Prediction by Replication: People Prefer Prediction Algorithms That Replicate the Event Being Predicted
Lin Fei, Berkeley J. Dietvorst · Journal of the Association for Consumer Research · 2023
Consumers can use automated prediction algorithms, like navigation apps and recommendation systems, to forecast the outcome of to-be-determined events like how long a drive will take or how much they will enjoy a movie. However, little is known about what drives consumers’ preferences for prediction algorithms. The authors propose that consumers prefer prediction algorithms that replicate the event being predicted by going through the same process that generates event outcomes. In seven studies, the authors find that consumers like replicative prediction algorithms, even when they may lead to worse performance. Furthermore, the authors find that consumers shift toward choosing better performing algorithms after receiving performance feedback but still prefer replicative algorithms holding performance constant. These findings suggest that marketers can increase adoption of prediction algorithms by framing them as replicating (or designing them to replicate) the process that generates event outcomes.