Computers as surrogate decision makers?

A fascinating article can be found here: Can computers make life-or-death medical decisions? The jist of the article is that research on surrogate decision making in hypothetical situations shows that people only make the 'correct' decision on behalf of others 68% of the time. (correct in this context means simply whatever the other would have chosen) However a computer using a fairly simple decision procedure was right 78% of the time.

The procedure used was that

most people want life-saving treatment if there is at least a 1% chance that following the intervention they would have the ability to reason, remember and communicate. If there is less than a 1% chance, people generally say they would choose not to have the treatment.

The researchers are now looking at boosting its effectiveness by taking into consideration other factors such as religious background etc. These sorts of initiatives provide an interesting challenge for defenders of strong accounts of autonomy in health care, since in some cases it seems others might know what we want better than our nearest and dearest and in some cases better than even we might know. This is suggested for example by the research on commercial kidney donors in Iran that I mentioned in this post: get out of jail free organs? over at INEIRA, if 76% of donors subsequently want commercial kidney donation banned, and 99% of them would have retrospectively preferred to have begged or got a loan rather than sell their kidney, does respect for autonomy mean we ought to respect the decision to sell or not?