The success of intelligent agents in clinical care depends on the degree to which they represent and work with human decision makers. This is particularly important in the domain of clinical risk assessment where such agents either conduct the task of risk evaluation or support human clini- cians with the task. This paper provides insights into how to understand and capture the cognitive processes used by clinicians when collecting the most important data about a person’s risks. It attempts to create some theoretical foundations for developing clinically justifiable and reliable de- cision support systems for initial risk screening. The idea is to direct an assessor to the most informative next question depending on what has already been asked using a mixture of probabilities and heuristics. The method was tested on anonymous mental health data collected by the GRiST risk and safety tool (www.egrist.org).
Artificial Life and Intelligent Agents: Second International Symposium, ALIA 2016, Birmingham, UK, June 14-15, 2016, Revised Selected Papers