Decisions Based on Qualitative and Quantitative Reasoning
Causal networks can be used to reason quantitatively or qualitatively about optimal decisions. Quantitative reasoning in causal probabilistic networks requires the specification of conditional probabilities, which in a medical application reflects the doctors current knowledge about the problem at hand. It is also necessary to specify a utility function, which reflects the priorities of the patient. A small medical example is analysed quantitatively. The results of the analysis are compared with the results from a previous qualitative analysis. Based on this example it is argued that although qualitative reasoning can eliminate a range of suboptimal decisions, the “interesting” decision must be based on a detailed quantitative analysis. The quantitative specification of a causal probabilistic network represents an unambiguous way of stating the opinion of the doctor (and of the patient). The firm theoretical foundation in probability and decision theory allows independent evaluation of the doctors’ and the patients’ assumption3.
Unable to display preview. Download preview PDF.
- Andersen, S. K., Andreassen, S., Woldbye, M.: Knowledge Representations for Diagnosis and Test Planning in the Domain of Electromyography. Advances in Artificial Intelligence, ed. Du Boulay, Hogg and Steels, Elsevier, 461–472, 1987.Google Scholar
- Andreassen, S., Jensen F.V., Andersen S.K., Falck B., Kjærulff U., Woldbye, M., Sorensen A.R., Rosenfalck A., Jensen F.: MUNIN An Expert EMG Assistant, In: Computer-Aided Surface or Needle Electromyography and Expert Systems in Diagnosis. p.p. 255–277. Desmedt, J. E. (ed.). Elsevier, 1989.Google Scholar
- Andreassen, S., Woldbye, M., Falck, B., Andersen, S. K.: MUNIN A Causal Probabilistic Network for Interpretation of Electromyographic Findings. Proc. of 10th IJCAI 1987, 366–372, 1987.Google Scholar
- Banks, G.: Artificial Intelligence in Medical Diagnosis. The Internist/Caduceus Approach CRR Critical Reviews in Medical Information, 1, 23–54, 1986Google Scholar
- Jensen, F.V., Olesen, K.G., Andersen, S.K.: An Algebra of Bayesian Belief Universes for Knowledge Based Systems. Institute of Electronic Systems, Aalborg University, R-88–25, 1988.Google Scholar
- Kim, J.H. and Pearl, J.: A Computational Model for Causal and Diagnostic Reasoning in Inference Systems. Proc. IJCAI 1983, 190–193, 1983.Google Scholar
- Lauritzen, S.L. and Spiegelhalter, D.J.: Local Computations with Probabilities on Graphical Structures and Their Application to Expert Systems.R.Statis.Soc. B, 2, 157–224, 1988.Google Scholar
- Long, W: A Program for Management of Heart Failure. Sigart Newsletter, n. 93, July 1985.Google Scholar
- Pople, H.E.Jr.: Heuristic Methods for Imposing Structure in Ill-structured Problems: The structuring of Medical Diagnosis. In: Artificial Intelligence in Medicine, Szolovits, P. (ed.). Westview Press, chap. 5, 1982.Google Scholar
- Schwarz, S., Griffin, T: Medical Thinking. The Psychology of Medical Judgment and Decision Making. Springer Verlag, 1986.Google Scholar
- Spiegelhalter, D.J. and Lauritzen S.L. Statistical Reasoning and Learning in Knowledge-bases Represented as Causal Networks. In: Lecture Notes in Medical Informatics, Reichertz, P.L (ed.), 105–112. Springer-Verlag Berlin Heidelberg, 1988.Google Scholar
- Patil, R.S. Causal Representation of Patient Illness for Electrolyte and Acid-Base Diagnosis. Laboratory for Computer Science, MIT, PhD Thesis, 1981.Google Scholar
- Wellman, M.P.: Qualitative Probabilistic Networks for Planning Under Uncertainty. In: John F. Lemmer (ed.). Uncertainty in Artificial Intelligence 2, 197–208. Elsevier Science Publishers B. V, 1987.Google Scholar