Robert Hawkins
@rdhawkins.bsky.social
asst prof @Stanford linguistics | director of social interaction lab 🌱 | bluskies about computational cognitive science & language
Reposted by Robert Hawkins
✨New preprint! Why do people express outrage online? In 4 studies we develop a taxonomy of online outrage motives, test what motives people report, what they infer for in- vs. out-partisans, and how motive inferences shape downstream intergroup consequences. Led by @felix-chenwei.bsky.social 🧵👇
November 11, 2025 at 4:34 PM
✨New preprint! Why do people express outrage online? In 4 studies we develop a taxonomy of online outrage motives, test what motives people report, what they infer for in- vs. out-partisans, and how motive inferences shape downstream intergroup consequences. Led by @felix-chenwei.bsky.social 🧵👇
Reposted by Robert Hawkins
Our paper @sarabogels.bsky.social covering our pre-registered multi-year research is now finally out in Cognition. We show that in conversations people reduce their multimodal signals non-linearly; the steeper this non-linear drop-off the more communicative success.
www.wimpouw.com/files/Bogels...
www.wimpouw.com/files/Bogels...
November 11, 2025 at 4:49 PM
Our paper @sarabogels.bsky.social covering our pre-registered multi-year research is now finally out in Cognition. We show that in conversations people reduce their multimodal signals non-linearly; the steeper this non-linear drop-off the more communicative success.
www.wimpouw.com/files/Bogels...
www.wimpouw.com/files/Bogels...
Reposted by Robert Hawkins
Starbucks workers this Thursday 11/13 are starting a long strike to win a first union contract
*** NOBODY SHOULD BUY ANYTHING FROM STARBUCKS AS LONG AS WORKERS ARE STRIKING ANYWHERE IN THE COUNTRY***
Consumer boycotts are very powerful when combined with workplace militancy
*** NOBODY SHOULD BUY ANYTHING FROM STARBUCKS AS LONG AS WORKERS ARE STRIKING ANYWHERE IN THE COUNTRY***
Consumer boycotts are very powerful when combined with workplace militancy
November 11, 2025 at 7:58 PM
Starbucks workers this Thursday 11/13 are starting a long strike to win a first union contract
*** NOBODY SHOULD BUY ANYTHING FROM STARBUCKS AS LONG AS WORKERS ARE STRIKING ANYWHERE IN THE COUNTRY***
Consumer boycotts are very powerful when combined with workplace militancy
*** NOBODY SHOULD BUY ANYTHING FROM STARBUCKS AS LONG AS WORKERS ARE STRIKING ANYWHERE IN THE COUNTRY***
Consumer boycotts are very powerful when combined with workplace militancy
Reposted by Robert Hawkins
In our new paper (open access in PSPB), we asked where children’s greater moral concern for animals stems from. We find that children place less moral weight on species membership alone (i.e. are less speciesist) than adults. #socialpsyc #devpsy #philsky journals.sagepub.com/eprint/ZMRUM...
November 11, 2025 at 1:09 PM
In our new paper (open access in PSPB), we asked where children’s greater moral concern for animals stems from. We find that children place less moral weight on species membership alone (i.e. are less speciesist) than adults. #socialpsyc #devpsy #philsky journals.sagepub.com/eprint/ZMRUM...
Reposted by Robert Hawkins
What factors influence alignment btw infant eye-gaze & parental report measures of early vocab? The excellent Haley Weaver & I found that parental confidence & mutual exclusivity effects (distractor knowledge) impact alignment btw these "gold standard" tasks onlinelibrary.wiley.com/doi/epdf/10....
Interrogating Early Word Knowledge: Factors That Influence the Alignment Between Caregiver‐Report and Experimental Measures
You have to enable JavaScript in your browser's settings in order to use the eReader.
onlinelibrary.wiley.com
November 11, 2025 at 2:18 PM
What factors influence alignment btw infant eye-gaze & parental report measures of early vocab? The excellent Haley Weaver & I found that parental confidence & mutual exclusivity effects (distractor knowledge) impact alignment btw these "gold standard" tasks onlinelibrary.wiley.com/doi/epdf/10....
Reposted by Robert Hawkins
Hi BlueSky! Georgetown Linguistics is here and ready to post about our department and community! Follow us for all things GU Linguistics! #linguistics #georgetown #GURT2026
November 6, 2025 at 5:57 PM
Hi BlueSky! Georgetown Linguistics is here and ready to post about our department and community! Follow us for all things GU Linguistics! #linguistics #georgetown #GURT2026
Reposted by Robert Hawkins
For years, advertising has been how Stereogum made money to pay its writers. While we made progress moving away from that model, Google AI search has decimated the business almost overnight. Today we relaunch with a greater focus on subscription. We’d love your support: stereogum.com/2478838/ster...
