When data is scarce, it is difficult to screen the opinions of informed and uninformed experts. In spite of this difficulty it is possible to deliver incentives for informed experts to honestly reveal their views, and for uninformed experts to do no harm to a principal in the sense
that uninformed experts report the view the principal held originally (i.e., without the expert's report). This follows even if there is only a single data point to evaluate the expert's opinions and the expert's preferences over risk are unknown.
"At Least Do No Harm: The Use of Scarce Data."
American Economic Journal: Microeconomics,
Asymmetric and Private Information; Mechanism Design
Economics of Contract: Theory