I've just come across a fascinatingly compact observation by I. J. Good:

Public and private utilities do not always coincide. This leads to ethical problems. Example - an invention is submitted to a scientific adviser of a firm...

The probability that the invention will work is p. The value to the firm if the invention is adopted and works is V, and the loss if the invention is adopted and fails is L. The value to the adviser personally if he advises the adoption of the invention and it works is v, and the loss if it fails to work is l. The losses to the firm and the adviser if he recommends the rejection of the invention are both negligible...

Then the firm's expected gain if the invention is adopted is pV - (1-p)L and the adviser's expected gain in the same circumstances is pv - (1-p)l. The firm has positive expected gain if p/(1-p) > L/V, and the adviser has positive expected gain if p/(1-p) > l/v.

If l/v > p/(1-p) > L/V, the adviser will be faced with an ethical problem, i.e. he will be tempted to act against the interests of the firm.

This is a beautifully simple recipe for a conflict of interest:

Considering absolute losses assuming failure and absolute gains conditioned on success, an adviser is incentivized to give the wrong advice, precisely when:

• The ratio of agent loss to agent gain,
• exceeds the odds of success versus failure
• which in turn exceeds the ratio of principal loss to principal gain.

You can see this reflected in a lot of cases because the gains to an advisor often don't scale anywhere near as fast as the gains to society or a firm. It's the Fearful Committee Formula.