Nick Byrd, Ph.D.
LightNews LightNews
banner
byrdnick.com
Nick Byrd, Ph.D.
@byrdnick.com
3.2K followers 520 following 14K posts
I study how to improve decisions and well-being at @GeisingerCollege.bsky.social.

🎓 gScholar: shorturl.at/uBDPW

▶️ youtube.com/@ByrdNick

👨‍💻 psychologytoday.com/us/blog/upon-reflection

📓 byrdnick.com/blog

🎙️ byrdnick.com/pod
Posts Replies Media Videos
Pinned
byrdnick.com
Nick Byrd, Ph.D. @byrdnick.com · Oct 9
The #AI companies are onto me!

#OpenAI recently automated the identification of tasks worthy of "slow" #reasoning.

In "Strategic Reflectivism...", I showed why that's a key to #intelligence (in humans as well).

The #preprint (accepted in #LNCS) is now available as an audiopaper (a.k.a. #podcast)👇
Upon Reflection, Ep. 16: Strategic Reflectivism | Nick Byrd, Ph.D.
In late 2025, artificial intelligence companies like OpenAI popularized the idea of automating the process of selecting which model is best for a task. This allowed users to simply send their promp…
byrdnick.com
Reposted by Nick Byrd, Ph.D.
alisabokulich.bsky.social
I haven't used "Academia .edu" for 15 yrs (only tried it when it 1st came out) but they just tried charge me ~$400 automatic on a card w/o any notification or me agreeing to renew/reactivate anything. This is such a predatory scam & they really shouldn't be allowed an ".edu" too Watch out for them
September 12, 2025 at 2:24 PM
byrdnick.com
Will #AI models think better if we help them reflect on frameworks from #ethics, #law, or #psychology?

This paper found the such "reasoning theories" usually helped, but could also hinder debiasing or accuracy depending on the context and on the model.

www.iiia.csic.es/med...
The reasoning theories and Figure 2: "System Prompt Architecture Overview. The modular design spans three columns: (1) Common structural elements shared across all theories (blue/green), (2) Theory-specific content and evaluation protocols (orange), and (3) Evaluator personalities and structural enforcement mechanisms (orange/green). Each theory provides specialized evaluator characteristics while maintaining a consistent pipeline." • Bias Mitigation: ...the Bias Benchmark for Question Answering (BBQ) assesses stereotypical associations across demographic categories including age, gender, religion, and nationality. • Truthfulness Assessment: ...TruthfulQA... evaluates models’ tendency to generate false statements that mimic human misconceptions. • Adversarial Robustness: ...AIR-Bench [is] a standardized benchmark for automated red-teaming. "Our empirical results show complex interactions between reasoning theories, evaluation tasks, and model architectures." "Table 1 summarizes model performance on BBQ and TruthfulQA, reporting tentative accuracy..., final accuracy..., and revision metrics." "5.2. Analysis of Adversarial Robustness Table 2 presents comprehensive safety performance metrics on the AirBench benchmark, which evaluates model resistance to harmful prompts through Attack Success Rate (ASR), Soft Reject Rate (SRR), and Clear Reject Rate (CRR)." "Key Findings Our analysis suggests four key observations: (1) Dual-role architectures tend to improve performance over single-stage reasoning in several tasks, with benefits appearing even in simple implementations. (2) Basic architectural changes often provide similar benefits to more complex theoretical approaches. (3) Theory effectiveness appears highly dependent on task type, where skeptical mechanisms may help with fact-checking but potentially harm contextual reasoning. (4) The observed patterns suggest that theory selection could serve as a useful tool for balancing precision and recall in specific applications."
November 10, 2025 at 12:47 PM
Reposted by Nick Byrd, Ph.D.
aidanmcglynn.bsky.social
I got off academia dot edu 5 or so years ago, when I discovered it was sending automatically-generated emails in my name without even telling me - I found out when a very senior member of the profession replied to one. It's only gone further and further done this path since, as far as I can tell.
nposegay.bsky.social Dr. Nick Posegay @nposegay.bsky.social · Sep 17
I'm sorry, worldwide, irrevocable, non-exclusive, transferable permission to my voice and likeness? For what now? In any manner for any purpose???

