[throwaway729]

But if there's no payout then there's no incentive to say don't know. Whereas if there is a payout then the optimal strategy is clearly "dont know", assuming equal payouts for correct and dont know. Why would you take the risk of being wrong?

[hx87]

Quote from paper:

> The amount paid for “don’t know” responses was also assigned randomly, and was a fraction of the amount offered for a correct response: 20% of the payment for a correct response with probability .33, 25% with probability .33, and 33% with probability .33.

[throwaway729]

Thanks, that's exactly what I was looking for but it's hard on a phone.

Qwrusz's comment may be relevant here.