Active Inference: A Competency for Making Decisions in Uncertain Situations
Abstract
Army officers need to make decisions while facing unprecedented levels of volatility, uncertainty, complexity, and ambiguity. We propose that active inference - the ability to infer the true state of the world by revising beliefs that generate invalid predictions - is pertinent to Army officers' ability to make sense of evolving situations they have never encountered before and about which they have incomplete information. A prototype active inference measure - the Decisions Over Time (DOT) task - was developed to measure how participants learn a complex structure of rules based solely on iterative exposures to stimuli that change systematically over time. Participants inferred the rules governing changes in stimuli across multiple iterations by making choices and inferring from outcomes, developing a mental model of the rules governing the pattern of change. We found that the DOT task differentiates active inference ability levels along two distinct dimensions: (a) an ability to learn from prediction errors, and (b) a sensitivity to prediction error. Future research will further refine the efficacy of the DOT task for assessing Army officers' capability to make sense of evolving and uncertain situations. The DOT task may be applied in assessing officers' active inference preferences and strategies to inform tailoring of decision making training to individual learning needs.
Document Details
- Document Type
- Technical Report
- Publication Date
- Apr 01, 2024
- Accession Number
- AD1226410
Entities
People
- Alex Francisco
- Cary Stothart
- James M. Nye
- Jared Peterson
- Rhett Graves
Organizations
- U.S. Army Research Institute for the Behavioral and Social Sciences