363 | Chandra Sripada on How LLMs and Humans are Cognitive Cousins
August 10, 2026
AI Summary
5 min readLLMs and Humans as Cognitive Cousins
In a 2024 paper, cognitive scientist Chandra Sripada and his colleague Rick Lewis discovered something striking: a small, untuned language model with just 2 billion parameters spontaneously reproduced the same dual-process dynamics that cognitive scientists have studied for decades. When given a prompt like "the crayon is red, so the crayon is," the model automatically wanted to say "red." But when a rule-based prefix reversed the color mapping, the model's in-context processing overrode that automatic response—and the resulting conflict patterns, including congruency facilitation and congruency sequence effects, matched human performance almost exactly. This wasn't engineered. It emerged from next-word prediction alone.
The Core Question and the Evidence
The central question Sripada addresses is whether large language models think like humans or merely simulate human-like outputs through alien mechanisms. He argues for what he calls the "cognitive cousin" hypothesis: LLMs and humans share deep similarities at the level of core cognitive principles, not just surface behavior.
Continue reading the full summary in the app — free to try.
Read Full Summary →Free • No credit card required
Never miss an episode of Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas
Get every new episode summarized in your inbox — free, ~5 minutes to read.
No spam. Unsubscribe anytime.
What you'll learn
- 1 (07:05) **Guest Introduction: Chandra Sripada** - Sean introduces philosopher and cognitive scientist Chandra Sripada, who studies both human cognition and LLMs, setting up the episode's central question.
- 2 (07:55) **How LLMs Actually Work** - Sripada provides a technical overview of LLM architecture, emphasizing prediction, transformers, and the residual stream.
- 3 (14:44) **Scratch Pads and Self-Discovered Strategies** - Sripada explains how LLMs learn to use their context window as a working memory for complex reasoning.
- 4 (19:34) **The Core Question: Cognitive Cousin or Alien Intelligence?** - Sean and Sripada frame the two competing hypotheses for how LLMs produce human-like outputs.
- 5 (23:46) **The Strawberry Problem and What It Really Means** - Sripada addresses the famous LLM failure to count letters, arguing it is not evidence of alien cognition.
- 6 (28:09) **Cognitive Science Phenomena Reproduced by LLMs** - Sripada catalogs multiple classic human cognitive effects that LLMs also exhibit.
- 7 (34:55) **The Car Wash Problem** - Sripada offers another behavioral example where LLMs make a mistake that is surprisingly human-like.
+ Full timestamped outline available in the app
Show Notes
Large Language Models display an uncanny ability to construct human-sounding speech, and can synthesize concepts in novel ways. Is this because they are truly thinking like human beings in some way, or have they found a way to be human-like without reproducing the internal mechanisms of human thought? Chandra Sripada argues that LLM cognition is more human-like than we suppose, and offers evidence from the ways that cognitive scientists study actual humans.
Blog post with transcript: https://preposterousuniverse.com/podcast/2026/08/10/363-chandra-sripada-on-how-llms-and-humans-are-cognitive-cousins/
Support Mindscape on Patreon.
Chandra Sripada received an M.D. from the University of Texas and a Ph.D. in philosophy from Rutgers University. He is currently a professor of philosophy and psychiatry at the University of Michigan, where he holds the Theophile Raphael Research Professorship and directs the Weinberg Institute for Cognitive Science. He writes Cognition, Decoded, a Substack newsletter about AI, cognitive science, and philosophy.
Babbel is offering listeners up to 60%
More from this podcast
Sean Carroll's Mindscape: Science, Society, Philosophy, Culture, Arts, and Ideas →