UCI cognitive scientist uses crowdsourcing to predict celebrity deaths
- May 17, 2013
- Algorithms by Michael Lee show Ranker users' predictions to be surprisingly accurate
Participating in a celebrity death pool is a pretty macabre way to pass the time, but when multitudes of people do it the results can be fairly prescient.
UC Irvine professor of cognitive sciences Michael Lee (pictured) and colleagues found that the collective opinions of users of Ranker—a website that solicits crowd opinions on a variety of topics—predicted recent celebrity deaths better than individual users, chance or age. Lee used algorithms developed by his research team to analyze lists provided by 27 users. Lee found that 99 celebrities were included in at least one list, and at the time of analysis, six of the 99 celebrities had passed away. Lee’s modeling included a list of all 99 celebrities in an order that combined user rankings. The top 5 in this aggregated list were Hugo Chavez (already a correct prediction) Fidel Castro, Zsa Zsa Gabor, Abe Vigoda and Kirk Douglas.
“Successfully aggregating opinions is a cornerstone of prediction markets like the New York Stock Exchange, InTrade, and the Hollywood Stock Exchange,” said Ravi Iyer, Ranker’s chief data scientist and Lee’s partner in the study. “Our preliminary success in predicting box office receipts and celebrity deaths leads us to believe that we can harness the wisdom of crowds to predict a whole range of future outcomes on our platform.”
According to Lee, celebrity death pools are useful in the analysis of “crowd wisdom” as they can be assessed for their accuracy, unlike opinion-based polls, which are plentiful on the Internet.
“Predictions, like celebrity deaths, do eventually have a ground truth, so we are focusing our initial collaboration with Ranker on prediction lists,” Lee said.
-Laura Rico, University Communications