AI summary by TwinMind
Steven Pinker | Rationality: What It Is, Why It Seems Scarce, Why It Matters | Talks at Google
Talks at Google
57m 6s
Introduction and Book Background
- Brian Welley (Google social scientist) interviews Dr. Steven Pinker about his book "Rationality"
- Pinker: Johnstone Professor of Psychology at Harvard, experimental psychologist, author of "The Language Instinct"
- Accolades: National Academy of Sciences member, 2x Pulitzer Prize finalist, 9 honorary doctorates, Foreign Policy World's Top 100 public intellectuals
Motivation for Writing "Rationality"
- Book originated from Harvard course on rationality during pandemic
- Students requested course on effective altruism, Pinker broadened to applied rationality
- Core tools: probability, logic, Bayesian reasoning, distinguishing correlation from causation
- Central paradox: humans simultaneously rational (evidence-based medicine, moneyball, evidence-based policing) and irrational (fake news, conspiracy theories, medical quackery)
- Challenge: understanding human nature that can be both rational and irrational
Evolutionary Psychology and Human Rationality
- Critique of simplistic evolutionary explanations ("cavemen out of time")
- Humans evolved in "cognitive niche" - ability to develop mental models, intuitive theories
- Capabilities: scenario planning, tool/trap fashioning, extracting medicines/poisons, cooperation through language
- Example: San people of Kalahari Desert
- Track animals through hypothesis testing from fragmentary evidence
- Challenge authority, don't accept arguments from authority
- Engage in Bayesian reasoning (consider base rates)
- Distinguish correlation from causation
- Depend on rationality for survival
Modern Irrationality vs. Ancient Rationality
- Mismatch between evolved environment and current environment
- Example: Pinker's refrigerator water dispenser "sacrifice" thinking
- Primitive intuitions persist despite scientific knowledge
- Trust in scientific authorities varies by political beliefs
- Climate change acceptance predicted by political position, not scientific literacy
- People accepting scientific consensus often have wrong understanding of underlying science
Replication Crisis in Science
- Started ~15 years ago with two key events:
- John Ioannidis article: "Why Most Published Research Findings Are False"
- Daryl Bem's ESP study in Journal of Personality and Social Psychology
- Students predicted computer-generated erotic image placement
- Used conventional statistical methods, passed peer review
- Obviously bogus result questioning research standards
Causes of Replication Crisis
- Publication bias: Journals favor new, counterintuitive, exciting findings over negative results/replications
- Bayesian reasoning failure:
- Scientists neglect prior probability (base rate neglect)
- Focus on likelihood of data given hypothesis
- Confuse p-values with posterior probabilities
- Majority of social scientists misunderstand statistical concepts
- File drawer problem: Negative results unpublished, positive results get media attention
- Texas sharpshooter fallacy: Post-hoc highlighting of positive findings while ignoring null effects
- Questionable research practices: Data analysis flexibility after seeing results (gender subgroups, excluding years, different statistical tests)
Solutions to Replication Crisis
- Pre-registration of studies and analysis plans
- Publishing failures to replicate alongside successful replications
- Recursive power of rationality: science can examine and improve itself
Algorithms vs. Human Judgment
- Consistent finding since 1950s: Simple regression equations outperform human experts
- Applies to medical diagnosis, predictions based on multiple variables
- Even "aided human mind" (expert + algorithm results) often worse than algorithm alone
- Humans better at vision, language, common sense - provide data for algorithms
- Medical applications: X-ray reading, genomic analysis, drug interactions
- Physicians over-interpret diagnostic tests, neglect base rates
- Society needs to accept algorithmic superiority in prediction tasks while keeping humans in the loop for values, bias detection, variable generation
Technology Dependence Concerns
- Personal examples: Difficulty reading analog watches, using paper maps
- Historical parallel: Slide rules, calculating square roots by hand became obsolete with calculators
- Generally not serious problem as long as technology remains available
- Knowledge can be relearned if necessary
Reality Zone vs. Myth Zone
- Distinction by Robert Abelson and Dan Sperber/Hugo Mercier
- Reality zone: Beliefs about verifiable facts (beer in fridge)
- Myth zone: Beliefs expressing moral values, attitudes, identity
- Historical questions unknowable without modern archives/science
- Cosmic questions (origin of universe, why bad things happen)
- Truth/falsehood irrelevant - beliefs valued for being ennobling, entertaining, empowering
- Example: Hillary Clinton pizzeria conspiracy = "Hillary boo" (expressing hatred, not literal belief)
- Few people actually called police about pizzeria despite professing belief
- Enlightenment gift/curse: Expectation that all beliefs should be testable, justifiable, literally true
Game Theory and Collective Action Problems
- Social belief dynamics: Rational to assert beliefs accepted by your social group, even if preposterous
- Climate change as tragedy of commons:
- Individual sacrifice doesn't save planet if others don't participate
- Rational for individuals/countries to free ride
- Only solution: Change incentives through rules/punishment
- Carbon pricing: Makes conservation individually rational through economic incentives
- Technological solution: Make clean energy cheaper than dirty energy
- Avoid dead-end strategy of individual guilt for conservation
Social Media and Rationality
- Pinker initially resistant to blaming social media for society's problems
- Influenced by Jonathan Haidt's "The Constitution of Knowledge"
- Institutions that promote rationality: Peer review, fact-checking, empirical testing, open debate, slow deliberation, filtering
- Social media does opposite: Instantaneous reaction, no filtering, reputation for shareability vs. accuracy
- Historical irony: 1990s criticism of information oligopoly (NY Times, Encyclopedia Britannica)
- Noam Chomsky's "Manufacturing Consent" - wanted democratized information
- "Be careful what you wish for" - now too democratic, proliferation of fake news
- Uncharted territory requiring experimentation with algorithmic tweaks, external criticism
- Heavy-handed censorship not the answer
Education and Cognitive Enhancement
- Critical thinking, probability, logic should be woven into curriculum from early age
- Public discourse should not tolerate logical fallacies (anecdotal evidence, base rate neglect, ad hominem, argument from authority)
- Fiction consumption: Can expand empathy by inhabiting other perspectives (peasants, minorities, marginalized groups)
- Openness to evidence: Key cognitive virtue
- Keynes quote: "When facts change, I change my mind. What do you do, sir?"
- Goes against valuing steadfastness/unwavering conviction
- Research shows correlation with less conspiracy theory belief, less paranormal belief, less susceptibility to pseudo-intellectual gobbledygook
Historical Perspective on Irrationality
- Difficult to measure whether irrationality is increasing
- Historical consequential irrationality:
- Protocols of Elders of Zion → Holocaust
- Gulf of Tonkin fake news → Vietnam War
- WMD reports/9-11 connection → Iraq invasion
- Belief in paranormal phenomena: Flat over 50 years (astrology, ghosts)
- May not be getting worse, but not getting better either
- Availability heuristic may make current irrationality seem worse than historical examples