November 10, 2025 at 11:40 PM
For years, advertising has been how Stereogum made money to pay its writers. While we made progress moving away from that model, Google AI search has decimated the business almost overnight. Today we relaunch with a greater focus on subscription. We’d love your support: stereogum.com/2478838/ster...
Reposted by Robert Hawkins
New work to appear @ TACL!
Language models (LMs) are remarkably good at generating novel well-formed sentences, leading to claims that they have mastered grammar.
Yet they often assign higher probability to ungrammatical strings than to grammatical strings.
How can both things be true? 🧵👇
Language models (LMs) are remarkably good at generating novel well-formed sentences, leading to claims that they have mastered grammar.
Yet they often assign higher probability to ungrammatical strings than to grammatical strings.
How can both things be true? 🧵👇
November 10, 2025 at 10:11 PM
New work to appear @ TACL!
Language models (LMs) are remarkably good at generating novel well-formed sentences, leading to claims that they have mastered grammar.
Yet they often assign higher probability to ungrammatical strings than to grammatical strings.
How can both things be true? 🧵👇
Language models (LMs) are remarkably good at generating novel well-formed sentences, leading to claims that they have mastered grammar.
Yet they often assign higher probability to ungrammatical strings than to grammatical strings.
How can both things be true? 🧵👇
Reposted by Robert Hawkins
So many nonsense ad hoc pipelines could be prevented by requiring that they work on synthetic data.
I tend to think of experiments as special cases of inference, since most of the problems I work on cannot be studied in experiments. But I get that many researchers see experiments as base analogy.
I tend to think of experiments as special cases of inference, since most of the problems I work on cannot be studied in experiments. But I get that many researchers see experiments as base analogy.
"Validate With Simulated Truth: A first habit is to test whether an analytical pipeline can recover known conditions."
Very good advice below. So much COVID nonsense (e.g. 'immunological dark matter') basically came down to a non-identifiable model that hadn't been properly tested.
Very good advice below. So much COVID nonsense (e.g. 'immunological dark matter') basically came down to a non-identifiable model that hadn't been properly tested.
Modelling Like an Experimentalist
Dahlin et al. (2024) apply experimental thinking to a model of mosquito-borne disease transmissions.
onlinelibrary.wiley.com
November 10, 2025 at 12:41 PM
So many nonsense ad hoc pipelines could be prevented by requiring that they work on synthetic data.
I tend to think of experiments as special cases of inference, since most of the problems I work on cannot be studied in experiments. But I get that many researchers see experiments as base analogy.
I tend to think of experiments as special cases of inference, since most of the problems I work on cannot be studied in experiments. But I get that many researchers see experiments as base analogy.
Reposted by Robert Hawkins
new pre-print,
"Chain of Time: In-Context Physical Simulation with Image Generation Models"
(by Wang, Bigelow, Li, and me)
arxiv.org/abs/2511.00110
"Chain of Time: In-Context Physical Simulation with Image Generation Models"
(by Wang, Bigelow, Li, and me)
arxiv.org/abs/2511.00110
November 10, 2025 at 1:57 PM
new pre-print,
"Chain of Time: In-Context Physical Simulation with Image Generation Models"
(by Wang, Bigelow, Li, and me)
arxiv.org/abs/2511.00110
"Chain of Time: In-Context Physical Simulation with Image Generation Models"
(by Wang, Bigelow, Li, and me)
arxiv.org/abs/2511.00110
Reposted by Robert Hawkins
Why do AI models struggle with social scenes? 🧐 Our new preprint with @lisik.bsky.social reveals a fundamental gap: most AI vision models lack explicit 3D pose information that humans rely on for social judgments.
Read the full work: arxiv.org/abs/2511.03988
Read the full work: arxiv.org/abs/2511.03988
Simple 3D Pose Features Support Human and Machine Social Scene Understanding
Humans can quickly and effortlessly extract a variety of information about others' social interactions from visual input, ranging from visuospatial cues like whether two people are facing each other t...
arxiv.org
November 10, 2025 at 12:38 AM
Why do AI models struggle with social scenes? 🧐 Our new preprint with @lisik.bsky.social reveals a fundamental gap: most AI vision models lack explicit 3D pose information that humans rely on for social judgments.
Read the full work: arxiv.org/abs/2511.03988
Read the full work: arxiv.org/abs/2511.03988
Reposted by Robert Hawkins
One year ago, with Stella Punselie and Bonnie McLean: The Anatomy of Iconicity, in Open Mind doi.org/10.1162/opmi...
In which we combine iconicity ratings, experimental evidence, and structure mapping to identify & explain crosslinguistic patterns of iconic words
#iconicity #cognitivescience
In which we combine iconicity ratings, experimental evidence, and structure mapping to identify & explain crosslinguistic patterns of iconic words
#iconicity #cognitivescience
November 8, 2025 at 2:24 PM
One year ago, with Stella Punselie and Bonnie McLean: The Anatomy of Iconicity, in Open Mind doi.org/10.1162/opmi...
