It's worth noting that humans completely suck at almost any task even simple statistics can be applied to. Even world class experts will typically be beat by simple linear regression. See here for many examples: http://lesswrong.com/lw/3gv/statistical_prediction_rules_out... And this is very old research with methods that are very primitive compared to modern machine learning (most of this was done on pencil and paper!):
>Wittman (1941) constructed an SPR that predicted the success of electroshock therapy for patients more reliably than the medical or psychological staff.
>Carroll et. al. (1988) found an SPR that predicts criminal recidivism better than expert criminologists.
>An SPR constructed by Goldberg (1968) did a better job of diagnosing patients as neurotic or psychotic than did trained clinical psychologists.
It's completely amazing we allow human doctors to make diagnoses at all. At the very least algorithms should always be part of the process (but note that humans given the results of an algorithm still do worse than just the algorithm on it's own.)
The problem is there has been enormous resistance to the use of algorithms in medicine. People irrationally distrust algorithms and strongly prefer human judgement. Even when they know the algorithms are superior. Psychologists have actually studied this phenomenon and have named it "Algorithm Aversion": http://opim.wharton.upenn.edu/risk/library/WPAF201410-Algort... This isn't even getting into the institutional resistance to change or having people lose their jobs to robots.
>Wittman (1941) constructed an SPR that predicted the success of electroshock therapy for patients more reliably than the medical or psychological staff.
>Carroll et. al. (1988) found an SPR that predicts criminal recidivism better than expert criminologists.
>An SPR constructed by Goldberg (1968) did a better job of diagnosing patients as neurotic or psychotic than did trained clinical psychologists.
It's completely amazing we allow human doctors to make diagnoses at all. At the very least algorithms should always be part of the process (but note that humans given the results of an algorithm still do worse than just the algorithm on it's own.)
The problem is there has been enormous resistance to the use of algorithms in medicine. People irrationally distrust algorithms and strongly prefer human judgement. Even when they know the algorithms are superior. Psychologists have actually studied this phenomenon and have named it "Algorithm Aversion": http://opim.wharton.upenn.edu/risk/library/WPAF201410-Algort... This isn't even getting into the institutional resistance to change or having people lose their jobs to robots.