Silicon Valley’s elite startup accelerator is getting more selective. Paul Graham recently announced that the number of companies accepted to Y Combinator, the Mountain View (Calif.) program called the “Harvard of entrepreneurship,” will likely shrink to fewer than 50, from 84, over the summer because “more things than usual broke” with the larger batch. Looking for signs indicating which applicants would run into trouble led Graham to an unusual admission of bias: The entrepreneurs that Y Combinator interviewed in the afternoon were more likely to fail than those accepted in the morning. Graham’s conclusion: “It turned out that, like judges, we were more tolerant after lunch.”
Graham’s discovery is just one example of how decisions can be influenced by unexpected factors. Hidden bias is a risk in all sorts of systems that aspire to be objective: Pharma-funded drug research finds more favorable outcomes than government-backed trials. Hiring managers are less likely to call back job applicants named Lakisha and Jamal than Emily and Brendan. University tech transfer officers reviewing the same invention are less likely to recommend commercializing it if they think the inventor was a woman. And even middle-aged white guys in Silicon Valley feel they have to dress down and shave their gray hair to compete with younger, hoodie-clad candidates for top jobs at startups.