Would you trust the judgment of a machine or a physician? That is a pretty simple question. A seminal paper by Dawes, Faust and Meehl (1989) claim that you should trust the machine. The compare statistical or actuarial methods that are not based on human judgement against physician’s own decisions.
In reviewing the literature, they find:
Eliminating research that did not protect sufficiently against inflated results for actuarial methods, there remain nearly 100 comparative studies in the social sciences. In virtually every one of these studies, the actuarial method has equaled or surpassed the clinical method, sometimes slightly and sometimes substantially
The statistical methods outperformed physicians even when physicians were provided additional information.
However, it is still possible that physicians may be superior to machines. How so?
- Information advantage. Physicians may be able to gain information from patients tone of voice and better extract information from patients to improve decision-making.
- Factoring in treatment experience. When picking between treatment A and B, treatment A may be superior on average. However, if a physician has extensive experience using treatment B–and say, has better experience knowing how to control side effects–a physician treatment decision could be the ‘wrong’ decision from a global perspective but the right one based on their experience.
- Better defining what the ‘right’ decision is. The right decision for the average patient may differ from the right decision for an individual patient. For instance, some patients may prefer more effective treatment even if there are more side effects; other patients may prefer treatments with fewer side effects even if the treatment is less efficacious. While machines could potentially weight outcomes if patients entered their preferences, physicians may be able to better understand patient preferences. Further, physicians may understand patients may up-weight or down-weight certain types of outcomes. Alternatively, some errors in judgement may be more severe than others and thus physicians may place different weights on Type I vs. Type II errors.
- Rare facts and the broken leg problem. “In psychology this circumstance has come to be known as the “broken leg” problem, on the basis of on an illustration in which an actuarial formula is highly successful in predicting an individual’s weekly attendance at a movie but should be discarded upon discovering that the subject is in a cast with a fractured femur…The clinician may beat the actuarial method if able to detect the rare fact and decide accordingly.”
- Dawes RM, Faust D, Meehl PE. Clinical versus actuarial judgment. Science. 1989 Mar 31;243(4899):1668-74.