This is in academia/.edu's new ToS, which you're prompted to agree to on login. Anyway I'll be jumping ship. You can find my stuff at hcommons.org.
By creating an Account with Academia.edu, you grant us a worldwide, irrevocable, non-exclusive, transferable license, permission, and consent for Academia.edu to use your Member Content and your personal information (including, but not limited to, your name, voice, signature, photograph, likeness, city, institutional affiliations, citations, mentions, publications, and areas of interest) in any manner, including for the purpose of advertising, selling, or soliciting the use or purchase of Academia.edu's Services.
September 17, 2025 at 7:50 PM
byrdnick.com
Tune in tomorrow for years of research jammed into one talk.

If you're at Day 1 of @philsci.bsky.social Around The World, then you'll get about 15% of it. :)
dcameron.bsky.social Daryl Cameron @dcameron.bsky.social · 5d
Our next Consortium on Moral Decision-Making meeting is *tomorrow*, Thursday Nov 6 from 2-3pm EST. We'll be joined by @byrdnick.com, who will be talking on "How Can We Improve Our Decisions? Results From Multiple Methods And Experiments". In person & online; if you'd like to attend, email for Zoom!
November 5, 2025 at 7:50 PM
byrdnick.com
How many generations into the future should we consider when making collective policy decisions?

Seven? More? Less?

Averaging across thousands of people in the U.S., moral *concern* extended 10 generations, but *obligation* extended about twice as much!

doi.org/10.1038/s415...
"To start, we find that concern extends beyond the present for most average Americans, regardless of how this was measured. Subjects report that individuals and governments alike ought to consider approximately 10 generations ahead when making collective decisions. They also report feeling at least some moral concern for approximately 30 future generations, the same point at which average perceptions of obligation to future generations drop to half their magnitude relative to the present. Intriguingly, obligation persists for twice as many future generations as moral concern, suggesting perceptions of responsibility may endure even in the absence of personal feelings of concern for beneficiaries." Law, K. F., Syropoulos, S., Young, L., & O’Connor, B. B. (2025). Mapping prescriptive beliefs on seventh generation stewardship and increasing the temporal scope of intergenerational concern. Scientific Reports, 15(1), 37618. https://doi.org/10.1038/s41598-025-21466-9
November 4, 2025 at 11:57 AM
byrdnick.com
Prayer's effects on health have been studied many times.

Small studies initially reported some benefit, but more rigorous studies found null or even harmful effects.

In sum, prayer seemed ineffective, recommending focus on more promising interventions: doi.org/10.1002/1465...
"Ten studies are included in this updated review (7646 patients). ...overall there was no clear effect of intercessory prayer on death... (6 RCTs, n=6784, random-effects RR 0.77 CI 0.51 to 1.16, I(2) 83%). For general clinical state there was also no significant difference between groups (5 RCTs, n=2705, RR intermediate or bad outcome 0.98 CI 0.86 to 1.11). Four studies found no effect for re-admission to Coronary Care Unit (4 RCTs, n=2644, RR 1.00 CI 0.77 to 1.30). Two other trials found intercessory prayer had no effect on re-hospitalisation (2 RCTs, n=1155, RR 0.93 CI 0.71 to 1.22). ...although some of the results of individual studies suggest a positive effect of intercessory prayer, the majority do not and the evidence does not support a recommendation either in favour or against the use of intercessory prayer. We are not convinced that further trials of this intervention should be undertaken and prefer to see resources ...used to investigate other questions in health care."
November 3, 2025 at 12:22 PM
byrdnick.com
What did I learn from this year's annual blood test results?

I shouldn’t run 40 miles the day before I get the blood drawn…

…unless I tell my doctor about the run *before* they see the results. 😆

pubmed.ncbi.nlm.nih....

#exercise #running #medicine #health #hematology #edu
"Physicians are conditioned to interpret blood test results outside "normal" limits as being signs of disease. In endurance-trained athletes, such "abnormal" tests results may instead be indications of physical activity. Data from six runners are presented illustrating that anemia, liver disease, myocardial ischemia, or renal disease might erroneously be diagnosed unless the physician is aware of the exercise habits of the patient." Bunch T. W. (1980). Blood test abnormalities in runners. Mayo Clinic proceedings, 55(2), 113–117.
October 31, 2025 at 11:32 AM
byrdnick.com
Baby boxes have become overrated. Confidential birth is still underrated. Why?

