Gordon Pennycook
banner
gordpennycook.bsky.social
Gordon Pennycook
@gordpennycook.bsky.social

Associate Professor, Psychology @cornelluniversity.bsky.social. Researching thinking & reasoning, misinformation, social media, AI, belief, metacognition, B.S., and various other keywords. 🇨🇦

https://gordonpennycook.com/ .. more

Gordon Robert Pennycook is a Canadian psychologist who is an associate professor at Cornell University. He is also an adjunct professor of Behavioural Science at the University of Regina's Hill and Levene Schools of Business. In 2020, he was elected to be a member of the Royal Society of Canada’s College of New Scholars, Artists, and Scientists. .. more

Political science 22%
Sociology 21%
Pinned
🚨 Now out in Psych Science 🚨

We report an adversarial collaboration (with @donandrewmoore.bsky.social) testing whether overconfidence is genuinely a trait

The paper was led by Jabin Binnendyk & Sophia Li (who is fantastic and on the job market!) Free copy here: journals.sagepub.com/eprint/7JIYS...

Also, secondarily, I made the mistake of focusing on the camera and not looking at myself... and hence did not realize what my face was doing.

I did an interview today while on vacation... and the only room free from the intrusion of children was my in-law's laundry room 😂

If you're curious (it's about using AI for political persuasion): www.cbc.ca/player/play/...

One reasonable guess is that censoring this report will actually cause *more* people to watch it and hear about it.
The full spiked 60 Minutes CECOT package, clean & subtitled. 1/5
The full spiked 60 Minutes CECOT package, clean & subtitled. 1/5

Reposted by Gordon Pennycook

"Some people do believe that they are able to perform relatively well on tasks even when there is little reason for that confidence. Our results support the claim that overconfidence might be a trait." pubmed.ncbi.nlm.nih.gov/41401360/
The Epstein file is the only thing Trump has ever taken his name off of

Thank you!!

Nothing particularly insightful to say about that, but I do think it's a great question a very promising direction for future research! (Also possible that there's already some work on this that I'm not aware of)

Ah, interesting! This paper (www.cambridge.org/core/journal...) looks at a broader set of associations and the correlations are quite modest

For estimated performance, there aren't consistent gender differences. However, when it comes to overplacement (thinking you're better than others on the task), men are more overconfident than women

Thanks to Cory Clark and Phil Tetlock and their "Adversarial Collaboration Project" for supporting this work! web.sas.upenn.edu/adcollabproj...
Research Team | Adversarial Collaboration Project
web.sas.upenn.edu

Ultimately, it may have taken several papers and much back-and-forth to come to a similar conclusion. And, most likely, we might never have come to a genuine agreement if we had not engaged in this adversarial collaboration together. I think it's a great approach to science!

Nonetheless, we came to an agreement on the following conclusion: "some participants are more confident than others across the seemingly random tasks we offered. These relationships provide support for a within-individual consistency and a trait interpretation of confidence and
overconfidence."

@donandrewmoore.bsky.social et al remain skeptical about trait overconfidence (see the discussion section of the article), but future work will hopefully help evaluate whether the tendency to be confident on these types of "unconfounded" tasks is predictive of real-world instances of overconfidence

Although performance on these tasks is (essentially) random, some ppl just think they are doing better than others. This is what I would like to call "generalized" overconfidence: The extra boost of confidence that people who tend to be overconfident bring with them to various situations/contexts.

We then tested whether the tendency to indicate confidence when it is not justified is consistent across the 8 tests that we devised. The results were straightforward: The tendency was very consistent across tasks. Factors analyses supported a single factor.

To test whether overconfidence is genuinely a trait, we created several tasks where ppl have no basis for confidence (e.g., guessing whether someone has an innie or outie belly button based on a pic of their face). We pretested them to make sure that they were unfamiliar tasks that involved guessing

Before I talk about the results, I just want to say that it was such a pleasure to work with @donandrewmoore.bsky.social more & his team. This ad collab did not require a moderator & it was very pleasant throughout! Don, of course, is extremely reasonable (my highest possible praise)!

Don was justifiably skeptical that this was genuinely measuring an underlying trait, as opposed to (once again, based on past work) an idiosyncratic form of overconfidence that would not strongly correspond with other measures of overconfidence.

So, we did an ad collab to test it further!

To address this concern, Jabin & I created a task where people have to, essentially, guess the answer ("the Generalized Overconfidence Task"). Even though people are guessing and confidence is not related to performance, some people are *still* more confident. And this predicts a bunch of things.

Ppl tend to be more overconfident for tasks that they are bad at and less overconfident for tasks that they are good at. I.e., Task performance confounds our estimates of *trait* overconfidence. Thus, as Don has pointed out many times, overconfidence measures are highly unreliable.

The ad collab emerged from a discussion with Don at @sjdm-tweets.bsky.social about this paper: www.cambridge.org/core/journal...

In the paper, Jabin & I try to address a longstanding problem for measuring overconfidence: One's level of overconfidence is highly dependent on the task in question.
Individual differences in overconfidence: A new measurement approach | Judgment and Decision Making | Cambridge Core
Individual differences in overconfidence: A new measurement approach - Volume 19
www.cambridge.org

Reposted by Gordon Pennycook

Meta-analysis shows autonomy-supportive parenting is positively related to child well-being and psychologically controlling parenting is predicts child ill-being. Effects hold across regions of the world & various measures of cultural ideas. #PsychSciSky #AcademicSky #EduSky
doi.org/10.1037/amp0...
Canadian job 👀 🇨🇦 ‼️ Second search this year! Our Dept of Psychology is hiring again, this time at the assoc/full level, for a prestigious, endowed position. Focus is social/affective neuroscience in any area (clinical, social, cog, or dev). Apply by Jan 30! Reposts, pls www.queensu.ca/psychology/n...
Employment Opportunities | Department of Psychology
Available Positions Position Title Position Details Posting Date Closing Date Queen's University National Scholar (QNS) Senior Faculty Position in Social and/or Affective Neuroscience, Department of P...
www.queensu.ca
one of the funniest pieces The Onion has done all year, and it's been a GOOD YEAR.

theonion.com/oprah-pursue...
Oprah Pursues Dr. Phil On Ship Through Arctic
THE ARCTIC CIRCLE—With a vow to destroy the abomination she had created if it was the last thing she ever did, television host Oprah Winfrey has spent weeks on a ship pursuing Dr. Phil through the Arc...
theonion.com
Motivated reasoning is a well-understood phenomenon - or is it?

In a new paper just published at @collabrapsychology.bsky.social we discuss three known unknowns.

doi.org/10.1525/coll...

Here is a 🧵
Known Unknowns in Motivated Reasoning: A Closer Look at Three Open Questions
Motivated reasoning denotes the phenomenon that individuals are more likely to arrive at conclusions that they want to arrive at. Properly understanding this phenomenon requires at least three things:...
doi.org
This is so disgusting.
🚨 New in Nature+Science!🚨
AI chatbots can shift voter attitudes on candidates & policies, often by 10+pp
🔹Exps in US Canada Poland & UK
🔹More “facts”→more persuasion (not psych tricks)
🔹Increasing persuasiveness reduces "fact" accuracy
🔹Right-leaning bots=more inaccurate