Aiding Human Operators with State Estimates.
Abstract
Three experiments are reported in which subjects must maintain a dynamic plant at or near some desired state by selecting actions based on a display of noisy state information and their own knowledge of the plant dynamic characteristics. In some cases, a normatively derived state estimate was displayed as an alternate information source, or decision aid. Results suggest that subjects simplify their own cognitive tasks significantly while degrading overall task performance only slightly. Models of this cognitive behavior include: substituting an adequate stimulus-response like algorithm for one which requires maintenance of an internal state estimate; retaining point estimates of state when distributions over states are presented; partitioning the state space so that the modelled state space is smaller and more manageable than the actual state space; and ignoring certain structural details of plant behavior which extend beyond convenient analogical models assumed to be present in the subject. the implications for the design of decision aids based on state estimation are discussed. Keywords: Supervisory control, Mental models, Human computer interaction.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jul 01, 1986
- Accession Number
- ADA174632
Entities
People
- James B. Rosenborough
- Thomas B. Sheridan
Organizations
- Massachusetts Institute of Technology