A collaborative study, in which 160 research teams built models to predict six life outcomes based on social science data collected over 15 years, finds that even the best predictive models were not accurate; the results suggest that the use of predictive models in criminal and social systems may need to be reevaluated, along with the assumption that life trajectories can be predicted, according to the authors.
Article #19-15006: "Measuring the predictability of life outcomes with a scientific mass collaboration," by Matthew J. Salganik et al.
MEDIA CONTACT: Matthew J. Salganik, Princeton University, NJ; tel: 609-258-8316; e-mail: mjs3@princeton.edu
###
Proceedings of the National Academy of Sciences