Primary Submission Category: Causal Inference and SUTVA/Consistencies Violations
Bias in sequential decision-making for stochastic service systems
Authors: Gabriel Zayas-Caban, Juan Camilo David Gomez, Amy Cochran,
Presenting Author: Gabriel Zayas-Caban*
Decision-makers in many service systems are often confronted with making a random number of decisions sequentially over time. How sequential decisions are made may depend on the decision-maker’s perception of prior and analogous decisions, and how these prior decisions led to specific outcomes. This phenomenon, whereby prior decision-making or experiences influence current or future decision-making, has been termed sequential bias. Sequential bias violates a core assumption in causal inference that the decision for one person does not interfere with the potential outcomes of another. By drawing the connection between sequential bias in service systems to dynamic treatment regimes, and extending these latter settings to allow for a randomized number of decisions, we are to define and identify average causal effects for quantifying sequential bias. Subsequently, we propose estimators, and derive properties thereof. In a case study, we use our approach to demonstrate that the decision of a provider to route a patient one way in the Emergency Department has a measurable impact on the care of future patients.