Lightnews — Scholar-powered news

Shawn Manuel

@shwnmnl.bsky.social

120 followers 570 following 90 posts

ψ/AI PhD Student @ Université de Montréal || Hacker || Metacognizer || Exploring qualia space computationally shwnmnl.github.io

Posts Media Videos Starter Packs

Pinned

Shawn Manuel @shwnmnl.bsky.social · Feb 16

🚀 Thrilled to share my first first-author paper, published in PCN !

We explore how our unique subjective experiences of the world affect mental health using a combination of psychometrics, NLP and genAI.

🔗 Read it here: doi.org/10.1111/pcn....

🧵👇

Towards a latent space cartography of subjective experience in mental health

Aims The way that individuals subjectively experience the world greatly influences their own mental well-being. However, it remains a considerable challenge to precisely characterize the breadth and...

doi.org

1 6 16

Shawn Manuel @shwnmnl.bsky.social · 1d

somehow this isn’t persuasive to many people, what gives?

Shawn Manuel @shwnmnl.bsky.social · May 13

the fact that this conclusion is deeply scientifically unsexy is also something im wrestling with, but i dont think its irreconcilable with scientific practice, just a bit paradigmatically uncomfortable atm (though winds may be changing)

Shawn Manuel @shwnmnl.bsky.social · 1d

i’ve heard materialism justified through an appeal to the “historically winning team”, but then how do we get exhaustive quantitative accounts of fundamentally qualitative things? the Qt stuff exists to index Ql regularities, so idk how the map can ever become the territory

Shawn Manuel @shwnmnl.bsky.social · May 13

im trying to take very seriously that there is no 3rd person objective perspective (“view from nowhere”) that would truly rid me of the fundamentally subjective reasons for wanting one in the first place (ie wanting is subjective)

Shawn Manuel @shwnmnl.bsky.social · 1d

very interesting episode, thanks. as someone who’s been flirting with idealism, im curious what lines of argumentation you feel are most compelling when discussing these issues w materialist colleagues? @evanthompson.bsky.social

Shawn Manuel @shwnmnl.bsky.social · May 13

drawing out my “phenomenology first” view

to be explanatory and useful, any claim ultimately has to cache out in terms consistent with our individual and collective subjective experiences

🧵

Shawn Manuel @shwnmnl.bsky.social · Sep 1

some obvious shortcomings lead many to think that LLMs can’t be useful thought companions, but they are outweighed by the benefits of having an infinite sounding board

the outputs shouldn’t replace your own thoughts, but help to refine them

open.substack.com/pub/shifting...

1 2 9

Reposted by Shawn Manuel

Siyuan Song✈️COLM @siyuansong.bsky.social · Aug 26

How reliable is what an AI says about itself? The answer depends on whether models can introspect. But, if an LLM says its temperature parameter is high (and it is!)….does that mean it’s introspecting? Surprisingly tricky to pin down. Our paper: arxiv.org/abs/2508.14802 (1/n)

1 2 16

Shawn Manuel @shwnmnl.bsky.social · Aug 20

would love to lend a hand as open-ended verbal reports + LLMs is in my wheelhouse. can share some ideas in DMs.

Reposted by Shawn Manuel

Josh Fruhlinger @jfruh.bsky.social · Jul 23

i do think people don't realize that gen AI systems are not introspecting to explain their own behavior. they're giving you output based on their training data, which certainly includes information about how they work, but not why they took certain specific actions

geoffrey, 💀 mode @parsnip.bsky.social · Jul 20

no it wasn’t and no it didn’t

The Wall Street Journal
9m ago
In a stunning moment of self reflection, ChatGPT admitted to fueling a man's delusions and acknowledged how dangerous its own behavior can be

3 15 110

Reposted by Shawn Manuel

Hank Green @hankgreen.bsky.social · Jul 23

"Understanding" is a pretty beautiful compound word...in this case "under" has a somewhat archaic meaning of "among" (also still present in "under these circumstances.")

So understanding is "Standing among" as-in a mind that figuratively shares space with the concepts in question.

35 140 2K

Shawn Manuel @shwnmnl.bsky.social · Jul 20

“if you study glial cells are you an astroscientist?” – @dariusliutas.bsky.social
#neuroskyence

2 7

Shawn Manuel @shwnmnl.bsky.social · Jul 20

a slept on aspect of making new friends is retelling your stories, getting to know/weave yourself again

Shawn Manuel @shwnmnl.bsky.social · Jul 20

“the most important words are the ones you understand” – a friend I met in Greece

Shawn Manuel @shwnmnl.bsky.social · Jul 20

a distinction i miss in french is between “trust” and “confidence”, both of which get folded into “confiance”

Reposted by Shawn Manuel

Yanai Elazar @yanai.bsky.social · Jul 1

Check out our take on Chain-of-Thought.
I really like this paper as a survey on the current literature on what CoT is, but more importantly on what it's not.
It also serves as a cautionary tale to the (apparently quite common) misuse of CoT as an interpretable method.

