomiechowska@bpomie.bsky.social · 2w ago

Hi #CogSci2026 folks! Anyone into #compositionality #geometry and/or babies' minds? Karthika, an amazing PhD student from the lab is presenting some cool preliminary results. We use geometry as a case study for compositionality and program induction in babies. Do stop by her (virtual) flash talk!

Karthika Kamath@karthikakamath.bsky.social · 2w ago

Can babies understand different geometrical shapes?👶🏼Join my Flash talk 41.1 at #cogsci2026 to find more on July 24th,17:30(Rio) @bpomie.bsky.social @mathiassablemeyer.bsky.social @sarahruthbeck.bsky.social @paulm-k.bsky.social @cogscisociety.bsky.social @bbabylab.bsky.social @thechbh.bsky.social

Cat Hicks@grimalkina.bsky.social · 3w ago

I won't show you all the demographic back and forth, but I'll show you just how easy it is to make an AI-administered survey good, even with the stupid AI tone ("honest" lol), you can dramatically improve LLM output with a good harness Here are inventory instructions: github.com/DrCatHicks/i...

Two honest boundaries up front, so they don't surprise you later: I won't recommend for or against EV, and I won't weigh in on dose changes — those are the oncology team's call. What I can do well here is help you build a clear picture of what to watch for, how strong the evidence behind the safety profile actually is, and what specific, well-timed questions to bring to their next appointment. We can also do this in pieces — answer what's easy now, and we can pick up the rest later. Here's the inventory. The most important thing to know about EV is that its risk profile is very regimen- and person-specific, so a handful of these answers will genuinely change what I search for and what I tell you to watch. Answer what you can, skip what doesn't apply or you don't know.1. The regimen and setting — this one matters most. Is this EV on its own, or EV plus pembrolizumab (often written "EV+P")? Pembrolizumab is an immunotherapy that adds a whole separate category of possible reactions, so knowing this reshapes the search. And roughly what setting — advanced/metastatic, or something else you were told? 2. Where they are in the course. When did treatment start, and which cycle are they on now (or how many infusions so far)? Some of the most serious reactions cluster in the early cycles; others (like nerve symptoms) build up cumulatively over time. Timing changes what's most likely right now. 3. Baseline health that maps onto EV's specific known risks. Any of these that apply: Blood sugar — any diabetes, pre-diabetes, or higher body weight? (EV can push blood sugar dangerously high, and this risk is concentrated in certain people.) Nerves — any numbness, tingling, or weakness before starting treatment, from any cause? Skin — any pre-existing skin conditions, or a history of bad drug rashes?
Stephen Turner@stephenturner.us · 3w ago

"The journal is a seventeenth-century technology asked to do four modern jobs at once: disseminate results, certify their quality, allocate scholarly attention, and confer career credit. It does none of them well." arxiv.org/abs/2607.05454

Publishing Without Journals: An Open, Forkable Archive with Attributed Review

The journal is a seventeenth-century technology asked to do four modern jobs at once: disseminate results, certify their quality, allocate scholarly attention, and confer career credit. It does none o...

arxiv.org

Richard Sever@richardsever.bsky.social · last mo.

“move from preprint to peer-reviewed publication leaves the central claims of most abstracts intact, indicating preprints are a reliable source…papers that were never posted as preprints were retracted at roughly twice the rate of those that were…” www.biorxiv.org/content/10.6...

Tracking claim changes from preprint to publication across 72,644 biomedical studies using large language models

Preprints now disseminate a large share of biomedical research before peer review. Because they have not yet passed peer review, some scientists regard preprint claims as unverified or potentially unr...

biorxiv.org

Dr. Rebecca Hirst@hirstrj.bsky.social · last mo.

We will be running out full official PsychoPy and Pavlovia workshop (virtually) again in October ! If you have starting PhD students who need equipping with skills to build their experiments then check it out and share !

PsychoPy@psychopy.org · last mo.

Join us for our next official PsychoPy and Pavlovia workshop! 📅 October 13th- 16th 🕐️ 4 half days to accomodate time zones. Learn how to use the tools with our team! workshops.psychopy.org#official-psy...

Mike Frank@mcxfrank.bsky.social · last mo.

Most of the statistical tests you learned in stats class — t-test, ANOVA, correlation — are actually special cases of a single thing: linear regression. Ch 7 of Experimentology argues that thinking in models, not tests, is more flexible and a better foundation for theory. 🧵 experimentology.io

Bild
Gareth Watkins@garethwatkins.bsky.social · last mo.

Me: Okay, there's been a thunderstorm, the heatwave should be over now. (New health bar appears) (The sun is now named 'Sol, Starflame of Regret') (Objectively the greatest piece of classical music ever recorded starts playing).

Dagmar Fraser@dagmarfraser.bsky.social · last mo.

You can run external commands from MATLAB 😺 External commands are anything you would normally do in a terminal. Very handy if you suddenly find yourself with thousands of ffmpeg jobs cutting up mp4s based on audio onsets… You do it with the wonderfully named bang operator!

Bryan Blessed@bryb.bsky.social · 2mo ago

Shout out to the Scots showing the lovely Americans what alcohol, travel insurance, and socialised health care waiting at home can achieve. "Big historical horsecunt was pure beggin' for a conin'", noted one onlooker.

Charley Wu@thecharleywu.bsky.social · 2mo ago

Look how far we've come in just a few years! We're fast as well! It turns out, recruiting reviewers is much easier when they're not being used as free labor for massively profitable publishers Support free-to-publish and Open Access science by submitting to Open Mind today! direct.mit.edu/opmi

Open Mind | MIT Press

direct.mit.edu

Llanguagemit.bsky.social@languagemit.bsky.social · 2mo ago

—— Open Mind is MIT Press’s Diamond Open Access Cognitive Science journal. We now have our first Impact Factor: 2.9 as of June 2026 Submit your papers to Open Mind direct.mit.edu/opmi