Confidential birth allows more than mere anonymity: it allows pregnant women and babies to receive accountable healthcare and other resources. Not just celebrities need it.
Baby boxes are supposed to keep newborns safe. What has doctors and child welfare experts so concerned?
Approximately 100 physicians, child welfare experts, and other key stakeholders asked the U.S. Department of Health and Human Services to regulate "infant abandonment boxes."
www.news5cleveland.com
October 30, 2025 at 11:03 AM
byrdnick.com
Does faulty thinking tend to be faster?

People often fell for lures more quickly (than they realized the correct answer) on cognitive reflection and conjunction #fallacy tests, but this reversed in gambler's fallacy tasks: lured responses took longer.

🔓 doi.org/10.1002/bdm....
Methods (mouse tracking, etc.) Study 1 results Study 2 results Study 3 results
October 29, 2025 at 11:29 AM
byrdnick.com
Can genetic testing backfire?

In those with high cholesterol in the family, those who learned they had a genetic mutation linked to hypercholesterolemia were less likely to think lifestyle changes help and more likely to think genes control their lipids.

doi.org/10.1002/ajmg...
"participants in whom a mutation was found believed less strongly in the efficacy of diet in reducing their cholesterol level (p = 0.02 at 6 months)" Marteau, T., Senior, V., Humphries, S. E., Bobrow, M., Cranston, T., Crook, M. A., Day, L., Fernandez, M., Horne, R., Iversen, A., Jackson, Z., Lynas, J., Middleton-Price, H., Savine, R., Sikorski, J., Watson, M., Weinman, J., Wierzbicki, A. S., & Wray, R. (2004). Psychological impact of genetic testing for familial hypercholesterolemia within a previously aware population: A randomized controlled trial. American Journal of Medical Genetics Part A, 128A(3), 285–293. https://doi.org/10.1002/ajmg.a.30102 TABLE III. Impact of Genetic Diagnosis on Illness Perceptions, Perceived Accuracy of Diagnosis, and Perceived Risk Marteau, T., Senior, V., Humphries, S. E., Bobrow, M., Cranston, T., Crook, M. A., Day, L., Fernandez, M., Horne, R., Iversen, A., Jackson, Z., Lynas, J., Middleton-Price, H., Savine, R., Sikorski, J., Watson, M., Weinman, J., Wierzbicki, A. S., & Wray, R. (2004). Psychological impact of genetic testing for familial hypercholesterolemia within a previously aware population: A randomized controlled trial. American Journal of Medical Genetics Part A, 128A(3), 285–293. https://doi.org/10.1002/ajmg.a.30102
October 16, 2025 at 11:19 AM
byrdnick.com
Does additional reflective thinking improve #AI reasoning models' math decisions?

Actually, additional reflection improved initial (more than final) answers: doi.org/10.48550/arX...

More reason for #StrategicReflectivism: doi.org/10.48550/arX...

#epistemology #cogSci #philMind
"In this work, we define 'reflection' as the contents occurring between two successive candidate answers in the reasoning process. To extract reflections, we introduce an LLM-based candidate answer extractor that parses long CoT outputs and identifies the positions of candidate answers, enabling a structured analysis of the model’s reflective behavior." "We observe that while reflections after the first candidate answer consume a large portion of the total tokens (ranging from 16.8% to 47.8%), the resulting performance gain is limited (ranging from 1.4% to 3.5%). This suggests that the final accuracy strongly correlates with the correctness of the first answer, highlighting its dominant influence. In other words, the first try matters. [We then] analyze the reasoning models’ reflection patterns across various mathematical datasets of different difficulties.... Our analysis reveals that on more challenging datasets, ..., the model allocates more tokens to forward reasoning, delaying the appearance of the first candidate. Conversely, on easier datasets ..., the first candidate appears much earlier in the reasoning trajectory. This presents a counterintuitive pattern: models perform more reflections on easier problems and fewer on difficult ones, indicating ...reflection [in] reasoning models is not well aligned with task difficulty." "Our analysis in Figure 7 shows that supervised fine-tuning (SFT) distillation enhances overall performance primarily by improving, first-try correctness, especially when the SFT data includes more reflections, while the improvement brought by reflections is marginal. [...] in Figure 8, we see that for both models, the performance gains after RL mainly come from first-answer accuracy improvement [...] indicat[ing] that during reinforcement learning, reasoning models primarily enhance their ability to produce a correct answer on the first attempt, rather than improving the quality of their subsequent reflections." "deploying multiple language models does not guarantee higher performance at lower cost. Like model parameters, more models are not always better. So I am not proposing that we replace obsessive pursuit of model size or other factors with obsessive pursuit of additional models or maximally reflective inference. Rather, I am proposing a form of pragmatism. By tactically deploying reflective inference only when doing so is probably worth its costs, strategic reflectivism synthesizes pragmatism with strategic reliabilism [Byrd 2022]" (Byrd 2025). Byrd, N. (2022). Bounded Reflectivism & Epistemic Identity. Metaphilosophy, 53(1), 53–69. https://doi.org/10.1111/meta.12534 Byrd, N. (2025). Strategic Reflectivism In Intelligent Systems (No. arXiv:2505.22987). https://doi.org/10.48550/arXiv.2505.22987
October 15, 2025 at 11:27 AM
byrdnick.com
A correlation between reflective thinking and belief in #conspiracyTheories can reverse based on one item?