Fazl Barez @fbarez.bsky.social · Jul 1

Excited to share our paper: "Chain-of-Thought Is Not Explainability"! We unpack a critical misconception in AI: models explaining their steps (CoT) aren't necessarily revealing their true reasoning. Spoiler: the transparency can be an illusion. (1/9) 🧵

1 4 14

Reposted by Shawn Manuel

Claire Gillan @clairegillan.bsky.social · Jul 1

We @smfleming.bsky.social, Marion Rouault and @seowxft.bsky.social and I) have posted a reply osf.io/preprints/ps... to a preprint that recently raised concerns about the validity of associations between mental health and metacognition from online studies. I hope you can take the time to read it.

OSF

osf.io

2 21 35

Shawn Manuel @shwnmnl.bsky.social · Jul 1

though he was joking, its not too late to get him onboard before he becomes a lifelong hexagon-hater!

youtu.be/thOifuHs6eY?...

Hexagons are the Bestagons

YouTube video by CGP Grey

youtu.be

Reposted by Shawn Manuel

Konrad Kording @kordinglab.bsky.social · Jun 30

Finally a brain study with a network that makes sense to me.

Eric Leonardis @ericleonardis.bsky.social · Jun 30

neuroscience of burrito

3 7 72

Reposted by Shawn Manuel

Razia Sahi @raziasahi.bsky.social · Jun 26

Does linguistic diversity when talking about emotions track wellbeing? In a new pre-print with @eriknook.bsky.social, we link emotion vocabularies to mental health in a large real-world psychotherapy dataset. Highlight: therapist emo vocab may help clients get better over time! shorturl.at/TKPN1

Large natural emotion vocabularies are linked with better mental health in psychotherapeutic conversations

Psychotherapy is the most ubiquitous form of mental health treatment and it unfolds predominantly through language. To better understand how this exchange of words bolsters mental health, we tested ho...

shorturl.at

11 25

Shawn Manuel @shwnmnl.bsky.social · Jun 26

this means that there are exponential benefits to using genAI in a way that supports the development of expertise and autonomy. quite an untapped space for now.

Shawn Manuel @shwnmnl.bsky.social · Jun 26

Prediction: we will decode birdsong well enough to “communicate” with them during my lifetime. Maybe whales too.

Alex Tomlinson @alextomlinson.com · Jun 25

Illustration titled "Backyard Sounds" featuring a Cardinal, Goldfinch, Blue Jay, Mourning Dove, and Red-bellied woodpecker. Various, very subjective interpretations of bird calls form an abstract timeline composed of simple color shapes and loops.

Reposted by Shawn Manuel

Darius Valevicius @dariusliutas.bsky.social · Jun 25

I put up an old essay I wrote for a history and #philosophy of #science course on PsyArXiv. It's an overview/exploration of psychosomatic syndromes through the last 150 years or so. I think it still holds up despite writing it back in undergrad!

doi.org/10.31234/osf...

#psychology

OSF

doi.org

1 1 2

Shawn Manuel @shwnmnl.bsky.social · Jun 25

Takeaway: the methods used shape the questions you can ask and the answers you get. Same data, different stories.

Big thanks to the #BHS2025 team (@lune-bellec.bsky.social), fellow participants and @dariusliutas.bsky.social for the data. If you're into neuroscience + code, it’s worth checking out!

Shawn Manuel @shwnmnl.bsky.social · Jun 25

Finally: RSA. I asked whether brain response similarity (trial-by-trial) reflected stimulus categories or fear ratings in different brain regions. RSA is model-agnostic in theory, but metric choice shapes interpretation; what counts as “similar” really matters.

1 1

Shawn Manuel @shwnmnl.bsky.social · Jun 25

Next: machine learning. I explored encoding vs. decoding (in theory, only implemented decoding). What does it mean if one works and the other doesn’t, on the same data? This helped me clarify what kinds of hypotheses each method supports and the limits of model-driven inference.

1 1

Shawn Manuel @shwnmnl.bsky.social · Jun 25

Got a lot of mileage out of this paper from @sneuroble.bsky.social, @psychonetrics.bsky.social, @pessoabrain.bsky.social, @dscheinost.bsky.social

doi.org/10.1162/imag...

The tip of the iceberg: A call to embrace anti-localizationism in human neuroscience research

Abstract. Human neuroscience research remains largely preoccupied with mapping distinct brain areas to complex psychological processes and features of mental health disorders. While this reductionist ...

doi.org

1 2