Leão B F, Rocha A F
Instituto de Cardiologia do Rio Grande do Sul, Porto Alegre, Brazil.
Methods Inf Med. 1990 Jan;29(1):30-40.
This paper proposes a methodology for knowledge acquisition (KA) from multiple experts, in an attempt to elicit the heuristic rules followed by the physician in diagnosing twelve frequently occurring congenital heart diseases (CHD). Twenty-two pediatric cardiologists and twenty-three general cardiologists were interviewed with this technique; 274 interviews were conducted, 169 with the 22 experts, 105 with the 23 non-experts. A graph formalism was employed to represent their reasoning model, leading to the construction of a "mean reasoning model" for each diagnosis, separately for experts and non-experts. The results indicate that experts, compared to non-experts, tend to build knowledge representation models (KRM) that are smaller and less complex. Qualitative differences in information utilization between the two groups were also observed. Entropy analysis suggests a greater objectivity and cohesion of the experts' model.
本文提出了一种从多位专家获取知识（KA）的方法，旨在引出医生在诊断十二种常见先天性心脏病（CHD）时遵循的启发式规则。采用该技术对22名儿科心脏病专家和23名普通心脏病专家进行了访谈；共进行了274次访谈，其中对22名专家进行了169次访谈，对23名非专家进行了105次访谈。采用图形形式主义来表示他们的推理模型，分别为专家和非专家构建了每种诊断的“平均推理模型”。结果表明，与非专家相比，专家倾向于构建更小、更简单的知识表示模型（KRM）。还观察到两组在信息利用方面的定性差异。熵分析表明专家模型具有更高的客观性和连贯性。
