Using Sequences of Life-events to Predict Human Lives

Over the past decade, machine learning has revolutionized text analysis through flexible computational models. Beyond text, emerging transformer-based architectures have shown promise as tools to explore multi-variate sequences from protein structures to weather forecasts due to their structural similarity to written language.Human life trajectories are another type of process that has a strong structural resemblance to language. From one perspective, lives are simply sequences of events. We present a life2vec model that uses this similarity to adapt innovations from natural language processing to examine the evolution and predictability of human lives based on day-to-day event sequences.

This is the presentaiton used during the ODISSEI Lecture on 26 September 2023.