Qualititive Interfaces for Quantitative Models

点击翻译按钮获取中文摘要

This interdisciplinary project is concerned with relationships between formal quantitative methods used to aid diagnosis and decision making under undertainty and the informal, largely qual- itative mental models employed by the human beings who apply such methods. The goal is to be able to design effective user inter- faces to the quantitative methods. Topics to be investigated include the use of hierarchically structured influence diagrams to specify and display the qualitative structure of models; ways of representing, propagating, and analyzing vagueness and uncer- tainty; and quantitative and qualitative explanations of model behavior. Research will include psychological experiments to evaluate graphic tools for the probability encoding of uncertain influences and to investigate probability encoding on the basis of causal and diagnostic relations. The knowledge sought in this research is fundamental to the design of machines that can effectively aid human beings in a quintessentially human function: deciding important issues where much is known, but much remains uncertain.