Religion on the Mind, Episode #368 — “Your Brain is a Large Language Model”
Host: Dr. Dan Koch
Guest: Dr. James W. Pennebaker, Professor Emeritus of Psychology, UT Austin
Release Date: December 15, 2025
Episode Overview
In this episode, Dr. Dan Koch interviews Dr. James W. Pennebaker, a renowned psychologist and expert in language, psychology, and the connections between language use, culture, and identity. They dive into the concept of human brains as “large language models,” exploring the parallels between our natural use of language and modern artificial intelligence (AI) systems like ChatGPT. The conversation weaves together themes from psychology, religion, wisdom traditions, politics, AI hallucinations, spiritual abuse, polarization, and how individuals and societies pursue (and evade) truth.
Key Discussion Points & Insights
1. Human Rationalization and Language Use
- Humans are Not Rational, but Rationalizing
- Dr. Koch recounts Jonathan Haidt’s idea that people use language as post-hoc justification for their actions (04:00–05:30).
- Pennebaker: “Humans are not rational. We're rationalizing... Our brains evolved to try to figure out the world, to be able to predict it as well as we possibly can. And that's what the brain is for.” (05:32)
- Language as Social Glue and Foundation of Culture
- Language enables us to communicate, share stories, and build complex cultures, not just survive as individuals.
- Pennebaker: “This is really kind of the beginning of what culture is, the ability to... identify who I am, who we are, who this group is, and that group that's over in the other field, five miles away.” (07:34)
- Language as Post-Hoc Rationalization
- People use language to protect their self-image and social status, sometimes at the expense of truth.
- Koch: “I am trying to find out truth, but I'm also trying to save my own ass all the time. And everyone's doing that all the time. More mature people maybe do it slightly less, but everybody does it.” (09:51)
2. Wisdom Traditions, Truth, and Cultural Evolution
- Wisdom Aggregation Across Generations
- Koch suggests wisdom traditions aggregate useful ‘claims’ over time, acting as containers for distilled, tested knowledge—not absolute truth.
- Koch: “Wisdom tradition is interesting because... in the aggregate [it figures out] which of those formulations... found the most truth, which we would find out through usefulness and help.” (10:34)
- Truth is Provisional and Contextual
- Pennebaker: “There is the thing that I find so intriguing. Humans are great at coming up with hypotheses and beliefs and then they start to believe them that they are true. My personal view is, I'm happy to say something's true for the time being until there's persuasive evidence that maybe the world does not work that way.” (12:31)
- Memorable Quote: “Gravitate to people who are seeking truth. Run away from those people who have found it.” (14:47, Pennebaker quoting a bumper sticker)
- Sorting Wisdom by Longevity and Usefulness
- Some teachings (“the honeymoon phase” in relationships, “fear/awe is the beginning of wisdom”) persist because they are repeatable or universally applicable; others (like sacrificial atonement) fall away over time (15:07–19:00).
3. Parallels Between Human Brains and AI Language Models
- AI “Hallucinations” and Human Bullshitting
- Both large language models and humans feel compelled to generate statements even when uncertain, often producing confident-yet-unverifiable explanations.
- Koch: “The large language model... wants to have something to say whether or not the thing that they say is true... That's kind of how we use language.” (22:00–23:52)
- Pennebaker: “That sounds just like me. It sounds like everybody. It sounds like how you talk to your wife about why you didn't do your chores.” (23:52)
- Example: Responding confidently with made-up statistics or references (24:51–26:08)
- Confidence Over Accuracy in Public Discourse
- Textual analysis shows a linear rise in confidence and simplicity in presidential speeches/public leaders' language since the early 20th century, with a decline in factual/logical rigor (26:08–27:25).
- Pennebaker: “What we found is since 1900, there's been a linear increase in terms of the degree to which presidents speak with confidence... and at the same time... we are being less internally logical over time.” (27:10)
- Memorable: “These are the Cheetos of leaders. So these are, you know, they are giving us what we want.” (27:25)
4. Media, Technology, and the Shaping of Language
- Mass Media and the Rewarding of Simplicity & Confidence
- Rise of radio and later media drives leaders and religious figures toward simplicity and confident delivery to attract followers (30:42–33:26).
- Junk food metaphor: “People buy a lot more garbage food if it tastes good, even though it's not nutritious, because that's what humans like.” (32:27)
- Churches and religions increasingly tailor their “product” to what gets engagement/clicks (32:41–33:26).
5. Spiritual Abuse and Dangers of Confident Simplicity
- Charisma and Authority as Vectors for Influence (Both Political and Spiritual)
- The psychological appeal of confident, simple language becomes more dangerous in religious contexts where leaders are perceived as speaking for God.
