Effect of Procedural Elements on Trust and Compliance with anImperfect Decision Aid

Abstract

We sought to establish to what extent incorporating a dichotomized procedural variable (in this case, maritime ‘rules of the road’) and incentives into a decision aiding algorithm would change a previously found non-compliance bias when the algorithm contradicted the known procedure. We also sought to examine the relationship between trust in and dependence on an automated system. An experiment was conducted using a simple, simulated maritime collision avoidance task featuring an imperfect, but highly reliable (87%), decision aid. Adding the dichotomous procedural variable into the algorithms recommendations increased compliance with the system, even for recommendations that violated learned procedures. Performance was still not perfectly calibrated to the actual reliability of the system (underreliance and under-trust). Results also revealed the dissociation between rated trust in, and behavioral dependence on decision aiding automation.