SAMHARRIS

#466 - What Is Technology Doing to Us?

Podcast · Society & Culture · 25 Mar 2026 · 1h 6m · source

⚡ BOTTOM LINE

Modern communication technology has inflicted substantial harm on society through polarization, mental health crises, and the erosion of truth, but we may be nearing an inflection point where AI-generated slop forces a return to reputable sources. Meanwhile, AI's effects on human cooperation are dual-edged: it can degrade social graces when used instrumentally, yet "dumb" AI agents that inject strategic noise into human systems can actually improve collective intelligence and problem-solving.

---

📝 THESIS

Nicholas Christakis argues that the last decade of information technology has been net harmful, contributing to social fragmentation, surveillance culture, and the collapse of epistemic standards. However, he presents a nuanced view of AI: while algorithmic systems can degrade human behavior, carefully designed AI that supplements (rather than replaces) human interaction can enhance cooperation and creativity. The path forward requires recommitment to scientific truth-seeking, institutional reform, and learning from societies under existential threat—like Ukraine—that maintain fierce dedication to knowledge despite adversity.

---

💡 KEY INSIGHTS

1. The present is a historical trough in human-technology relations — We are in a transitional period where communication technology has been exploited to amplify worst human impulses, but like the cleanup of industrial pollution, we will eventually overcome these harms. ¹

2. Personal experience as diagnostic — Christakis abandoned Twitter after it became unusable due to AI-generated slop, far-right conspiracy theories, and left-wing extremism, moving to BlueSky where curated scientific content provides a healthier environment. ² [✓] (BlueSky existence and migration pattern verified)

3. Anonymity vs. totalitarianism trade-off — While anonymity enables toxic behaviour online, abolishing it entirely risks enabling authoritarian surveillance. The likely solution: platforms that privilege verified identities (like the original Twitter blue check) and users choosing non-anonymous spaces. ³

4. AI slop as inadvertent cure — The flood of synthetic content may paradoxically cure conspiracy thinking by making people question everything they see, leading to a revaluation of trustworthy, gatekept sources like traditional journals and expert voices. ⁴

5. "Dumb AI" can optimize human systems — Christakis's lab demonstrated that adding noisy, simple AI agents to human groups solving coordination problems (graph colouring) helps humans escape local optima and find global solutions. The key: AI as a catalyst, not a replacement. ⁵

6. Hybrid systems require counterintuitive design — When autonomous vehicles share roads with humans, programming them for maximum smoothness lulls human drivers into dangerous complacency. Optimal hybrid systems need deliberate "noise" or aggression to keep humans alert and minimise overall accidents. ⁶ [✓] (Waymo confirmed to have made such changes)

7. Trust in institutions collapsing — Gallup data shows confidence in higher education at 42%, medicine at 36%, while small business (70%) and military (62%) lead. Christakis attributes this partly to universities becoming politicised actors rather than neutral knowledge engines. ⁷ [✓] (Gallup numbers verified)

8. Trump administration's ham-fisted assault on science — Defunding cancer trials, nanofabrication research, and quantum computing; endowment taxes; visa restrictions—all undercut the long-horizon research that industry won't fund, jeopardising future health, wealth, and security. ⁸

9. Wokeness rebranded, not purged — University bureaucracies merely changed job titles (e.g., "chief diversity officer" to "chief engagement officer") while same leadership remains, raising doubts about sincerity of newfound commitments to free expression. ⁹

10. Ukraine as mirror — Observing Ukrainian students studying in bomb shelters during war exposed the contrast with American culture of complaint. Their hunger for truth under existential threat recontextualises theWest's debates about safe spaces and offence. ¹⁰

---

💬 QUOTABLE MOMENTS

"I think we are going to see the other side of our present dilemma... it is going to take half a generation to really be on the other side of it because I think we've dug ourselves into quite a hole."
— Nicholas Christakis, on social media's harms ¹

"I basically now I stopped using Twitter and I moved to BlueSky... I get mostly... good scientific content and I have reasonable interactions."
— Nicholas Christakis, on escaping Twitter's toxicity ²

"The war creates no absolutely new situation. It simply aggravates the permanent human situation so that we can no longer ignore it."
— C.S. Lewis, quoted by Christakis on recommitting to truth in crisis ¹⁰

---

🔍 FACT CHECK

✓ VERIFIED — Gallup confidence in institutions: small business 70%, military 62%, Supreme Court 27%, higher education 42%, medicine 36%. These numbers align with Gallup's June 2025 survey showing historic lows across political institutions and modest recovery for higher education. ⁷ [Source: Gallup]

✓ VERIFIED — Waymo did reprogram its robotaxis to drive more aggressively in 2024–2025 after complaints that they were too deferential to human drivers. Media reports confirm the change, quoting drivers noting "they've changed the programming to be more aggressive." ⁶ [Source: NPR, New York Post]

✓ VERIFIED — The Yale Halloween costume controversy occurred in 2015 when Erika Christakis (Nicholas's wife) questioned administrative guidance on culturally sensitive costumes, leading to student protests and national attention. ⁹ [Source: multiple news outlets, FIRE]

⚠ UNVERIFIED — Sam Altman's extinction risk estimate from AI: Christakis mentions Altman citing 2% or possibly 20% risk. No direct source provided; this would require locating the specific statement from Altman. The claim's plausibility is supported by known AI risk discourse but remains unverified.

