@timoroettger.bsky.social
Cognitive scientist / Linguist - Full Professor at University of Oslo Director of @manylanguagesc.bsky.social, PI of ICONIC (https://www.hf.uio.no/iln/english/research/projects/iconic/) #MetaScience #OpenScience #DataViz #PresentationDesign #SciComm
I think my dog has Tourette
if you try to estimate sample size for a bayesian version of these models through simulations, you don't think you run into practical problems? or what do you mean?
these models tend to be computationally very expensive and time consuming, which makes for example simulation-based sample size justifications much more difficult.
regardless of the psychological
nature of acceptability judgments, Likert and VAS scales require more complex mixture models because both tend to elicit many extreme responses at the end of the scales that might reflect a different (part of the) decision process and need to be modeled too!
This month’s #ManyLanguagesChallengeOfTheMonth comes from the world of **grammatical acceptability judgments** by the ML2 CLAPS team. @ambridge.bsky.social
Dear #LabPhon20 organizers: Will there be a registration fee for online participants presenting a digital poster? And if so do you know the exact figure yet?
#LabPhon @labphon.bsky.social
Tenure track position in the multimodal language department at the MPI!
its amazing how chatgpt knows everything about subjects I know nothing about, but is wrong like 40% of the time in things im an expert on. not going to think about this any further
How many times, in how many contexts, in how many ways am I going to have to say that this is what "AI" does— what it fundamentally *is*— before it sinks in? That all Bullshit engines do is statistically correlate training data & inputs via their weights to produce outputs you are likely to accept…
I get that, particular for software development, the LLM-assisted vibe coded prototype-building is aw-inspiring (it really is) but an AI CEO telling you you should buy AI products and "seriously engage" with AI because its coming for all our jobs? Yeah. Right. No conflict of interest whatsoever...
⚠️ Despite all the hype, chatbots still make terrible doctors. Out today is the largest user study of language models for medical self-diagnosis. We found that chatbots provide inaccurate and inconsistent answers, and that people are better off using online searches or their own judgment.
...JNDs are notoriously misinterpreted as absolute thresholds (which they are not), so we are left again with arbitrary threshold...
The SESOI literature is so unsatisfying. It almost always discusses applied cases. Remains unclear how to apply SESOIs to basic psychological mechanisms that we measure e.g. with RTs or accuracy. The only direction with potential was the discussion of JNDs, but...
ManyLanguages is on two social platforms:
✨ LinkedIn: www.linkedin.com/company/1107...
✨ Bluesky (this page): bsky.app/profile/many...
Follow us to stay in the loop, join discussions, and help us spread the word about Big Team Science.
The next one (though it's adults only) is a team-science project with @manylanguagesc.bsky.social looking at passives. We're keen to get as many diverse languages as possible so please get in touch if you'd like to join as a coordinator or a co-investigator for your language bsky.app/profile/many...
Instead of suffering in silence, let’s share them. We’d love to hear your stories, opinions, comments—or, in rare and glorious cases, actual solutions.
We’re trialling a new social activity: #ManyLanguagesChallengeOfTheMonth
Running large-scale projects means encountering puzzles, hiccups, and the occasional “why is this happening” moment.
Nevermind the jobs you had, tell me five classes you took in college
Ontological Arguments
Roman Archeology
Fictional Writing
Model Thinking
Universal Design
All of them gave me invaluable perspectives on how to think about the world, science, and humans!
"These machines take OUR jobs now?" 😂
New blog post introducing Causion - a web app for causal inference teaching and learning: pedermisager.org/blog/causion....
"P-Hacking with one prompt" by Shigeto Kawahara.
TL/DR: Ask LLMs to find a significant effect, and they will oblige.
www.researchgate.net/publication/...
Science comms drowning in PISS?
Really comprehensive but digestible. Consider sharing it with your students. I will.
Amazing, Andy! Just started teaching and thought im gonna spend some time on discussing LLMs but now I might just share your content with them. Just watched the first video and this is the kind of balance between comprehensive yet easily digestible treatment that rly fills a gap for me. Thank you!
Please share your work for educational purposes :) #language #linguistics
adding some folks who might have something that fits the bill?
@casillas.bsky.social @jofrhwld.bsky.social @bodowinter.bsky.social @linguistbrian.bsky.social @fernandaedi.bsky.social
Feel free to broadcast for reach.
#linguistics
Dear exp linguist, for my advanced stats class, I am looking for articles that
- are experimental linguistics-ish
- report on linear mixed models that did not converge with the intended random effect structure (using lme4)
- published open data and scripts
Please @ me, I would appreciate it!
Get involved! Ben's project is a fantastic opportunity to contribute to Big Team Language Science, to learn from the many challenges involved when scaling up our empirical efforts to many languages, and to be part of large interdisciplinary and global project.
⭕Translation Coordinator:
- Recruit translators for individual languages
- Coordinate translation process
- Coordinate with the method lead for study implementation when translations are finished
- Coordinate translation implementation checks
⭕Scaling Coordinator:
- Increase the coverage of underrepresented languages / cultures
- Ensure typological diversity and proportional representation of languages
- Actively recruit team members and labs outside of virtual academic commons