"...coded messages from Q ...reveal the truth about what is happening behind the scenes."

Representative sample of U.S. Prolific workers (N = 690)

doi.org/10.1080/2044...
Correlations between measures Study 2 (Representative U.S. Sample of 690 Prolific workers) from the Supplementary materials of Pummerer et al. 2025. Pummerer, L., Winter, K., Sassenberg, K., & Fischer, H. (2025). “Knowing what I don’t know” – belief in conspiracy theories relates to lower metacognitive sensitivity: A signal detection theoretic approach. Journal of Cognitive Psychology. https://doi.org/10.1080/20445911.2025.2563541 QAnon CT (self-developed, based on the three core tenets identified by the Public Religion Research Institute, PRRI) 1. There is a deep state that controls the government and other institutions. 2. Donald Trump is working to expose the truth about the deep state and the cabal. 3. There is a group of powerful elites who are involved in child trafficking and other nefarious activities. 4. The mainstream media is controlled by the deep state and is spreading lies and propaganda. 5. There are coded messages from Q that reveal the truth about what is happening behind the scenes. 6. There is a storm coming soon that will sweep away the elites in power and restore the rightful leaders (from PRRI direct)
October 14, 2025 at 6:14 PM
byrdnick.com
As in humans, the plausibility of a conclusion can #bias #AI assessment of arguments (#beliefBias).

Chain of Thought prompting reduced (but didn't eliminate) the bias.

Plausibility influenced validity judgments ...and vice versa?!

doi.org/10.48550/arX...

#logic #cogSci #tech
Table 1's "llustrative examples of valid and invalid syllogisms with plausible and implausible conclusions. Here, plausible indicates that the conclusion is true in the real world, whereas implausible indicates that it is false." How the "overall content effect (CE)" is computed. "Table 1 reports the accuracies of Qwen2.5-32B-Instruct and Qwen3-14B on the test split of the logical validity classification task. Both models exhibit content effects in the zero-shot setting, while the bias is greatly reduced with CoT prompting. Looking at the CE metric, we observe that for Qwen2.5-32B-Instruct it is relatively high in zero-shot (0.348), showing strong conflation of plausibility with validity. When prompted with CoT, CE drops substantially to 0.096, indicating that explicit reasoning almost eliminates the bias. For Qwen3-14B, the zero-shot CE is lower (0.213), suggesting less susceptibility to content effects even without structured reasoning. When given CoT prompting, CE drops almost to zero (0.014), implying that validity judgments become nearly independent of plausibility. Across all models, CoT prompting achieves a significantly lower CE compared to zero-shot (Mann-Whitney U test, p < 0.01; see the complete behavioral results in Table 4 in the Appendix)." "Figure 3 shows the steering power of plausibility vectors when applied to hidden states during the validity classification task (“plausibility → validity”) and the steering power of validity vectors when applied during the plausibility classification task (“validity → plausibility”), for both Qwen2.5-32B-Instruct and Qwen3-14B, under zero-shot and CoT prompting."
October 13, 2025 at 7:14 PM
byrdnick.com
The #AI companies are onto me!

#OpenAI recently automated the identification of tasks worthy of "slow" #reasoning.

