Build and evaluate models that turn user-authorized signals and feedback into accurate, calibrated understanding of a person, and measure whether that understanding improves.