2.8 Transferable Utility Example

Loading...

En provenance du cours de Stanford University

Game Theory II: Advanced Applications

293 notes

Stanford University

Game Theory II: Advanced Applications

293 notes

Popularized by movies such as "A Beautiful Mind", game theory is the mathematical modeling of strategic interaction among rational (and irrational) agents. Over four weeks of lectures, this advanced course considers how to design interactions between agents in order to achieve good social outcomes. Three main topics are covered: social choice theory (i.e., collective decision making and voting systems), mechanism design, and auctions. In the first week we consider the problem of aggregating different agents' preferences, discussing voting rules and the challenges faced in collective decision making. We present some of the most important theoretical results in the area: notably, Arrow's Theorem, which proves that there is no "perfect" voting system, and also the Gibbard-Satterthwaite and Muller-Satterthwaite Theorems. We move on to consider the problem of making collective decisions when agents are self interested and can strategically misreport their preferences. We explain "mechanism design" -- a broad framework for designing interactions between self-interested agents -- and give some key theoretical results. Our third week focuses on the problem of designing mechanisms to maximize aggregate happiness across agents, and presents the powerful family of Vickrey-Clarke-Groves mechanisms. The course wraps up with a fourth week that considers the problem of allocating scarce resources among self-interested agents, and that provides an introduction to auction theory. You can find a full syllabus and description of the course here: http://web.stanford.edu/~jacksonm/GTOC-II-Syllabus.html There is also a predecessor course to this one, for those who want to learn or remind themselves of the basic concepts of game theory: https://www.coursera.org/learn/game-theory-1 An intro video can be found here: http://web.stanford.edu/~jacksonm/Game-Theory-2-Intro.mp4

À partir de la leçon

Mechanism Design

The description goes here

2.1 Mechanism Design: Taste 3:18

2.2 Implementation 18:01

2.3 Mechanism Design: Examples12:00

2.4 Revelation Principle 9:23

2.5 Revelation Principle: Examples5:18

2.6 Impossibility of General Dominant-Strategy Implementation 11:26

2.7 Transferable Utility 9:05

2.8 Transferable Utility Example8:31

2.9 Mechanism Design as an Optimization Problem 19:03

Rencontrer les enseignants

Matthew O. Jackson
Professor
Economics
Kevin Leyton-Brown
Professor
Computer Science
Yoav Shoham
Professor
Computer Science

Hi all, it's Matt again.

So let's take a look at an example which will illustrate some of our transferable

utility mechanisms.

And in particular,

let's think of a fairly canonical example where we've got a set of citizens and

they're trying to make a decision between either undertaking a project or not.

So this might be say building a library, building a new school,

making some sort of investment if it's a,

could be a club deciding on making an investment of some kind.

So, there's a set of people and we'll keep track.

So N is just our set of individuals.

And now the outcomes we'll just keep track of as either 0 or 1.

So 0 means status quo, you don't do anything.

1 means you undertake the project, make the investment, and so forth.

And so the idea now is that the types represent somebody's

private value in this particular example for the two possible options.

So this is a valuation function.

So they basically have utilities for doing nothing, and utilities for

undertaking the project.

And undertaking the project is going to involve, possibly, some taxes and

other kinds of things, and maybe benefits.

So each person has utility of doing nothing and then the utility for

what happens if they undertake this investment.

Okay, so for

instance, one possibility is that somebody has a utility of 0 for doing nothing.

We can normalize it to 0.

And then if they did, undertook the project, this is a person who

likes the project, they would get some 4 utils for doing this.

So given that vi(0) is 0 and vi(1) is 4,

we can just represent that by a vector (0, 4).

So we could say that person has a vector of utility (0, 4), okay?

Another possible utility, maybe there's somebody who doesn't like the project.

That type could be represented by a 0 and a -2.

So what this does is it keeps track of how much each person values

not doing the project, how much they value doing the project.

And so now a person's type is just represented by these valuations.

And these valuations can be thought of also as transferable in the sense that we

can just pay them to compensate them.

So if you wanted to compensate this person down here for

the pain that they're enduring for building this thing, you would

have to pay them 2 utils in order to make them happy if you undertake this.

This person may be willing to pay up to 4 utils to have the project built.

Okay, so

in this kind of mechanism now we can think of this as kind of a direct mechanism.

People are going to announce their type.

So for instance if we keep things simple and they're either types who like it and

get 4 utils for it or don't like it and get -2, then there would be two possible

type vectors, the (0, 4) type and the( 0, -2) types.

In general you could have lots of types, but for this example,

let's just keep it simple.

