iiinova — where your digital self keeps evolving

Wiktoria Kozyra @wiktoriakozyra.bsky.social · 19/09/2026

New commentary! Are scientists and laypeople alike under persistent illusions of understanding? Shiffrin, Stigler, & Keil (2026) seem to think so. I argue this isn't the case w/ @sharonchungsl.bsky.social, Adam Harris, @davidlagnado.bsky.social & @singmann.bsky.social: rdcu.be/3swzbgH3ePva

rdcu.be

12513

Reposted by Kevin O’Neill

I’m very grateful to @smfleming.bsky.social for his supervision and @kevingoneill.github.io for a great collaboration. We show that the PE bias in confidence judgments can be a Bayes-optimal consequence of inference in high-dimensional spaces, and in CNNs, the bias also increases with dimensionality

1157

Kevin O’Neill @kevingoneill.github.io · 18/09/2026

the DOI for the target article is linked as “original paper” in last post of my thread! your link is a preprint of the same paper

010

this doesn’t usually result in a hard non-identifiability, so in principle I’d say retain the levels. but esp with multilevel models can make the gradients/posterior geometry difficult, so in practice this may not be possible

020

because the modeled response probabilities depend on the area between thresholds, a zero probability category will either result in two successive thresholds being approximately equal (for middle categories), or thresholds that approach +/- infinity. ditto for adjacent empty cells

130

original paper: doi.org/10.1007/s421... our commentary: rdcu.be/3swzbgH3ePva all commentaries: link.springer.com/collections/... response to commentaries: doi.org/10.1007/s421...

Of course, we agree that people often get things wrong. But that doesn't mean that scientific understanding is an illusion, and we think such extraordinary claims must be supported by evidence

100

3. SSK present people as inherently bad at causal reasoning, but our best models of causal cognition are normative! No matter whether we look at causal selection, causal inference, or diagnostic reasoning, the data suggest that people tend to make the best inferences possible under the circumstances

120

2. SSK present causal understanding as shallow and ever incomplete, but causal theory tells us exactly what information can be learned from experimental data! When we discover mechanisms or alternative causes, that doesn't invalidate our previous theories, it supplements them

1. SSK present science as a theory-free enterprise, but we have an entire field dedicated to causal inference! Notably, their examples dissolve if one specifies a research question, fixes an estimand, and deploys a valid causal identification strategy in line with modern practice

100

Putting aside the optics of making such a claim as right-wing forces use alarmist metascience to dismantle research institutions, as scientists studying causal cognition, we think that these kinds of arguments should be supported by empirical evidence. We make three counter-arguments:

100

After backing up this claim up with a few familiar statistical examples (Simpson's paradox, Lord's paradox, Stein's paradox, and regression to the mean), SSK map out a partial list of the kinds of illusions that scientists purportedly find themselves under & propose ways to resist these illusions

In their paper, SSK claim that scientists are susceptible not only to misunderstandings, but also to *illusions of understanding* in which we deceive ourselves into thinking we have better causal explanations for phenomena than we do

120

joannad-c.bsky.social @joannad-c.bsky.social · 16/09/2026

rdcu.be

12513

Reposted by Kevin O’Neill

Hiring! 🧠❤️🤔If you’re a philosopher working in ethics/ x-phi/ empirically informed moral philosophy / moral psychology / applied ethics, consider applying for a 2.5 year research postdoc with me in my lab at Oxford. Deadline noon 16th Oct. Details 👉 philjobs.org/job/show/32101 Hit me up with q’s!

static.klipy.com

Shameless Fiona Gallagher

ALT: Shameless Fiona Gallagher

12924

Reposted by Kevin O’Neill

Rochelle Kaper @rochellekaper.bsky.social · 16/09/2026

PREPRINT UPDATE🥳👇: 🧠learning under uncertainty & metacognitive processes are not as generalizable as previously thought! both evolve differently depending on structure & stimulus set osf.io/preprints/ps...

osf.io

OSF

14214

Reposted by Kevin O’Neill

Micah G. Allen @micahgallen.com · 16/09/2026

How do you model what your heartbeat feels like? Our new Bayesian toolbox makes models of cardiac and respiratory interoception easier to use and interpret, with familiar R formulas and worked tutorials. New publication 🧪: link.springer.com/article/10.3...

link.springer.com

Hierarchical Bayesian modeling of interoceptive psychophysics - Behavior Research Methods

Interoception, the capacity to sense, perceive, and metacognitively appraise viscerosensory and homeostatic signals, is a growing focus in psychology and psychiatry. Adaptive psychophysical tasks now ...

1257

Reposted by Kevin O’Neill

Martin Modrák @modrakm.bsky.social · 14/09/2026

I will be looking for a PhD student in Bayesian statistics (in collaboration with Paul Bürkner), located in Prague. Forward to students if you have somebody who could be interested. www.martinmodrak.cz/grammo-call/

martinmodrak.cz

Grammo - Let's Research Bayes!

03224

so, even with *no* data, there is nothing stopping us from drawing on a range of methods, frameworks, and assumptions to do integrative & interdisciplinary science

160

as a new postdoc, my role was mostly to add some related derivations and additional simulations. and even though all of these paths relied on different frameworks & assumptions, they all converge on the same result and mutually inform each other!

150

on a final note, one thing I really like about this project is that it showcases a breadth of formal tools for cognitive science. when I joined the project, @wiktoriakozyra.bsky.social and @smfleming.bsky.social already had a number of convincing simulations and model fitting experiments

160

in this case, if people choose to represent additional possibilities that are unavailable for response, their behavior looks a lot like suboptimal heuristic models

160

I won’t get into details, but there is a strong take home message: as experimentalists, we can present participants with a task. but how they choose to represent that task and reason over it is up to them!

