Human causal discovery from observational data

Hashem AI, Cooper GF. Human causal discovery from observational data. In: Proceedings of the Annual Symposium of the American Medical Informatics Association (1996) 27-31.  PMID:  894721

Utilizing Bayesian beliefnetworks as a model of causality, we examined medical students' ability to discover causal relationshipsfrom observational data. Nine sets ofpatient cases were generatedfrom relatively simple causal beliefnetworks by stochastic simulation. Twentyparticipants examined the data sets and attempted to discover the underlying causal relafionships. Performance waspoor in general, except at discovering the absence ofa causal relationship. This work supports the potentialfor combining human and computer methodsfor causal discovery.

Publication Year: 
1996
Faculty Author: 
Publication Credits: 
Hashem AI, Cooper GF.
AttachmentSize
PDF icon Hashem.pdf784.8 KB
^