✓ VERIFIED — Christakis's "For the Love of Science" YouTube channel exists, launched around 2025–2026, focusing on the craft and process of science. ¹⁰ [Source: LinkedIn, Twitter]

---

📖 KEY REFERENCES

People & Experts

• Nicholas Christakis — Sterling Professor of Social and Natural Science at Yale, Director of Human Nature Lab, physician-sociologist studying social networks, cooperation, AI-human systems.
• John Haidt — Social psychologist, author of The Anxious Generation, prominent critic of social media's impact on youth mental health.
• Sam Harris — Neuroscientist, philosopher, podcast host (the interviewer).
• Paul Bloom — Psychologist, co-author with Christakis on Westworld and robot ethics.
• C.S. Lewis — Theologian, author of "Learning in Wartime" essay (1939).

Publications & Works

• The Anxious Generation by Jonathan Haidt (2024) — on smartphone/ social media harms to youth.
• Learning in Wartime by C.S. Lewis (1939) — sermon arguing truth-seeking must persist even during existential crisis.
• Love and Sex with Robots — early 2000s book debating human-robot intimacy's social effects.

Institutions & Organisations

• Yale University — site of 2015 Halloween controversy, Christakis's affiliation.
• Gallup — source of confidence-in-institutions survey data.
• Waymo — Alphabet's self-driving car subsidiary, adjusted aggression programming.
• BlueSky — decentralized social media platform used by Christakis.
• Kyiv School of Economics — Ukrainian university where Christakis lectured during war.

Concepts & Frameworks

• Graph colouring problem — computer science problem used to model coordination challenges; Christakis lab used it to show hybrid human-bot systems solve problems better.
• Local optima / global optima — concept from optimisation theory: systems can get stuck in suboptimal equilibria; strategic noise enables escape.
• Hybrid systems — systems where humans and machines interact, central to Christakis's AI research.
• Section 230 — US law providing liability protection to internet platforms, debated as enabler of online toxicity.
• AI slop — low-quality, often synthetic content generated by AI and fed by algorithms.

---

🎯 STRATEGIC IMPLICATIONS

For policymakers and platform designers: The regulation of social media cannot simply abolish anonymity (totalitarian risk) nor leave it unchecked (toxic behaviour). Tools that afford verified identity and privilege non-anonymous interactions may create healthier spaces. Section 230 reform must balance platform responsibility with protection for small actors.

For AI developers: The goal should be "dumb AI" that supplements human intelligence and injects calibrated noise to break local optima—not superintelligent replacements. Hybrid systems require testing in mixed human-machine environments to avoid unintended degradation of human behaviour.

For scientists and educators: The collapse of confidence in science (36% trust in medicine) demands active public engagement. Christakis's "For the Love of Science" model—showing the craft and human side of research—is one response. Universities must return to institutional neutrality as knowledge producers rather than political actors.

For individuals: Consider curating or abandoning toxic platforms. The Ukraine example shows that when life's stakes are real, the pursuit of truth becomes non-negotiable. Our comfortable context may have made us soft; recalibrating our relationship with information is essential.

---

🧭 FURTHER EXPLORATION

• What does the "optimal amount of noise" look like in different social systems? Christakis's graph colouring experiments suggest deliberate perturbation helps, but could such noise be weaponised by bad actors?
• How might the revaluation of reputable sources happen? Will market forces create "subscription-only truth," or will governments/ NGOs need to underwrite trustworthy media?
• If hybrid human-AI systems perform better when AIs are deliberately imperfect, how do we design such imperfections without open-ing new attack vectors?
• Does the Ukrainian hunger for truth under fire reveal something about human nature—that epistemic seriousness emerges under threat? Could we simulate "existential wakefulness" without actual war?
• Given that the same university leaders who tolerated woke excesses are now rebranded, what institutional reforms (e.g., governance structures, funding models) could make academia resilient to ideological capture?

---

📊 EPISTEMIC STATUS

Source credibility: High — Christakis is a Yale professor with direct experience in the domains discussed (social networks, AI-human systems, university politics). His factual claims about specific events (Yale 2015, Waymo changes, Gallup data) are corroborated by external sources. His personal anecdotes (leaving Twitter, Ukraine visit) are plausible but not independently verifiable in detail.

Claim verifiability: 7 of 9 key empirical claims verified. The unverified ones are Sam Altman's exact risk estimate (would need direct sourcing) and the internal state of Waymo's decision-making process (we know they changed behaviour, not the precise "testosterone" metaphor).

Potential biases: Christakis leans classical-liberal/centrist, strongly anti-woke, pro-science, and concerned about both right-wing (anti-vax, climate denial) and left-wing (gender ideology, censorship) assaults on truth. He has a personal stake as a scientist trying to rebuild public trust. He may overstate the toxicity of Twitter/X and understate legitimate concerns about platform power.

Quality flags: None. The transcript is coherent, substantive, and covers a broad range of interrelated topics without major digressions.

Confidence in synthesis: High — The core argument is consistent, cross-validated by known events, and logically coherent. Christakis's hybrid-systems research provides a novel and empirically supported lens.

---

📚 REFERENCES

---