The answer is probably from a different area of computer science/game theory: consensus.
This has been practiced by patients for a long time by asking for a second opinion. However, this no longer works, because doctors now all share your medical record, and the second doctor risks getting sued if they disagree, while they bear no risk at all for agreement with the first doctor.
So, the consensus rules have to actively break the model for collusion. I’m not sure how to do this, and actually I think this is a relatively unexamined area of game theory, because most models assume tautological correctness of consensus.
Breaking consensus requires an oracle, which is by definition, a source of truth that operates outside of the consensus model, and cannot be affected by participants. This would be something like a piece of diagnostic equipment that can’t be altered or censored. It also requires a very large incentive for correctly predicting the oracle result against the consensus opinion.
So here is a straw man mechanism:
1. Assume condition can be assessed by a diagnostic test oracle
2. See many doctors, solicit opinions on treatment
3. For each proposed treatment, the doctors bet on the post-treatment test result
4. Choose the treatment with the best most money wagered on a good result
5. Add your service payment to the pool
6. Undergo treatment, take post-treatment diagnostic test
7. Distribute all money to the correct predictors
This has been practiced by patients for a long time by asking for a second opinion. However, this no longer works, because doctors now all share your medical record, and the second doctor risks getting sued if they disagree, while they bear no risk at all for agreement with the first doctor.
So, the consensus rules have to actively break the model for collusion. I’m not sure how to do this, and actually I think this is a relatively unexamined area of game theory, because most models assume tautological correctness of consensus.
Breaking consensus requires an oracle, which is by definition, a source of truth that operates outside of the consensus model, and cannot be affected by participants. This would be something like a piece of diagnostic equipment that can’t be altered or censored. It also requires a very large incentive for correctly predicting the oracle result against the consensus opinion.
So here is a straw man mechanism: 1. Assume condition can be assessed by a diagnostic test oracle 2. See many doctors, solicit opinions on treatment 3. For each proposed treatment, the doctors bet on the post-treatment test result 4. Choose the treatment with the best most money wagered on a good result 5. Add your service payment to the pool 6. Undergo treatment, take post-treatment diagnostic test 7. Distribute all money to the correct predictors