And one possible mechanism would be to say, okay, look, if a majority of people

announce (0,4), if a majority of people like the project, then we'll build it.

If a majority of people don't like the project, we won't build it.

So we'll just take a majority vote, and then we'll build it or not.

So in terms of our notation of a mechanism, if a majority of people

announced the (0,4) types, then x, the outcome is going to be 1.

You undertake the project.

And if we're not doing any compensation, then everybody gets a pi of 0.

So basically, we just go ahead, we undertake the project, and

there's no transfers or subsidies back and forth in between individuals.

In contrast if the majority of people announce (0,

-2), then we don't undertake the project and

everybody ends up again, with a payment of 0.

So this is a simple example, just a straight majority vote over the project.

What's interesting about this mechanism?

Well it's actually a dominant strategy in this particular example

to announce your type, and better off voting to not

get the project built if I don't like it and voting for it if I do.

So truth is an equilibrium.

But sometimes what are the downfalls of this?

Sometimes some of the people get negative utility.

So sometimes the project is built, and there's a bunch of people getting -2s.

And also the choice doesn't maximize the total utility in this society, right?

So this mechanism isn't good in terms of maximizing the total utility.

It's not accounting for

the fact that these people like it twice as much as these people dislike it.

So one possibility would be to say, okay look, let's try a different mechanism.

Why don't we try, instead of having a majority,

if a third of the population likes it, then we'll go ahead and build it.

So now we're trying to do something where, instead what we take into account is these

people like it twice as much as these people.

So we'll weight their votes twice as much.

And we only need a third of the population to announce the v hat types, sorry,

the (0, 4) types, in order to have the project undertaken.

So we could change that mechanism, right?

So now, again, people announce their types.

And now if the number of people,

let's let m be the number of people who voted for it, so the (0, 4) types.

If that's bigger than a third of the population,

then we go ahead and undertake the project.

But now what we're going to do is we're going to try and

also compensate the people who don't like it, right?

So, there's a bunch of people suffering for it.

So, what we'll do is the people who announced the positive, so

they're the people voting for the project.

They're going to have a payment that they have to make which is

going to be n-m over m times 2.

That's going to be enough to cover the payments that we make to the n-m

other people who are voting against the project.

And those people will receive two payments.

So this is a mechanism where now we're allowing the people who are for

the project to have a stronger vote, but

then we're going to compensate the people who have pain inflicted on them.

And otherwise, if we get too few votes for it, then we don't build the project and

we don't make any payments, okay?

So this is a mechanism which now is making the choice that actually is best for

society in terms of the over all maximization of total utility.

It's also one that compensates the people who are receiving pain for it.

What's the difficulty with this particular mechanism?

Well the problem with this one, it does maximize total utility.

You never get a negative utility, but

now we've lost the fact that truth is an equilibrium, right.

Why have we lost the fact that truth is an equilibrium?

Well, let's suppose just for an example, think of a really, really large society.

And suppose you're a type that is a (0, 4) type.

And suppose for instance there's half of the people who you think might like it,

half might not like it.

Well, if we have a really large and everybody else is honest,

then the chance that my announcement is going to make a difference in

the undertaking of the project is really vanishing.

So the larger the society is then the chance that I have an impact on the vote,

for instance, if people's types are drawn independently becomes really tiny.

And so then all that happens is now my type is going to affect my payment, right?

So it affects whether I'm making a payment into the system or getting a payment out.

And we're going to lose truth in equilibrium because I realize my vote's

not going to make a difference.

Now if I'm a (0, 4) type,

maybe I pretend to be a type who's receiving pain from the project so

that people make a payment to me instead of me having to make a payment in.

So you can begin to see the different mechanisms

in these kinds of settings are gonnao have very different properties.

Some of them might be good in terms of truth.

Some might be good in terms of maximizing overall utility.

Some might be good in terms of compensating people, balancing so

that payments are balanced across a society.

There's going to be a whole series of different conditions we are interested in.

And the question is,

can we find mechanisms that are going to have nice incentive properties,

nice efficiency properties, and nice balance properties and so forth?

Can we get all of those at once?

And we'll take a look and

see that some mechanisms come pretty close to doing that.

It's not always possible to get everything, but

in a lot of contexts, we'll have mechanisms that do pretty well.

Explorer notre catalogue

Rejoignez-nous gratuitement et obtenez des recommendations, des mises à jour et des offres personnalisées.

Coursera propose un accès universel à la meilleure formation au monde, en partenariat avec des universités et des organisations du plus haut niveau, pour proposer des cours en ligne.

© 2018 Coursera Inc. Tous droits réservés.

Télécharger dans l'App Store

Disponible sur Google Play