Effects of Hybridization on Trust in Forecasting Advice

In three previous studies we studied people's trust in forecasting advice from different sources. We established that when offered advice by default, people don't differentially update their beliefs as a function of the source--human expert, algorithmic, or hybridized (a combination of both human and algorithmic) advice. Overall, they revise their forecasts only about 50% of the time, even though the advice would benefit accuracy on average (in each condition the advice was based on aggregate wisdom-of-crowds judgments). However, when asked to express an explicit preference, people prefer hybridized advice to both human and algorithmic advice across different subject domains (and in a direct comparison, prefer human or algorithmic advice depending on subject domain). In the third study, we also demonstrated that, when given a choice, people will choose hybridized advice more frequently than human or machine only advice, including in cases where viewing advice would be costly (their potential accuracy incentive bonus would be reduced). In the current study, we hypothesize that people will update more frequently when hybridized advice is available than when it is it not (instead only human or machine advice is available), and that the availability of hybrid advice will benefit accuracy.