Modern WisdomHow Much Does Google Know About Me? | Seth Stephens-Davidowitz | Modern Wisdom 134
CHAPTERS
- 0:00 – 1:59
From ad-click optimization to “data for social good”
Seth and Chris open by discussing Seth’s day-to-day work as a data scientist and author, and why many people enter data science for money before seeking meaning. Seth frames big data as a tool that can be redirected from ads and finance toward socially valuable insights.
- •Seth’s split identity: data scientist + writer (R code and Google Docs)
- •Why many data scientists feel unfulfilled optimizing ads/finance outcomes
- •The growing desire (especially among younger cohorts) to apply data to social impact
- •Early framing: data can reveal truth and guide better decisions/policy
- 1:59 – 5:43
The “How Big Is My Penis?” story: what private searches reveal
The conversation turns to the provocative early insight that men search Google about penis size more than any other body part. Seth uses it to illustrate how online behavior exposes insecurity and absurdity that people hide in public.
- •Men’s most common penis-related searches and the insecurity behind them
- •Why a search engine becomes a confessional space
- •The contrast between public persona and private curiosity
- •Humor as a gateway to serious human-psychology insights
- 5:43 – 8:41
Why the book is called Everybody Lies: surveys vs. behavioral data
Seth explains the central thesis: people lie to others and even to anonymous surveys, skewing traditional research methods. Internet data—especially Google searches—often provides a more accurate window into what people actually think and do.
- •Social desirability bias in surveys (voting, sex frequency, media consumption)
- •Behavioral data can outperform self-reports (e.g., search intent vs. claimed intent)
- •Google searches as high-signal indicators of real concerns
- •Pornography as a major new dataset for understanding sexuality
- 8:41 – 10:34
The art of data science: creativity, pattern-finding, and cultural oddities
Seth argues that data science is not purely technical—it requires creativity to identify meaningful “nuggets” in massive datasets. He shares surprising examples (like India-specific searches) that traditional research would likely miss.
- •Creativity prevents ‘drowning’ in data and helps find signal
- •Example insight: India-specific “my husband wants…” searches
- •Regional fetish patterns and how they emerge unseen
- •Why some truths only appear at scale and in aggregates
- 10:34 – 15:01
Getting Pornhub data—and why academia underuses it
They discuss Pornhub’s annual stats and Seth’s experience obtaining and analyzing Pornhub datasets for his book. Seth notes the methodological conservatism of academia and the missed opportunity to study large-scale, revealed-preference sexual behavior.
- •Pornhub’s internal analytics and annual reporting as a public window
- •How Seth convinced Pornhub to share data
- •Why sociologists/sex researchers often stick to older methods
- •Credentials and legitimacy: why some topics get taken seriously (or not)
- 15:01 – 19:41
What porn data shows: women’s preferences, universality, and “why” questions
Seth shares findings that surprised audiences: certain violent/rape-themed porn is disproportionately popular among women, and this pattern appears broadly across cultures. The discussion expands into how difficult it is to explain “why,” not just report correlations.
- •Women’s porn-viewing patterns can differ sharply from assumptions
- •Cross-cultural comparison: fantasies not strongly linked to women’s societal treatment
- •Distinguishing descriptive facts from causal explanations
- •Limits of interpretation and the need for theory-building beyond data
- 19:41 – 22:58
Sexual orientation signals: lesbian porn among straight women and closeted men
Seth describes the popularity of lesbian porn among women who identify as straight, contrasting it with much lower rates of gay porn consumption among men. He then explains how regional patterns suggest closeted homosexuality in areas where it’s harder to be openly gay.
- •A substantial share of women’s porn views are lesbian-themed despite straight identity
- •Men’s gay porn share is smaller but geographically stable
- •Mismatch between self-reported gay identity and gay porn searches in conservative states
- •“Gay test” searches as a sign of internal conflict and stigma
- 22:58 – 25:24
Search strings as narratives: suicide, stigma, and the herpes insight
Chris asks about other revealing ‘search sequences,’ and Seth describes work on suicide-related searches. A key discovery: herpes diagnoses can trigger suicidal ideation among young people, driven largely by stigma and fear rather than physical severity.