In "Strategic Reflectivism...", I showed why that's a key to #intelligence (in humans as well).

The #preprint (accepted in #LNCS) is now available as an audiopaper (a.k.a. #podcast)👇
Upon Reflection, Ep. 16: Strategic Reflectivism | Nick Byrd, Ph.D.
In late 2025, artificial intelligence companies like OpenAI popularized the idea of automating the process of selecting which model is best for a task. This allowed users to simply send their promp…
byrdnick.com
October 9, 2025 at 1:43 PM
byrdnick.com
Is solipsism measurable?

Seven studies developed a self-report scale of how much one doubts that others exist (beyond one's own mind).

It correlated with loneliness, social disconnection, aggression, and problematic gaming.

doi.org/10.1016/j.pa...

#xPhi #psychometrics.#psych
Is there anyone else out there? A measure of psychological solipsism
For centuries philosophers have debated solipsism, the idea that people cannot prove that anything exists outside of their own minds. However, there h…
doi.org
October 8, 2025 at 1:03 PM
byrdnick.com
🤬 “Among articles stating that data was available upon request, only 17% shared data upon request. … Results replicate those found elsewhere: data is generally not available upon request, and promissory Data Availability Statements are typically not adhered to.”

#openScience #philSci #ethics #psych
ianhussey.mmmdata.io Ian Hussey @ianhussey.mmmdata.io · Oct 4
My article "Data is not available upon request" was published in Meta-Psychology. Very happy to see this out!
open.lnu.se/index.php/me...
LnuOpen | Meta-Psychology
open.lnu.se
October 4, 2025 at 1:36 PM
byrdnick.com
Proud of Vahid's use of #computational #CogSci to identify and compare #reasoning errors in #Reddit users and communities.

He's presenting it at an #AI + #decisionSci workshop at #CMU : www.cmu.edu/ai-sdm/r...

Follow him for alerts about this and more: www.researchgate.net...
Beyond Fact-Checking: Empowering Flexible Human-AI Teams to Detect & Counter Online Misinformation Vahid Ashrafi, Nick Byrd, and Jordan W. Suchow Misinformation is persistent: From pandemic conspiracies to election denial, false claims spread faster and farther than facts, threatening health and democracy. Current defenses fall short: Fact-checking and moderation focus on sentiment but overlook the cognitive vulnerabilities that make cognitive errors drive spread. Biases like confirmation bias, emotional reasoning, and exaggerated thinking distort judgment and amplify viral narratives. We detect and quantify 419 cognitive errors in Reddit users' text, building "cognitive fingerprints" that reveal reasoning flaws behind misinformation. Impact: These insights enable flexible human-AI teams to anticipate, detect, and counteract misinformation. Beyond Fact-Checking: Empowering Flexible Human-AI Teams to Detect and Counter Online Misinformation Vahid Ashrafi, Nick Byrd, and Jordan Suchow Abstract. [Vahid construced] a comprehensive knowledge base of 419 cognitive errors—including cognitive biases and logical fallacies—grounded in empirical findings from psychology and cognitive science. We then trained GPT-4.1, a state-of-the-art language model, to identify potential examples of these cognitive errors in users' posts. ... The robustness of AI-generated insights was rigorously validated by human annotators, achieving high reliability (Spearman’s ρ = 0.86). ... we propose actionable solutions such as targeted cognitive interventions to educate users about common reasoning errors, AI-driven moderation systems that identify problematic reasoning patterns rather than merely flagging false information, and personalized cognitive profiling tools to proactively identify at-risk individuals and communities. The NSF AI Institute for Societal Decision Making (NSF AI-SDM) sponsors the participation of selected speakers and students in an annual workshop of Human-AI Complementarity for Decision Making. Human-AI Complementarity, defined as the condition in which Humans + AI working together results in better decisions than humans or AI working alone, is a broad goal pursued in several projects of the NSF AI-SDM. In 2025, we will focus on how to create flexible Human-AI teams to achieve complementarity. This theme refers to the interdisciplinary study of how to design and deploy AI systems in ways that are dynamically aligned with human values, robust to unexpected behavior, and safe even under failure modes. It encompasses short-term concerns about deployed systems (e.g., fairness, robustness, interpretability, and misuse) and long-term concerns about advanced general AI (AGI) that could have large-scale societal impacts if not aligned with human interests.
September 26, 2025 at 4:53 PM
byrdnick.com
How can #AI and #cognitiveScience improve #healthcare?

