Back to Main Agenda

Back to Post-Meeting Feedback

Human behaviors are emerging to be critical determinants of human health since more than half of premature mortality can be attributed poor health behaviors. Successful mitigation of this situation will require computational predictive modeling.  Human behaviors are inherently embedded in multiple temporal scales and multiple grains of analysis. Because of their complexity and inherent uncertainty, assessing and predicting individuals’ behaviors is challenging for classical ML approaches and would benefit from combination of mechanistic, principle-based models with ML.  While several speakers (and the title of the workshop) included the word “behaviors” the actual number of projects in that area was limited. I worry that many researchers in this workshop may have concerns with the uncertainty associated with behavioral measures and the strength of the basic principles.  Although I understand these concerns, I would like to point out that some 200 years ago, scientists were similarly challenged by physical and biological measurements.  Understanding these challenges in psychology and behavioral sciences gave rise to the field of “Mathematical Psychology” where measurement processes are treated with equal or even greater rigor than in physics, see for example very accessible text by F. S. Roberts ((1985) Measurement theory, New York: Cambridge University Press).  The challenges of measuring subjective responses in humans to perceptual (physical) stimuli gave rise to increased rigor in models of physical phenomena. These approaches are now routinely used in measurement in psychology and economics. These developments combined with rapid advances in behavioral and physiological sensing can provide solid foundations for the development of hybrid, dynamical and stochastic models of behaviors that could predict human behavior, perhaps as well as good car salesmen do ;-).  In addition to their theoretical implications, such models would be useful in many engineering areas including self-driving cars and human-robot interactions.  Perhaps we should consider organizing an entire workshop on rigorous modeling of human behaviors.

- Misha Pavel