In which we combine iconicity ratings, experimental evidence, and structure mapping to identify & explain crosslinguistic patterns of iconic words
#iconicity #cognitivescience
In which we combine iconicity ratings, experimental evidence, and structure mapping to identify & explain crosslinguistic patterns of iconic words
#iconicity #cognitivescience
congrats, @haoranzhao.bsky.social !!!
Had an amazing experience at EMNLP 2025 @emnlpmeeting.bsky.social. Glad to present my work in an oral session and honored to win the "SAC Highlight" award. Feel free to check the work below. Big thanks to my amazing advisor @rdhawkins.bsky.social!
November 9, 2025 at 4:27 PM
congrats, @haoranzhao.bsky.social !!!
Reposted by Robert Hawkins
📢 We’re looking for a researcher in in cogsci, neuroscience, linguistics, or related disciplines to work with us at Apple Machine Learning Research! We're hiring for a one-year interdisciplinary AIML Resident to work on understanding reasoning and decision making in LLMs. 🧵
November 7, 2025 at 9:19 PM
📢 We’re looking for a researcher in in cogsci, neuroscience, linguistics, or related disciplines to work with us at Apple Machine Learning Research! We're hiring for a one-year interdisciplinary AIML Resident to work on understanding reasoning and decision making in LLMs. 🧵
Reposted by Robert Hawkins
The Cornell AAUP chapter has consistently stated that any deal with the Trump administration would be strategically unwise and a betrayal of Cornell’s principles. This remains the case. At best, we can say that this deal could have been worse.
Read our full statement here:
Read our full statement here:
Statement on Cornell’s agreement with federal government
The Cornell AAUP chapter has consistently stated that any deal with the Trump administration would be strategically unwise and a betrayal of Cornell’s principles. This remains the case. We ar…
aaup-cornell.org
November 7, 2025 at 11:24 PM
The Cornell AAUP chapter has consistently stated that any deal with the Trump administration would be strategically unwise and a betrayal of Cornell’s principles. This remains the case. At best, we can say that this deal could have been worse.
Read our full statement here:
Read our full statement here:
Reposted by Robert Hawkins
As I mentioned at my talk today for the @stanfordaaup.bsky.social teach-in, over 4,500 federal grants have been terminated at over 600 institutions. Those grants are valued at~$8B, with ~$3.5B being stolen from universities when those grants were terminated. The type of grants terminated varies +
November 7, 2025 at 9:18 PM
As I mentioned at my talk today for the @stanfordaaup.bsky.social teach-in, over 4,500 federal grants have been terminated at over 600 institutions. Those grants are valued at~$8B, with ~$3.5B being stolen from universities when those grants were terminated. The type of grants terminated varies +
this moment calls for deep solidarity — a kind of collective politics faculty haven't been used to practicing. proud to stand with colleagues across stanford and across the country today.
We’ve started our Teach-In as part of the @aaup National Day of Action with @palumboliu.bsky.social talking about the implications of institutional neutrality in this moment. If we don’t engage with topics such as immigration, freedom of speech, etc, we are letting our whole community down.
November 7, 2025 at 9:31 PM
this moment calls for deep solidarity — a kind of collective politics faculty haven't been used to practicing. proud to stand with colleagues across stanford and across the country today.
Reposted by Robert Hawkins
Hope the folks in our Stanford community can join us for a teach-in this coming Friday as part of the @aaup.org National Day of Action for Higher Ed! 12-1pm at White Plaza. Please share!
November 6, 2025 at 12:17 AM
Hope the folks in our Stanford community can join us for a teach-in this coming Friday as part of the @aaup.org National Day of Action for Higher Ed! 12-1pm at White Plaza. Please share!
Reposted by Robert Hawkins
Well over 150 UC law faculty have just made public this open letter to the UC Regents arguing, point by specific point, for why the Regents should not accept any of the major demands the Trump administration has made of UCLA.
We argue it's not a genuine settlement offer, but a form of extortion.
We argue it's not a genuine settlement offer, but a form of extortion.
Home
UC Law Faculty to Regents: Resist the Unlawful Demands
sites.google.com
November 7, 2025 at 8:25 PM
Well over 150 UC law faculty have just made public this open letter to the UC Regents arguing, point by specific point, for why the Regents should not accept any of the major demands the Trump administration has made of UCLA.
We argue it's not a genuine settlement offer, but a form of extortion.
We argue it's not a genuine settlement offer, but a form of extortion.
Reposted by Robert Hawkins
Noise ceilings are really useful: You can estimate the reliability of your data and get an index of how well your model can possibly perform given the noise in the data.