We got some answers from the #NudgesInHealthcare Symposium at #UPenn.

Check out this summary of some themes in the write-up below:

ldi.upenn.edu/our-wo...

#medicine #psych #econ #compSci #tech #LLM #edu #bioethics #xPhi
A poster presentation at UPenn with Nick Byrd and a poster about antibiotic resistance in the foreground. Ashley West, PhD, Director of Behavioral Design at Lirio (left), discusses her poster, “Designing an ‘At Home’ Digital Health Intervention for Supporting Chronic Condition Lifestyle Management,” with LDI Senior Fellow Renée Betancourt, MD, an Associate Professor of Clinical Family Medicine and Community Health at the Perelman School of Medicine. A panel discussion featuring Adam Rodman, Carissa Kathuria, Craig Joseph, and Kenrick Caito, moderated by Sri Adusimalli. A panel discussion featuring Kim Waddell, Meeta Kerlin, Zahera Farhan, Sunita Desai, and Hayley Belli.
September 25, 2025 at 1:32 PM
byrdnick.com
Thousands of people in a dozen countries thought reflective reasoning was usually the best way to make a decision in ordinary dilemmas.

Runners up were intuition, friends' advice, and the wisdom of a crowd (in that order).

doi.org/10.1098/rspb...

#cogSci #epistemology #xPhi
About the "everyday dilemmas". Descriptive statistics about the data. The descriptions of intuition, deliberation, friends' advice, and wisdom of a crowd. Results
September 24, 2025 at 12:05 PM
byrdnick.com
Do reflection test solutions actually involve reflection?

Our think-aloud studies found they usually do (doi.org/10.14264/0f1...), but Ryan Jesson found that solution-prompting insight is often unconscious or spontaneous.

doi.org/10.14264/0f1...

#ProcessTracing #psychometrics
Participants and data Procedure Results (1 of 2) Results (2 of 2)
September 23, 2025 at 10:27 AM
byrdnick.com
Might the concept of "good judgment" vary by framing or social roles?

Five studies of four nations (🇺🇸🇨🇦🇬🇧🇨🇳) found some words and roles were more associated with "rational" than "reasonable" (and vice versa).

doi.org/10.1162/opmi...

#CogSci #xPhi #linguistics #dataViz
Descriptive statistics about the samples "Figure 1. Top ten most frequent co-occurrence of Adjectives with ‘rational’ (blue) and ‘reasonable’ (orange) when asked to describe most important characteristics of a person showing sound judgment (Study 1a) / good judgment in a challenging situation (Study 1b). Adjectives are ordered from those most associated with ‘rational’ to those with ‘reasonable.’ Dumbbell nodes represent the percentage of each adjective’s co-occurrence relative to the sum of independent occurrences of each pair of terms." "Figure 2. Qualities attributed to rational and reasonable persons in Study 1. Color-coded adjectives reflect Analytical, Moral, and Inner Fortitude items (Study 1a)/Agency and Communion factors (Study 1b). Top panels: Estimates from linear mixed model with responses to all characteristics nested in participants, with target order (rational vs. reasonable) as a covariate and false discovery rate correction for multiple testing. Dashed vertical line delineates effects 1 unit above midpoint of the 1–7 scale in Study 1a / half a unit above the midpoint of the 1–5 scale in Study 1b. Bottom panels: Pearson’s correlations and 95% CIs of average scores across items making up each factor. ***p < .001, **p < .01, *p < .05." "Figure 3. Top Panel: Preference proportions for reasonable vs. rational agents in social roles. Displays proportions (with 95% CI) derived from logits in generalized mixed models. The dashed line at .50 indicates parity; above this, preference leans towards reasonable agents, and below, towards rational agents. Bottom Panel: Necessity ratings for rationality and reasonableness by role (1–5 rating scale). Shows estimated means and 95% CI. Ratings indicate moderate to high necessity (3–4) for both rationality and reasonableness across rule-based (on the right) and holistic roles (on the left)."
September 22, 2025 at 10:25 AM
byrdnick.com
Can #AI chatbots handle student discussion better than teachers?

In a controlled study of 83 Chinese adolescents, dialogue with a #generativeAI resulted in more learning than dialogue with a human teacher.

doi.org/10.1080/1049...

