Rachel Ryskin

ryskin.bsky.social

07 October 2026 18:36

🧠 Takeaway: we understand speech by reasoning in real time over a generative model—noise included—updated by implicit stats & explicit causal beliefs. 🤖 HuBERT's representational geometry offers a window into that noise model & a test of AI speech models vs. human gaze. 💬 Feedback very welcome!

Rachel Ryskin

ryskin.bsky.social

07 October 2026 18:36

Finally, after the Noisy block, some participants saw "FATAL ERROR: Audio files corrupted," and the experimenter "fixed" it. These listeners tended to show lower τ in the next block than those given no explanation. The noise can be "explained away!"

Rachel Ryskin

ryskin.bsky.social

07 October 2026 18:35

🧵 New preprint! Speech is noisy, yet we usually understand it. I used eye-tracking + a self-supervised speech model (HuBERT) to show that speech comprehension unfolds as real-time Bayesian "noisy-channel" inference & that listeners tune their noise model to their environment 👇 osf.io/preprints/ps...

Emma James

emljames.bsky.social

01 October 2026 11:51

We have 3 departmental PhD studentships available for 2027 entry, alongside opportunities to apply for ESRC funding. I'm keen to support applications relating to cognitive mechanisms of word learning over adolescence. Please share and get in touch if interested! www.findaphd.com/phds/project...

Open Encyclopedia of Cognitive Science bot

oecs-bot.bsky.social

04 September 2026 18:45

Brain research reveals that our neural systems treat human interaction as a distinct computational process, separate from general reasoning. Social Cognitive Neuroscience by Rebecca Saxe https://doi.org/10.21428/e2759450.0276fa5e #CognitiveScience