But, contrary to what you may think, noise ceilings do not provide an absolute index of data quality.
Let's dive into why. 🧵
But, contrary to what you may think, noise ceilings do not provide an absolute index of data quality.
Let's dive into why. 🧵
November 7, 2025 at 2:58 PM
Noise ceilings are really useful: You can estimate the reliability of your data and get an index of how well your model can possibly perform given the noise in the data.
But, contrary to what you may think, noise ceilings do not provide an absolute index of data quality.
Let's dive into why. 🧵
But, contrary to what you may think, noise ceilings do not provide an absolute index of data quality.
Let's dive into why. 🧵
increasingly convinced that syntactic knowledge is not (primarily, straightforwardly) implemented by specific attention heaeds in transformers, but is instead much more distributed. congrats to @taka-yamakoshi.bsky.social on a great paper!
I’m excited to share our Findings of EMNLP paper w/ @cocoscilab.bsky.social , @rtommccoy.bsky.social, and @rdhawkins.bsky.social !
Language models, unlike humans, require large amounts of data, which suggests the need for an inductive bias.
But what kind of inductive biases do we need?
Language models, unlike humans, require large amounts of data, which suggests the need for an inductive bias.
But what kind of inductive biases do we need?
November 7, 2025 at 3:17 PM
increasingly convinced that syntactic knowledge is not (primarily, straightforwardly) implemented by specific attention heaeds in transformers, but is instead much more distributed. congrats to @taka-yamakoshi.bsky.social on a great paper!
Reposted by Robert Hawkins
We’re hiring!
Our new resume portal is live, and we’re looking for top talent in NYC to help build this administration and deliver on our affordability agenda.
Could that be you? Apply using the link below.
transition2025.com/apply
Our new resume portal is live, and we’re looking for top talent in NYC to help build this administration and deliver on our affordability agenda.
Could that be you? Apply using the link below.
transition2025.com/apply
November 6, 2025 at 1:36 PM
We’re hiring!
Our new resume portal is live, and we’re looking for top talent in NYC to help build this administration and deliver on our affordability agenda.
Could that be you? Apply using the link below.
transition2025.com/apply
Our new resume portal is live, and we’re looking for top talent in NYC to help build this administration and deliver on our affordability agenda.
Could that be you? Apply using the link below.
transition2025.com/apply
Reposted by Robert Hawkins
New preprint led by @pablooyarzo.bsky.social together with @kohitij.bsky.social, Diego Vidaurre & Radek Cichy.
Using EEG + fMRI, we show that when humans recognize images that feedforward CNNs fail on, the brain recruits cortex-wide recurrent resources.
www.biorxiv.org/content/10.1... (1/n)
Using EEG + fMRI, we show that when humans recognize images that feedforward CNNs fail on, the brain recruits cortex-wide recurrent resources.
www.biorxiv.org/content/10.1... (1/n)
www.biorxiv.org
November 7, 2025 at 9:39 AM
New preprint led by @pablooyarzo.bsky.social together with @kohitij.bsky.social, Diego Vidaurre & Radek Cichy.
Using EEG + fMRI, we show that when humans recognize images that feedforward CNNs fail on, the brain recruits cortex-wide recurrent resources.
www.biorxiv.org/content/10.1... (1/n)
Using EEG + fMRI, we show that when humans recognize images that feedforward CNNs fail on, the brain recruits cortex-wide recurrent resources.
www.biorxiv.org/content/10.1... (1/n)
Reposted by Robert Hawkins
I’m excited to share our Findings of EMNLP paper w/ @cocoscilab.bsky.social , @rtommccoy.bsky.social, and @rdhawkins.bsky.social !
Language models, unlike humans, require large amounts of data, which suggests the need for an inductive bias.
But what kind of inductive biases do we need?
Language models, unlike humans, require large amounts of data, which suggests the need for an inductive bias.
But what kind of inductive biases do we need?
November 7, 2025 at 9:17 AM
I’m excited to share our Findings of EMNLP paper w/ @cocoscilab.bsky.social , @rtommccoy.bsky.social, and @rdhawkins.bsky.social !
Language models, unlike humans, require large amounts of data, which suggests the need for an inductive bias.
But what kind of inductive biases do we need?
Language models, unlike humans, require large amounts of data, which suggests the need for an inductive bias.
But what kind of inductive biases do we need?
Reposted by Robert Hawkins
a couple years ago i contracted with one of subway's ad agencies for a small rotating sandwiches collab that ultimately they decided not to publish and i never did anything with. today feels like a good day to post one of them. here's my favorite subway sandwich, the veggie delight
November 6, 2025 at 8:23 PM
a couple years ago i contracted with one of subway's ad agencies for a small rotating sandwiches collab that ultimately they decided not to publish and i never did anything with. today feels like a good day to post one of them. here's my favorite subway sandwich, the veggie delight