- Koch: “There’s a meaningful difference, I think, between a presidential candidate saying, ‘I’m your best bet for solving immigration, economy issues...’ and a pastor saying, ‘the Lord is saying to you today...’”
- Pennebaker: “A powerful charismatic political leader is able to say all... Everything we know shows that immigration leads to the breakdown of society... total, total deceptive, false information. That person is going to have a huge impact.” (36:35)
- On new churches/leaders: “You create your own, your own church... Is it actually the religious fundamentalism and simplicity that is drawing them...? Religions are just like toothpaste.” (39:18)
6. Religion, Polarization, and Tribal Division
- Tribalism and Information Silos
- Isolation driven by both technological changes (the “Christian bubble,” social media) and psychological preferences, resulting in walling off alternative views (57:39–59:21).
- Koch: “Evangelical American Christianity, for the most part... created their own parallel institutions... that trained evangelicals over a long period to sort of just trust their own stuff.” (61:47)
- Social Nature vs. Siege Mentality
- Humans are social by nature but certain environments (isolated, less diverse communities) reward information “walling off.” (59:28–61:47)
- Two Strains within Evangelicalism
- “Crossroads” (engaging, open) vs. “Castle/Garters” (siege, fundamentalist) types—and how the latter more easily fuel polarization and Trump-style politics (61:47–65:55).
7. Engagement Across Differences and Political Psychology
- Shared Humanity and Common Values
- Both speakers wrestle with the tension between dialogue vs. disgust/exhaustion in the context of polarized communities.
- Pennebaker: “The reality is you share, probably 90% of your beliefs are the same. That's what we forget as humans.” (67:56)
- The Exhaustion of Engaging Fundamentalism
- Koch: “I do have a discomfort with it. I have a real lack of interest... a real boredom with certain topics that I've been over a thousand times, and there's a certain exhaustion with feeling judged that I recognize as well.” (68:33)
8. AI, Large Language Models, and the Search for Truth
- How Well Can LLMs Reduce Polarization?
- LLMs (like GPT) can present both sides and play a useful role—but outcomes depend on how people use them.
- Pennebaker: “The art form in using GPT is it's a great place to get ideas and to look at things from a different perspective... approach it with playfulness.” (72:32–74:49)
- Limits, Biases, and Responsibilities
- LLMs are not a source of singular truth; useful prompts ask for counterarguments and alternative perspectives.
- Pennebaker: “There's not necessarily one truth. In fact, there's probably never one truth.” (72:32)
- Societal effect? Probably as good or better than echo chamber systems, as LLMs often suggest follow-up questions and opposing views (75:31–77:21).
- Critical Use and Personal Agency
- LLMs should be treated like “a smart friend who is not judging you, who can give you... honest answers. And by honest, it thinks it's honest. Keep in mind there is no truth, but there is an approximation of truth.” (77:21)
- Maturity and Digital Literacy
- Using LLMs wisely is akin to CBT: “We're training our clients to become their own CBT therapists... That's what I'm hoping for with my clients. So that I guess that is hopeful that it... helps [people] learn to think and adjudicate and all of that.” (78:01–79:53)
Notable Quotes & Memorable Moments
-
On confidence vs. truth in public figures:
“[Presidents] are giving us what we want. We want somebody who speaks simply and with confidence. Well, that's a...horrible combination.”
(27:25, Pennebaker) -
On shifting wisdom traditions:
“Gravitate to people who are seeking truth. Run away from those people who have found it.”
(14:47, Pennebaker quoting a bumper sticker) -
On “hallucination” and language:
“The large language model...wants to have something to say whether or not the thing that they say is true...That's kind of how we use language.”
(22:00, Koch) -
On AI’s value and limitations:
“Assume [GPT is] probably better than your average friend, but it's not 100% correct...Approach it with playfulness.”
(74:49, Pennebaker) -
On future prospects:
“I find [the advance of AI] both absolutely thrilling and terrifying. So, you know, it's going to be a fascinating time ahead.”
(80:19, Pennebaker)
Suggested Timestamps for Key Segments
- 05:30 — Are humans rational? The evolution of language and culture
- 10:30 — Wisdom traditions as evolving containers of knowledge
- 14:47 — “Gravitate to people who are seeking truth...”
- 23:52 — Human “hallucinations” and BS as language models
- 26:08 — Confidence vs. logic in political rhetoric
- 34:45 — Spiritual abuse and dangerous charisma
- 57:39 — Technology, media, and social fragmentation
- 72:32 — LLMs, truth, and digital literacy
- 80:19 — The thrilling and terrifying future of AI
Episode Tone
Candid, humorous, reflective, and occasionally irreverent; the conversation is rich with empirical insights, playful analogies, and an underlying call for humility and curiosity in the search for truth—both human and digital.