I am very new to mechanism design so I am a bit lost among the concepts, even the most basic ones.

One thing that is not clear to me is the concept of direct mechanism and its relationship with the revelation principle.

From what I got a direct mechanism is a mechanisms, in which, instead of messages, the agents report their preference/type. Are we assuming this is a truthful report?

I am asking that because the revelation principle is stated as "anything that can be accomplished by any mechanism can actually be accomplished by a direct revelation mechanism that is individual rational and incentive compatible" in this notes but as "a social choice function can be implemented by an arbitrary mechanism (i.e. if that mechanism has an equilibrium outcome that corresponds to the outcome of the social choice function), then the same function can be implemented by an incentive-compatible-direct-mechanism (i.e. in which players truthfully report type) with the same equilibrium outcome (payoffs)." on Wikipedia.

So I don't get if truthfulness is implied by the direct mechanism or if it is an additional requirement.

  • $\begingroup$ A social choice function is either direct (by royalties) or fraudulent. An honest mechanism is blind. $\endgroup$ Jul 30, 2022 at 14:31

2 Answers 2


Incentive compatible here means that you have no incentive to lie. A direct mechanism that is not incentive compatible would entice some agents to report a false type.

  • $\begingroup$ Thanks! Thus, for example, the revelation principle in dominant strategies is saying if a social choice function is implementable, than it is implementable using a direct mechanism, i.e. truth-telling is always a dominant strategy. Is it the right way to think about it? $\endgroup$
    – PhDing
    Dec 7, 2018 at 1:24
  • $\begingroup$ Not sure waht you mean by "always a dominant strategy", the always seems redundant to me. If a social choice function is implementable in dominant strategies you can construct a mechanism where truth-telling is a dominant strategy. $\endgroup$
    – Giskard
    Dec 7, 2018 at 5:44

A direct mechanism is just a mechanism in which the action space of each participant is equal to the type space and an outcome function maps these type reports into an outcome. There is no requirement that this type report has to be equal to the true type. That is the answer to your question "Are we assuming this is a truthful report?" is NO. However, in an incentive-compatible or truthful direct mechanism, players prefer to reveal their true type.

Take an arbitrary (direct or indirect) mechanism in which all participants $i \in I$ can take one or more actions from some action space and together all players' actions determine some outcome. A strategy in this game maps a private type $t_i$ into actions $\sigma_i (t_i)$. Hence, a type profile $t = (t_i)_{i \in I}$ leads to an equilibrium action profile $\sigma(t)=(\sigma_i(t_i))_{i \in I}$ which then leads to an outcome determined by the mechanisms outcome function $o_{indirect} (\sigma(t))$.

Instead of letting the participants play some complicated game, we can ask all participants for their type and then "let the mechanism play the game for them". That is, each player reports her type $t_i$ and the mechanism takes the type profile $t$ and maps it into an outcome $o_{direct}(t) = o_{indirect} (\sigma(t))$ for all $t$. We have replicated the indirect mechanism. Is it incentive compatible to report the TRUE type? Yes, it is, because $\sigma$ is an equilibrium in the original mechanism. If a participant had an incentive to report some type $t'_i \neq t_i$, there would also be a profitable deviation $\sigma_i (t'_i) \neq \sigma_i (t_i)$ in the indirect mechanism. A contradiction to $\sigma$ being an equilibrium. Note that this equilibrium does not have to be in dominant strategies.

  • $\begingroup$ To be clear: The way in which I constructed the incentive-compatible direct mechanism is essentially the proof of the revelation principle. $\endgroup$
    – Bayesian
    Mar 17, 2019 at 19:25

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.