Joseph Jay Williams

Assistant Professor



Computer Science

Fields of Study

Areas of Interest

  • Technology-Based Interventions
  • Education
  • Health
  • Algorithms and Tests for Adaptive Experiments
  • (Bayesian) Multi-Armed Bandits


My Intelligent Adaptive Interventions lab spans Bayesian & Frequentist statistics, applied machine learning, human-computer interaction, psychology, education and health. Our applied statistics research focuses on designing/modifying algorithms and tests/analyses for Adaptive Experimentation. In an Adaptive Experiment (e.g. randomizing people 50/50 to two arms, like an old vs new version of a website), an algorithm automatically analyses data (e.g. which website get more people to click on paper links) to adapt the probability of assignment (e.g. 50/50 to 70/30 to 90/10). We are investigating "Statistically Considerate" Multi-Armed Bandit algorithms, which trade off giving participants better arms/interventions (e.g. websites) with enabling scientifically and statistically rigorous inferences. We are also investigating which statistical tests are appropriate for analyzing data from Adaptive Experiments. Students can read more.