Nick Byrd

Upon Reflection, Ep. 16: Strategic Reflectivism

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 prompt or request and let the system determine whether to respond using a “fast” or a “slow” reasoning model.

In a 2022 paper, I suggested that this kind of strategic deployment of slow, more reflective reasoning could be crucial to good judgment and decision-making. Since then, I have been synthesizing the arguments and evidence for this view in a paper titled, “Strategic Reflectivism in Intelligent Systems”.

After the new paper was accepted in Lecture Notes In Computer Science, I recorded a podcast of me reading the final proofs of the paper for the podcast. I review evidence suggesting that one key to intelligence in humans and machines is pragmatic switching between intuitive and reflective thinking based on the goals of the system. The paper has a wide range of implications for applied science, computer science, decision science, and epistemology.

Podcast: Play in new window | Download (Duration: 29:00 — 39.8MB)

A 3-Paragraph Writing Assignment

Since 2018, I’ve been adapting a 3-paragraph writing assignment that I stole from John Roberts at Florida State University. In this post, I’ll explain the assignment, its requirements, my advice to students, and what I like about this assignment. You’re welcome to adapt this assignment for your own purposes, of course.

Upon Reflection, Ep. 15: A Two-Factor Explication Of ‘Reflection’

You may have heard me drone on and on about this thing called “reflective thinking”. We philosophers and cognitive scientists are preoccupied with it. However, the term ‘reflection’ is sometimes used in different ways by scholars. To unify, make sense of, and guide our research, I synthesized a unified account from hundreds of years of English language, from philosophers, and from cognitive science. The result is this paper.

In this episode, I’ll read the paper, which explains the two key features of ‘reflection’ and how we measure them. This two-factor account of reflective thinking has implications for theories of rationality, self-knowledge, and dual-process theories.

Podcast: Play in new window | Download (Duration: 43:28 — 59.7MB)

Upon Reflection, Ep. 14: Analytic Atheism & Analytic Apostasy Across Cultures

You may have heard that atheists tend to score better on reflection tests than theists? But why do scientists find this “analytic atheism” correlation?

Many studies have attempted to answer this question. Of course, even the best studies had limitations. So Steve Stich, Justin Sytsma, and I developed better methods and studied over 70,000 people on 6 continents. What did we find?

Apostasy was key. Those who shed their religion since childhood were the most reflective. Lifelong atheists were not necessarily more reflective than theists. In other words, the analytic atheism correlation seems to be explained by analytic apostasy.

In this episode, I’ll explain the methods, results, and implications in our paper “Analytic Atheism & Analytic Apostasy Across Cultures” which will be published in Religious Studies.

Podcast: Play in new window | Download (Duration: 49:02 — 67.4MB)

Upon Reflection, Ep. 13: Reflection-Philosophy Order Effects and Correlations Across Samples

Suppose you glance at a clock that, unbeknownst to you, is broken, showing the same time all day. Nonetheless, you happened to look at the clock precisely when it showed the correct time. So your belief about the time is correct. My question is this: did you know what time it is?

Perhaps you think that you did. After all, you formed a belief on the basis of a device that most people trust and the belief was true! What else would it mean to know something? Well, in academic philosophy, the orthodox answer to this kind of thought experiment is “no”.

People who perform better on tests of reflective thinking tend to report philosophers’ orthodox answer to this kind of thought experiment. And, if you’ve been following my research, you know that philosophers are particularly reflective thinkers. These correlations may make you wonder about causation. Does thinking reflectively cause people to accept philosophers’ orthodoxy? Or is it the other way around: does studying thought experiments like the broken clock case somehow result in people performing better on reflection tests?

In this episode, I’ll tell you about the experiment I ran to find out. The paper is titled “Reflection-Philosophy Order Effects and Correlations Across Samples” and has been accepted for publication in Analysis. The paper will also mention a bunch of other thought experiments, tests of reflective thinking, and measures of research participants’ data quality.

Podcast: Play in new window | Download (Duration: 30:51 — 42.4MB)

The Bat And Ball Problem 20 Years Later

In 2002, a chapter from Kahneman and Frederick mentioned “the bat and ball problem”.

A bat and a ball cost $1.10 in total.
The bat costs $1.00 more than the ball.
How much does the ball cost?

By 2005, Frederick’s Cognitive Reflection Test paper added the lesser known Widgets and Lily Pad problems. In the intervening 20-ish years, each paper seems to have accrued over 5000 citations.

In 2023, Meyer and Frederick published a massive follow-up paper about the first problem: 59 studies, over 73,000 participants, and more pages of Appendixes than pages in the main article. As someone studying various reflection tests and interventions, I had to take a look right away. In this post, I list five initial takeaways and two things to like about the paper.

Oppenheimer: ‘Philosopher-Scientist-Statesman’

J. Robert Oppenheimer “was widely known not just for his scientific success but for his remarkably wide-ranging knowledge of the humanities [,…] an extraordinary combination…”, says Ashutosh Jogalekar in his 8th and final post about Oppie over at 3 Quarks Daily (2023). Why do I start with this quote? I think it alludes to an important lesson for our time. I explain in less than 700 words below.

Upon Reflection, Ep. 12: Tell Us What You Really Think

I have a question for you: “If a bat and a ball cost $1.10 in total and the bat costs $1.00 more than the ball, how much does the ball cost?“. Did 10 cents seem right? The authors of questions like this are attempting to lure you to accept this incorrect answer in order to test whether you thought reflectively when you solved the problem. However, there may be problems with this method of testing reflective thinking. So my colleagues used some underrated methods to determine the degree to which tests like this misclassify correct responses as reflective or lured responses as unreflective. I’ll read the paper in this episode.

Podcast: Play in new window | Download (Duration: 32:57 — 60.3MB)

Here’s to the Philosopher-Scientists!

Sometimes philosophers complain that scientists do philosophy badly and that philosophers may thereby be underrated. The idea is that people could have better philosophy if they just turned to academic philosophers rather than the popular scientists that have done philosophy badly. (Perhaps analogous complaints about philosophers circulate among scientists). In this post, I want to turn our attention to scientists that do philosophy well and philosophers that do science well.

Upon Reflection, Ep. 11: Testing for Implicit Bias

In this episode, I read my short paper with Morgan Thompson in WIRES Cognitive Science titled, “Testing for Implicit Bias: Values, Psychometrics, and Science Communication“. You may have heard about implicit bias. It is measured by indirect rather than direct measures of bias. We reconstruct arguments from debates about these measures, reveal some instances of talking past one another, highlight how debate has changed, and highlight how the debate is laden with value judgments about psychometrics and science communication. As always, free preprints of my papers are available on my CV under “Publications”.

Podcast: Play in new window | Download (Duration: 23:05 — 42.3MB)