Cool! Will be very interested to hear your thoughts

Not specifically (although I just got a grant on the topic, so hopefully soon 🙂)

I did a post once on the other place noting that LLMs like ChatGPT have no sense of time (ask them)

There's also this paper, with some great insights: arxiv.org/abs/1905.13469

I once met a couple of Australians who were terrified at the idea of Ticks and Lyme Disease. I countered that Australia had spiders and they said “yeah but we can SEE them!”

It isn’t?

Very cool! Any chance to play around with this yet? (I see paper says code available upon publication tho...)

Whoa, huge congratulations!!

I took a class with Rescorla and recall these findings. Curious what your take (or the students) was on them. Evidence for a temporal prior for association?

All this was the great work of Giuliana Macedo, a killer student in my lab. Huge thanks as well to Brady Roberts, Wilma Bainbridge, and Mathias Sablé-Meyer for providing their stimuli!

Beyond all this, we found that shapes/symbols with more edges were also processed faster in recurrent models of vision, and were also predicted to be more memorable.

Altogether, we think this provides evidence that visual time is guided by information content.

Here, we found that the durations of images that required more steps to generate were both dilated AND more precise. In the image below the ones on the right were seen as longer than the ones on the right.

ALSO, the effect was moderated by how subjectively meaningful subjects found the stimuli

Notably, Exp1 affected only precision, whereas Exp2 affected only accuracy. What's the link? In our final experiment we used line drawings generated by an algorithm from @mathiassablemeyer.bsky.social according to the minimum number of steps to create them.

doi.org/10.1016/j.co...

So what's going on? Do angles matter? We turned to more complicated symbols, both artificial and real:

Here, we found that symbols that contained more edges were dilated. In the attached image, the symbols on the right all seemed to last longer than the ones on the left.

osf.io/preprints/ps...

To answer this, we turned to intermediate layers of vision. We started out with the popular "Bouba Kiki" set. We found that the durations of more angular shapes ("Kiki"-like) were perceived more precisely and faster, modified by how large people perceived them and how consistently they named them.

Last year, we published a paper showing that scene images that were more memorable were also dilated. We found a link to how "fast" these images were processed through recurrent deep learning models of vision.

But, we didn't know WHY they were processed faster

www.nature.com/articles/s41...

Ok, so our sense of time is malleable, right? All kinds of stimuli alter perceived duration: bigger things seem longer than smaller things, red things seem longer than blue things. Scenes, contrast, emotion, biomotion, etc.

We can split these into "high" and "low" levels of vision