International Day Of Women in Statistics and Data Science

This session examines how scientific judgment can be made explicit in research design and decision making with mixed methods evidence, causal knowledge, expert elicitation, and Bayesian priors. Noor Qaragholi opens by showing how qualitative and quantitative evidence can inform strong, untestable causal assumptions in study design, illustrated through CACE estimation in a randomized behavioral intervention with substantial noncompliance. Utkarshani Jaimini then turns to Causal Neuro-Symbolic AI, a framework for building systems that are explainable, interpretable, and causally aware rather than limited to correlational pattern matching. Her work spans knowledge representation, causal inference, and applied machine learning. Anna Heath discusses expert elicitation for Bayesian clinical trial design, focusing on remote, real time exercises that help experts translate experience into probabilistic statements for prior distributions. Lu (Maggie) Qian closes by examining how Bayesian trial design can make scientific judgment explicit without allowing prior choice to become opaque, drawing on FDA guidance and reverse-Bayes methods as an audit tool. Together, the talks show how making assumptions and priors explicit can strengthen research design and support credible, decision relevant conclusions.