[EM] manipulation free method?

[Forest W Simmons]

It's true that there can be some incentive to lie about whom you think 
is the most likely to win, but there is no point in lying about your 
preference order.

Here's a version that reduces the incentive to lie about perceived 
probabilities:

Each person indicates both a guess as to the winner, and a number alpha 
between zero and ten indicating how confident they are in their guess.  
The larger alpha, the greater weight in the community estimate.  

Here's the kicker.  On your ballot, the probability distribution is a 
weighted average between the guess you submitted and the community 
distribution.

If your guess is bogus, then you wouldn't want it to have much weight 
on your own ballot, so you will choose a small value of alpha, and it 
won't influence the community distribution much either.

FWS


Warren Smith wrote:


>this is almost certainly manipulable.
>
>Cannot we set up scenarios where if you be dishonest about the
>most-likely-to-win-guy, you change the winner?
>
>It seems interesting nevertheless.
>