#AI reasoning models may seem to reason reflectively when they say things like, "Let me rethink that".
But do these "reflective" phrases predict better reasoning performance?
Not in #Deepseek R1 Zero: https://doi.org/10.48550/arXiv.2503.20783
#AI reasoning models may seem to reason reflectively when they say things like, "Let me rethink that".
But do these "reflective" phrases predict better reasoning performance?
Not in #Deepseek R1 Zero: https://doi.org/10.48550/arXiv.2503.20783
Jon Baron shared Peter Wakker's annotated bibliography of #decisionTheory
> 9000 entries!
DocX http://personal.eur.nl/wakker/refs/webrfrncs.docx
PDF http://personal.eur.nl/wakker/refs/webrfrncs.pdf
BibTeX (no annotations, I merged redundancies) https://www.dropbox.com/scl/fi/vbkki82h62ydq0fol1g8d/Decision-Theory.bib?rlkey=84m8zx3tyaa4uy6p0zptkybnx&st=qx9myebb&dl=0
Not what I expected!
More people preferred high diagnostic uncertainty than low, even though the higher uncertainty caused more worry.
Within- and between-subject, quant and qual results: https://doi.org/10.1136/jme-2024-109932
#Microsoft's viral paper about *correlations* between #AI use and #criticalThinking also has "impact" in the title (despite admitting "Our analysis does not establish #causation").
#Confidence in #GenAI predicted LESS critical thinking.
SELF-confidence predicted MORE critical thinking.
PREDICTED ≠ CAUSED
Df. "clinical #equipoise describes a situation of evidentiary uncertainty among experts"
Opel et al. argue such equipoise is neither necessary nor sufficient for #SharedDecisionMaking.
Rather SDM is most useful where tradeoffs between options are KNOWN.
What medical problem statement features help or hinder #chatGPT's #diagnosis accuracy?
- more comorbidities and historical items hurt differentials
- more examination items and use of 3-part statement format helped
A former colleague and old friend just got laid off. She has an MS in ITDS (information technology and decision science) and a ton of experience with SQL, and is a very reliable and consistent worker. She's also comfortable with Python and C#. Please let me know of any opportunities you may know of in that area, and I'll pass it along!
#ITDS
#InformationTechnology
#DecisionScience
#SQL
#Python
#CSharp
#FediHire
#FediHired
Good news in our preprint about #polarization:
Demand for BIPARTISAN #news analysis was strong!
People in the #US preferred fact-checking teams that engaged in #AdversarialCollaboration at least as much as copartisan and/or professional teams.
Follow the manuscript or authors on #GoogleScholar: https://scholar.google.com/scholar?oi=bibs&hl=en&cluster=1017561778629165218
If you prefer a link directly to our preprint: https://doi.org/10.31219/osf.io/gp9w7
An adversarial collaboration reports that mathematical reflection tests assess #math ability about as much as anything else.
And what is the evidence that the paper's “independent” measure of reflection actually measured reflection?
Excited for 2025 presentations in #NYC at the #APA (January 8 to 11): apaonline.org/mpage/2025eastern?utm_source=mastodon&utm_medium=social&utm_campaign=fedica-Text-posts
The talk: https://researchgate.net/publication/370132037?utm_source=mastodon&utm_medium=social&utm_campaign=fedica-Text-posts
The poster: https://researchgate.net/publication/371248872?utm_source=mastodon&utm_medium=social&utm_campaign=fedica-Text-posts
I’m sharing talks and posters from this weekend’s Society for #JDM #conference over on #BlueSky: https://bsky.app/profile/byrdnick.com/post/3lbmo4zclhc2d
The first talk’s about the #jobMarket. By the way, I’m hiring a 3-year #postdoc who can do #quant #decisionScience — #WFH or on campus: https://geisinger.wd5.myworkdayjobs.com/GeisingerExternal/job/Danville-PA/Postdoctoral-Fellow---Decision-Sciences--Remote-Capable-_R-69755
@knutson_brain @Andrewpapale @tdverstynen
3/3 But it may also be time to incorporate #sociology studies of world-understanding and news gathering into decision theory practical consequences.
To my knowledge there isn't a strong connection between people studying these consequences of #propaganda and #news eco-systems with modern theories of #DecisionMaking and #DecisionScience.
But if anyone is interested, please reach out to me. It is something I'm very interested in pursuing. #ChangingHowWeChoose
Excited to share YEARS of research about how to get people to think reflectively and how reflection impacts philosophical judgments at the 2025 #APA in #NewYorkCity (January 8 to 11): https://www.apaonline.org/mpage/2025eastern
Can't make it?
- More about my talk: https://researchgate.net/publication/370132037
- More about my poster: https://researchgate.net/publication/371248872
Thanks to the #APA, James Beebe, and the Experimental Philosophy Society for the opportunity!
Suppose you ask kids about their school work *while they work on it*: will it help or hurt their learning?
Asking fifth graders about their reading or math improved pre- v. post-test learning, compared to not asking (Ns > 350):
#OpenAccess paper: https://doi.org/10.1016/j.heliyon.2024.e40055
ICYMI: a large language model was better at diagnosis than doctors?
...and letting the doctors use the large language model didn't improve doctor's diagnoses?!
#OpenAccess article in #JAMA: https://doi.org/10.1001/jamanetworkopen.2024.40969
A #health literacy intervention reduced the perceived accuracy of inaccurate #cancer headlines ...as well as *accurate* headlines.
Cheers to the authors and the journal for publishing this less-than-encouraging result: https://doi.org/10.1093/abm/kaae054
Looks like prior (rather than new) beliefs about cancer were the primary predictors of discernment about cancer headlines.
Which subreddits score higher on deliberation metrics?
A new metric of deliberation (designed to overcome prior issues) found that the following subreddits had the highest deliberation scores (across 72 subreddits)
- Subreddits that are based on geographical regions
- Political subreddits
- Sports subreddits
Reflective thinking often improves judgment and decisions, but across three experiments reflective thinking didn't seem to predict FEELING better about one's decision: https://doi.org/10.1017/S1930297500009414
This may help explain why we so often underrate reflective thinking.
Emotion can inhibit reflective thinking, but are all emotions equally inhibitory?
Reflective thinking was inhibited more by a video designed to elicit disgust than a video designed to elicit fear/thrill (among about 50 people who completed 60 reflection tests): https://doi.org/10.2147/PRBM.S471624
Filed under #AI overhype:
#LLMs perform near perfectly on common reflection tests, but slightly altering the #math caused LLMs to fail, suggesting the best models (including #OpenAI's o1) are not great at reflective, mathematical reasoning.
V1 of #preprint: https://doi.org/10.21203/rs.3.rs-5273334/v1