Radio Replay: I, Robot

Episode Summary

Executive Summary: The episode argues that digital traces—especially Google searches, platform algorithms, and robot interactions—reveal hidden truths about human behavior that surveys often miss. Seth Stephens-Davidowitz shows search data exposing boredom, sexuality, racism, bias, and illness risk, while Kate Darling explains how naming and anthropomorphizing robots shapes empathy, ethics, and behavior toward machines and possibly toward other humans.

Main Topics: Search data as a map of hidden behavior (Priority: 5/5): Stephens-Davidowitz explains how anonymous searches capture what people really think, want, and fear, often more honestly than surveys or self-reports. Correlations in big data reveal social patterns (Priority: 5/5): Examples such as unemployment, porn searches, solitaire, and health queries show that seemingly odd online behavior can predict real-world conditions and outcomes. Racism, politics, and implicit bias in search behavior (Priority: 5/5): Searches for racist terms correlate with regional racism and political support patterns, while parental searches reveal gendered expectations about sons and daughters. Big data versus self-knowledge in relationships and media (Priority: 4/5): Netflix, Facebook, and other platforms use behavioral data to outperform people’s own predictions about what they will watch, want, or need in relationships. Ethics of predictive systems and health warnings (Priority: 5/5): The discussion raises whether companies should alert users to possible illness based on search histories, highlighting the promise and creepiness of predictive health models. Human attachment to robots and anthropomorphism (Priority: 5/5): Kate Darling shows that naming robots and giving them lifelike motion leads people to treat them as social beings, even when they know they are machines. Robot ethics, bias, and possible legal protections (Priority: 4/5): The conversation explores whether abusive behavior toward robots is morally troubling in itself and whether robots may merit limited legal protections similar to animals.

Key Arguments: Anonymous digital behavior is often more truthful than direct questioning because people lie to others and themselves in surveys. Odd correlations in search data can expose large-scale social realities, such as unemployment patterns, sexuality, and regional racism. Search behavior can predict political outcomes, suggesting that racial animus played a measurable role in Obama’s and Trump’s electoral support. Big data can reveal implicit bias, as shown by gender differences in how parents search about sons versus daughters. Algorithms often outperform human self-assessment because people are overly aspirational about future habits and choices. Predictive systems can be beneficial, especially in medicine, but they also raise ethical questions about privacy and unsolicited warnings. People readily anthropomorphize robots; naming, movement, and familiar design cues trigger empathy and moral hesitation. How people treat robots may matter ethically because repeated abuse of lifelike machines could desensitize users or signal deeper callousness.

Data Points: Unemployment-linked search correlation: Pornography site was the single most highly correlated search with unemployment rate - Stephens-Davidowitz described search terms that tracked unemployment better than expected queries like jobs or benefits. Another unemployment correlation: Solitaire was also highly correlated with unemployment - Used as an example of boredom/leisure behavior among unemployed people. Self-reported male attraction to men: 2.5%–3% - Survey estimate in the U.S. for men saying they are primarily attracted to men. Gay male pornography search share: About 5% of male pornography searches - Search data suggested a higher level of same-sex interest than surveys reported. Pornhub viewing volume: 2.5 billion hours in 2015 - Used to illustrate the gap between public self-reporting and actual consumption of pornography. Obama-related racist searches: More searches for a racist term than for 'first black president' in some states - Highlighted after Barack Obama’s election as evidence of hidden racism in search behavior. Regional racism pattern: East of the Mississippi higher than west of the Mississippi - Search data suggested the strongest divide in racist search volume was East-West, not North-South. Parent searches about sons: More likely to use terms like 'gifted' or 'genius' - Parents searching 'is my son...' often expressed intellectual optimism. Parent searches about daughters: More likely to use terms like 'overweight' or 'ugly' - Parents searching 'is my daughter...' revealed appearance-focused concern. Loan default correlation with God: 2.2 times more likely to default - A peer-to-peer lending study found loan applicants mentioning God were much more likely to miss repayment. Pancreatic cancer search study: Indigestion followed by abdominal pain flagged higher risk - Microsoft Bing data revealed subtle symptom sequences predictive of future pancreatic cancer diagnosis.

Pivotal Quotes: "there's just some things you just can't ask a real person in real life and you need to ask Google" — Shankar Vedantam: Opening framing for why search data can reveal hidden truths people won’t disclose directly. "I think that, yeah, that kind of is what I'm saying" — Seth Stevens-Davidowitz: Response to whether racist search patterns imply racial animus among many Trump supporters. "if you can be diagnosed with pancreatic cancer four weeks earlier, you have a much better chance of survival" — Shankar Vedantam: Ethical question about whether search platforms should warn users of possible illness. "we really do perceive them as lifelike and it is really offensive to us to see them be abused" — Kate Darling: Explaining why people resist harming robot dinosaurs and why robot ethics may matter.

Implications: Digital exhaust is becoming a powerful social sensor, but it can expose prejudice, vulnerability, and health risks in ways that outpace current ethics and law. As robots become more lifelike, society may need new norms for treating machines—and for preventing hidden biases from being baked into technology.