B1156
Title: The origins of unpredictability in life trajectory prediction tasks
Authors: Ian Lundberg - Cornell University (United States) [presenting]
Abstract: Why are life trajectories difficult to predict? Inspired by recent developments in public policy, machine learning, and computational social science, the question is engaged from a perspective that embraces qualitative and mathematical reasoning. The research design combines in-depth, semi-structured interviews with 40 families, using recent scientific mass collaboration results and an ongoing multi-decade longitudinal study of thousands of families. The qualitative evidence uncovered in these interviews, combined with a well-known mathematical decomposition of prediction error, helps identify some unpredictability origins. These lead to conjecture that high unpredictability will be the norm, rather than the exception, for life trajectory prediction tasks. Ideas about how the conjecture could be assessed empirically and its implications for social scientists and policymakers are presented.
