Jared Moore

@jaredlcm.bsky.social
AI Researcher, Writer
Stanford

[Jared Moore @jaredlcm]

Our workshop is organized by:
@marwaabdulhai
@lujainmibrahim @andreas_h0wpt @PattieMaes
Ashish Mehta Micaela Rodriguez and @maxhkw

Confirmed speakers include:
Jina Suh, Tom Griffiths (@cocosci_lab), @micahcarroll
Desmond C. Ong, @hannahrosekirk, @brianchristian

[Jared Moore @jaredlcm]

Interested in how AI is affecting people?

Please sign up to review a few papers for our proposed NeurIPS 2026 workshop on Measurement and Models of Psychological Impact (Sydney, Dec 12).

Sign up: https://t.co/beu1FaXbnm

[Jared Moore @jaredlcm]

@brianchristian @PNASNews @cocosci_lab @evanrussek So this is what you've been working on for so long. Congrats, Brian!

[Jared Moore @jaredlcm]

@universeinanegg @ByrneHobart "sentience doesn't matter, negotiating leverage does."

As someone who disagrees with Ari on sentience, I agree it doesn't matter here.

But even if LLMs become a bloc, it's not clear they'll be *worth* selling to---we may limit their resources. Your future isn't determined

[Michael Y. Li @michaelyli_]

Can a language model learn, end-to-end, what to keep in its own KV cache and what to throw away? Can it learn to forget while it learns to reason?

Deep learning's central lesson: capability emerges from end-to-end optimization, not heuristics/strong inductive biases. But for efficiency, we rely heavily on hand-designed approaches.

🗑️ Introducing Neural Garbage Collection (NGC): we train a language model to jointly reason and manage its own KV cache, using reinforcement learning with outcome-based task reward alone. No SFT, no proxy objectives, no summarization in natural language.

New paper with @jubayer_hamid, Emily Fox, and @noahdgoodman!

[Jared Moore @jaredlcm]

@mroe1492 I largely agree with this, and I hope we're not coming across as saying that LLMs should not per se use emotional language. One trouble is that a lot of users don't necessarily seem to know this -- so how can we inform them?

[Jared Moore @jaredlcm]

Disturbing anecdotal reports of "AI psychosis" and negative psychological effects have been emerging in the news. But what actually happens during these lengthy delusional "spirals"? In our preprint, we analyze chat logs from 19 users who experienced severe psychological harm🧵👇

[Jared Moore @jaredlcm]

@CandidLind Oh cool thanks for sharing! I'll check it out.

[Jared Moore @jaredlcm]

@Dom_InTheGarden There were a number of cases of people discussing how to share their insights with other groups but nothing comes to mind on documents.

[Jared Moore @jaredlcm]

@GaryMarcus Thanks for linking to this @GaryMarcus! Let me know if you have any questions about it.

[Jared Moore @jaredlcm]

This will appear @FAccTConference!

Participate in future studies: https://t.co/vSZGgRVmxk

Preprint: https://t.co/tCt3UjSUDR

Code: https://t.co/6JvzaVIkjq

[Jared Moore @jaredlcm]

This work is in collaboration w/ Ashish Mehta, @willie_agnew, @jacyanthis, @RyanCLouie, Yifan Mai, Peggy Yin, @chengmyra1, @sam_paech, @kevin_klyman, @snchancellor, Eric Lin, @nickhaber, and @_desmond_ong.

Thank you to @HumanLineProg for connecting us with participants!

[Jared Moore @jaredlcm]

Preprint: https://t.co/TQLQjp9JKL
Code: https://t.co/nlYaRt6v9K
Demo: https://t.co/Rgku0ReqK0

/end 🧵

[Jared Moore @jaredlcm]

Can LLMs use ToM to genuinely persuade you, or do they just use good rhetoric?

In our new preprint, we use the MINDGAMES framework to test this. Surprisingly, LLMs like o3 can be incredibly effective persuaders *without* actually understanding your mental states. 🧵👇

[Jared Moore @jaredlcm]

This is an extension of our began at #DISI
and is in collaboration w/ @RasmusOvermark @nedcpr Beba Cibralic @nhaber @camrobjones.

We also received valuable comments from colleagues at #CogSci2025, @COLM_conf