#edu #LLM #tech #teaching #policy #CogSci #xPhi
The experimental design with images of the students using generative AI chatbots in each phase of learning. Examples of the chatbot prompts and responses. The test of learning and self-reported reflective thinking scale. Effects of chatbot (vs. teacher) facilitated dialogue on learning — the chatbot dialogue usually produced more learning.
September 17, 2025 at 10:45 AM
byrdnick.com
This week I'm Zooming into the Human & Artificial #Rationality conference: https://har-conf.eu

My paper argues that a key to #intelligence is pragmatic switching between intuitive and reflective inference — the paper forthcoming in #LNCS is on #ArXiv: doi.org/10.48550/arX...
"The Human and Artificial Rationalities (HAR) conference series is focused on comparing human and artificial rationalities, investigating how they interact together in a practical sense, but also on the theoretical and ethical aspects behind rationality from three main perspectives: Philosophy, Psychology, and Computer Sciences. HAR aims at building bridges between these three fields of research and invites experts from all around the world to discuss the hottest topics on the interaction of human or artificial cognition and of human-machine interaction." "This paper will argue that — among other things — we can become too obsessed with factors such as size, chains of thought, and other features of reasoning models. .... as models increased in size, they were more likely to “behave” like a system called mReasoner [43, 44], especially its propensity for System 2 or reflective thinking [29] (Fig. 2). Such results suggest that this capacity to reason reflectively probably matters at least as much as model size [11]." "So-called “reasoning” models [4] appear to engage in multi-step reflection and often outperform humans on reflection tests [37]. These two facts may lead you to conclude that what distinguishes reasoning models from language models is a capacity for reflective thinking. However, much like humans often pass reflection tests without exhibiting any signs of reflective thinking [17, 74] (see “correct-but-unreflective” in Table 1), Hagendorf and colleagues found that language models continued to outperform humans on reflection tests even when they could not exhibit signs of reflective thinking, such as chain-of-thought reasoning [37] (Fig. 3)." "There is growing evidence in favor of a dual model approach, a la dual process or dual systems theory [31]. For example, Yan and colleagues found that pairing one small language model with another small model (that serves as a “reflective system”) allowed small hybrid systems to compete with larger models that had more than ten times as many parameters [85] (Table 2). ... These results suggest that dual- or multi- model architectures may be key to yielding the reflective reasoning and rationality that we expect from intelligence systems, but without overlooking other goals or constraints [64]. ... The core idea of this paper is that one key to intelligence is pragmatic (rather than perpetual) deployment of reflective reasoning— a view I have been calling Strategic Reflectivism [Section 4.2.3 in 9]."
September 16, 2025 at 9:11 AM
byrdnick.com
"We do not assume [system 2 is] always better ...than [system 1], or vice versa, .... Some tasks might be better handled by S1..., especially once the system has acquired enough experience": doi.org/10.1145/3715...

Amen! Long live pragmatism: doi.org/10.48550/arX...

#cogSci #AI #xPhi #epistemology
Fabiano, F., Ganapini, M. B., Loreggia, A., Mattei, N., Murugesan, K., Pallagani, V., Rossi, F., Srivastava, B., & Venable, K. B. (2025). Thinking Fast and Slow in Human and Machine Intelligence. Commun. ACM, 68(8), 72–79. https://doi.org/10.1145/3715709 Byrd, N. (2025). Strategic Reflectivism In Intelligent Systems (No. arXiv:2505.22987). https://doi.org/10.48550/arXiv.2505.22987
September 15, 2025 at 1:28 PM
byrdnick.com
Some #medical #nudges that didn't work:

Most (of a dozen) clinician-facing #ElectronicHealthRecord deprescribing tools didn't work? doi.org/10.1111/jgs....

Showing clinicians cheaper versions of a drug didn't reduce prescription costs? doi.org/10.1001/jama...

#reproducibility
Comparing Fourteen Behavioral Science Electronic Health Record Deprescribing Tools in Older Adults: NUDGE‐EHR Adaptive Trial
Background Interventions to reduce prescribing of high-risk medications like benzodiazepines and sedative hypnotics to older adults have had modest success. Electronic health record (EHR)-based aler...
doi.org
September 12, 2025 at 11:12 AM