150

Mediterranean Society for Consciousness Science @mesec-community.bsky.social · 04/09/2026

delighted to have been able to chip in on this newly published paper by @wiktoriakozyra.bsky.social & @smfleming.bsky.social on how confidence “biases” arise from normative Bayesian models!

1157

Reposted by Kevin O’Neill

How can students and postdocs shape their field of research? We're excited to share our TiCS piece, showcasing MESEC as a case study for exactly this question! So grateful for our wonderful community of MESECeers 🤗 doi.org/10.1016/j.ti... *Free access* link: authors.elsevier.com/a/1njPQ4sIRv...

doi.org

Redirecting

02511

Reposted by Kevin O’Neill

Nina Van Rooy @ninavanrooy.bsky.social · 27/08/2026

Next week Saturday, I am giving a public lecture on philosophy of animal minds at NC State! Full details: What: lecture by Nina Van Rooy, 'Human-Centered Bias in the Study of Animal Minds' When: Saturday Sept 5th, 11 am - 1 pm Where: Talley Student Union, Room 3285, NC State University, Raleigh

011

Reposted by Kevin O’Neill

Nina Van Rooy @ninavanrooy.bsky.social · 14/08/2026

The ball is rolling! Really excited that our paper showing no norm effects on causal judgments with continuous causes is now up as a preprint! See the thread by @kevingoneill.github.io in the post below for an overview of what we do in this paper. #psychology #x-phi

021

Kevin O’Neill @kevingoneill.github.io · 17/08/2026

having almost made this mistake myself, I’m glad somebody wrote this paper!

120

Reposted by Kevin O’Neill

Aki Vehtari @avehtari.bsky.social · 13/08/2026

All four books I've co-authored are freely available online for non-commercial use: - Bayesian Workflow at avehtari.github.io/Bayesian-Wor... Links to other three books are in the quoted post 👇 (too many books to fit in one post!)

avehtari.github.io

Bayesian Workflow book: Website – Bayesian Workflow book

Website for the Bayesian Workflow book by Gelman, Vehtari, McElreath, et al. — case studies, code, and exercises in R and Stan.

8283133

Reposted by Kevin O’Neill

Nora Harhen @noraharhen.bsky.social · 11/08/2026

new preprint! in my final grad school project led by @kmiceli.bsky.social & @ninavanrooy.bsky.social w/@felipedebrigard.bsky.social, we challenge ~20yrs of causal cognition research by finding a context in which people's causal judgments are completely *unaffected* by norms! osf.io/preprints/ps...

osf.io

OSF

2209

Reposted by Kevin O’Neill

How do we make good decisions with little experience? Adults can rely on memories of past choices. In our new study in @pnas.org we show kids & teens do something similar–but what they remember turns out to be quite different! www.pnas.org/doi/10.1073/...

pnas.org

Developmental changes in memory structure and precision alter the use of retrieved episodes during decisions for reward | PNAS

Most widely studied option evaluation strategies rely on knowledge accumulated across repeated experiences. But how should options be evaluated in ...

29340

I'm really excited about this project and would love to hear any thoughts!

010

most importantly, we think our results help shift the explanatory target from not only *that* norms influence causal judgments, but *when* and *in which contexts* norms matter. they also show that it will take some hard work to generalize our theories of causal judgment to more realistic domains

130

while we don't have a full model, participants' written reports point to a focus on actual contribution over necessity and sufficiency. though some might interpret this as evidence for productive concept of causation, we are optimistic that difference-making theories can stand up to this challenge

100

this is a stark contrast from lots of research finding the opposite pattern, constraining the supposed fundamental and pervasive role of normality in modal thought. so what gives?

100

nevertheless, participants gave the same causal judgments of C regardless of whether it violated a statistical norm or not! they were also highly confident regardless of normality, indicating that this is not because they were confused about the task

a violin plot showing that (a) causal judgments and (b) confidence were the same regardless of whether the focal cause was normal or abnormal

100

we then presented participants with a final case where C and A contributed equal amounts to E. after judging the extent to which C caused E, participants rated how surprised they were that C and A took the values they did, confirming that they saw C as violating a statistical norm

A graph showing ratings of surprisal. Ratings were higher for the focal cause when it was abnormal, but lower for the alternate cause and for the focal cause when it was normal

100

next, we had participants learn the structure (within a cover story) and the distributions of C and A over 40 trials. the variables were set so that C either had a lower variance (experiment 1) or mean (experiment 2) to A, and manipulation checks confirmed that participants learned this

A line graph showing ratings of perceived variability by block. Participants learned that the focal cause had lower variability when it was abnormal

A line graph showing ratings of perceived variability by block. Participants learned that the focal cause had a lower mean when it was abnormal

100

but all of this work focuses on judgments of *binary* causes: events that either happen or don't. what happens when candidate causes can contribute in degrees? to find out, we adapted a well-studied binary causal structure (b) so that (c) E happens if C+A > theta (where theta is some threshold)

(a) a causal structure with C -> E <- A. (b) a binary version of the structure where C occurs with p_C, A occurs with p_a, and E occurs if C and A occur. (c) a continuous version of the structure where C and A are normally distributed and E occurs if C + A > theta, where theta is some threshold.

researchers have now found similar effects for different causal structures (e.g., double prevention) and kinds of norms (e.g., prescriptive norms), as well as other kinds of judgments (e.g., intentionality). this has led to the notion that norms have a general and pervasive effect on modal cognition

100