- •Using sequences of searches to infer evolving mental states
- •Herpes as a surprisingly common trigger for suicidal thoughts among youth
- •How stigma amplifies distress beyond medical reality
- •The ethical potential: using search insights to design better interventions
- 25:24 – 29:43
Role-model searches: “celebrities with herpes” and what people are really seeking
Seth explains that people often search for celebrities who share their condition as a way to reduce shame and find hope. The tragedy: for herpes, search results often surface celebrity denials rather than supportive disclosures—potentially worsening stigma.
- •“Celebrities with X” as a common coping/normalization behavior
- •Depression and other conditions: many celebrities openly disclose to fight stigma
- •Herpes results skew toward denial and rumor-control
- •Policy/communications implication: targeted stigma reduction could save lives
- 29:43 – 34:01
Elections and prediction: what search behavior can reveal about voting
The discussion shifts to the 2020 U.S. election and how internet data might help forecast outcomes. Seth shares a subtle indicator: the order people type candidate names (e.g., “Trump Clinton” vs. “Clinton Trump”) correlates with preference and may reveal subconscious leanings.
- •Why early-cycle prediction is hard: turnout signals emerge late
- •Resource asymmetries (Bloomberg spending) as model edge-cases
- •Candidate-name order in searches as a predictive behavioral cue
- •Links to subconscious choice and free-will debates
- 34:01 – 38:27
Dating insights from recorded speed dates: what predicts a second date
Chris pulls a practical example: what language increases second-date probability. Seth summarizes research that mined recorded speed-date conversations, revealing cues like supportive phrases, laughter, hedge words, and how much a woman talks about herself.
- •Supportive language (“that must have been tough”) increases attraction
- •Laughter is a strong indicator of interest in men’s success rates
- •Hedge words (“maybe,” “kinda”) predict lower interest regardless of content
- •More ‘I’ usage by women correlates with mutual success on the date
- 38:27 – 53:16
The next book: data-driven life decisions (happiness, parenting, neighborhoods)
Seth outlines his upcoming book focused on using data to make better choices across life domains. He shares counterintuitive happiness findings about alcohol timing and major parenting insights suggesting neighborhood and role models matter more than most in-home tactics.
- •Kindle highlights showed readers wanted actionable self-improvement insights
- •Happiness pings: alcohol boosts mood more during boring tasks than fun ones (with caveats)
- •Parenting research: household effects are smaller than people assume
- •Neighborhood and peer role models strongly shape children’s outcomes; ‘subtle’ parenting works best
- 53:16 – 59:23
Other datasets and controversies: abortions, racism, Wikipedia births, Facebook fandom
They survey additional platforms Seth has analyzed beyond Google: DIY abortion searches mapping to access restrictions, Stormfront activity, Wikipedia birthplace patterns, and Facebook fandom in the NBA. Seth notes that some problems are easier to find insights in than the highly-optimized stock market.
- •DIY abortion searches spike where legal access is restricted; ‘missing pregnancies’ inference
- •Stock market prediction is harder due to intense competition and incentives
- •Wikipedia: notable people disproportionately born in college towns/cities (genetics + exposure)
- •Facebook NBA analysis: black players show higher fandom after controlling for performance
- 59:23 – 1:02:01
Wrap-up: anonymity, personal tech habits, and where to find Seth
Chris closes by reflecting on the power—and limits—of anonymous aggregated data: it reveals patterns without identifying individuals. Seth jokes that the biggest behavioral change is Googling himself more, then shares how listeners can find his work online.
- •Anonymous/aggregate data can be powerful without targeting individuals
- •Big data’s promise: insight into human nature and policy levers
- •Minimal personal behavior change despite studying sensitive datasets
- •Where to follow: search “Seth Everybody Lies” for his profiles and work
