Masoud

thank you! i enjoyed it too!

that's why I would not say such claims are place holders. they may be imprecise, but that is just a function of our current state of the field

I agree! :DA claim like "negation is innate" is underspecified and can be made precise in the three ways you mentioned and many more. I'd say all innateness claims ultimately require genetic evidence + specification of neurodevelopmental mechanisms that result in the learning, processing, & behavior

makes sense. It could be just terminological differences but once we separate them, innateness claims about each is substantial. Someone that claims "negation is innate" is making a substantial and (in principle) verifiable claim about representational constraints, regardless of onto/philogeny

2. learning mechanism (Bayesian update, trigger setting, gradient descent, etc) 3. Learning biases (e.g. priors over symbols or rules in a Bayesian system, hyper-parameter settings etc).As far as I can see these are quite separate components of learning that can be innate/learned to varying degrees

I'm not sure if we are using the immune system example the same way so I'll set it aside to avoid confusion. More generally, any learning system can have more or less of the following innate/inbuilt components: 1. representations (e.g symbols & combinatorial rules, vector space, etc) ...

or specific learning mechanisms. A good analogy may be the immune system. We have innate immunity that targets pathogens based on innate representations of what pathogens look like from evolutionary exposure and we have adaptive immunity that learns new representations of pathogens in our lifetime

I'm using innate knowledge to refer to all that is in a system prior to learning e.g. representations, learning mechanisms, biases... As Locke said no one should doubt some learning mechanisms and some capacity to represent ideas is innate. At issue is if specific ideas/representations are innate

I agree that inductive bias is a good term for a subclass of innate knowledge but not all innate knowledge is reducible to inductive bias obviously. In the extreme case a system can have innate stored information literally. Or innate knowledge of what info is computed how etc

not sure."X is innate" is a verifiable claim independent of how it evolved or developed. More clear in cases like innate vs. learned/adaptive immune system. It is currently harder to verify for cognitive constructs than biological ones but still a claim worth verifying or falsifying in its own right

Does the culture you grow up in shape the way you see the world? In a new Psych Review paper, @chazfirestone.bsky.social & I tackle this centuries-old question using the Müller-Lyer illusion as a case study. Come think through one of history's mysteries with us🧵(1/13):

how much of that is private funding/grants vs the federal government?

oh absolutely

I'm convinced over-reliance on forced choice tasks have created a lot of task-specific findings that lack external validity in different fields of behavioral sciences and one day we have to face that crisis

each proposition is a note? a cord is a conjunction? disjunction and negation would be hard to play but maybe we gotta limit it to one octave :))

Kinda fascinating how researchers (systematically?) overestimate cognitive abilities of human-built systems like LLMs while underestimating the cognitive abilities of natural systems like animals or children

LA fires are a good reminder that the only difference between me and any refugee is luck.

RIP Bob Bayley RIP Bob Bayley   We are so sad to learn that Dr. Robert Bayley , Professor Emeritus of Linguistics , UC Davis, and an associate member of t...

Very sad to have to share news of my old friend and colleague Bob Bayley's passing. He was a true scholar, mentor, and upright human being. I'll miss him a lot. utlinguistics.blogspot.com/2024/12/rip-...

Episode #61: Carole Chaski - Linguistics Careercast "You need to have a heart of service to do this work" Carole Chaski is a forensic linguist who is considered one of the leading experts in the field. Her research has led to improvements in the methodology and reliability of computational linguistics and inspired research on the use of this approach for authorship identification.

Linguistics Careercast #podcast: With guest Carole Chaski, a forensic linguist. Her research has led to improvements in the methodology and reliability of stylometric analysis and inspired research on the use of this approach for authorship identification. www.linguisticscaree...

sorry you make a good point I had no disagreement there. Just added a thought :)

a lot of ppl warned the industry that AI models are not ready to be deployed for high stakes tasks. They simply didn't listen. Many still don't listen or care. And it makes sense. the end game for a business is increasing revenue and profits not better decisions

CogSci 2025 Call for Reviewers CogSci is looking for abstract reviewers for the 2025 conference in San Francisco. If you are interested in this role, please complete the following form. If you have any questions, please contact cog...

Want to review for #CogSci2025? Click this thing. @cogscisociety.bsky.social
docs.google.com/forms/d/e/1F...

Yeah my bet would be that we just don't know. I also don't know what to make of a language having or not having a word for a concept. It seems like people make obvious inferences based on that information which to me seems not straightforward

I'd really doubt that claim can be properly substantiated. First, not having a word for X is hard to verify for a language. Second across many understudied languages with no written form! what's the book?

Black and white photographs of Victor Ferreira (left) and Fernanda Ferreira (right) who will be jointly giving the Nijmegen Lectures in January 2025.

Don't miss the Nijmegen Lectures, a unique 3-day series of talks & discussion by renowned psychologists @fernandaedi.bsky.social & Victor Ferreira, Jan 7-9, 2025! Topics will span the psycholinguistic continuum from intention to articulation. More info & registration here: www.mpi.nl/events/nijme...

that would be a good thread I'd think about it!

that's a great question. I gotta think about it for the ranking but a lot of early neural network papers and Chomsky's papers with Miller come to mind

needs to be on Bluesky too

There is so much really brilliant research from the 50-70s in cogsci and linguistics that is not being read or referenced anymore and I find that a real pity. We shouldn't reinvent the wheel or rediscover things we already know. Older literature should be actively read, taught, and discussed

Speech recognition in adverse conditions by humans and machines In the development of automatic speech recognition systems, achieving human-like performance has been a long-held goal. Recent releases of large spoken language

ASR systems are getting *really* good. But how do they actually compare to humans when the listening environment is challenging? In our JASA-EL paper led by Chloe Patman, we compare SoTA ASR systems in pub and speech-shaped noise and directly compare their performance against L1 English speakers