Crowds of online workers can be used to guess human preferences where there is little data to work with, according to the University of Illinois at Urbana-Champaign’s Peter Organisciak and colleagues. The researchers hired crowdsourcing workers on Mechanical Turk and had them make personal recommendations. They presented 100 different salt and pepper shakers and 100 photos of different types of meals to the crowdsourcing workers and asked them to give each one a suitability rating out of five for a target person. The team only provided the workers with a small sample of the individual’s actual taste in shakers and food. The researchers report the crowdsourcing workers did well in that the average rating from the top three recommenders matched the target person’s own ratings to within half a star. Organisciak says websites such as Amazon and Netflix use algorithms to guess what consumers want, but they need to learn from large amounts of data. He notes using crowds instead of algorithms to determine preference is useful in personal data sets for which training an algorithm is impossible. The findings will be presented at the Conference on Human Computation & Crowdsourcing in Pittsburgh in November.