However, in my mind,(2)is much more important than(1). The reason is that in my perhaps-not-so-humble opinion, GT is best viewed not as a "bag of tricks but as a methodology -a select collection of internally consistent, general principles which helps you structure your own reasoning about strategic interaction in a given setting Games as Multiperson Decision Problems Decision theory enjoys an enviable status in economics. It is, often seen as the prototypical example of a rigorous, well-founded theory, which is easy to apply to the analysis of a wide variety of relevant decision situations in the social sciences: from Econ 101-type models to GE and finance. The link with GT might not be apparent. Indeed, it has been somewhat down- played in the past, because certain conclusions one can draw from strategic reasoning seem to be so much at odds with decision theory that the very existence of such a link might appear questionable(think about: "more information can hurt you. " However, the link is definitely out there, and recent research has brought it to the ore. What might be surprising is that the link was quite evident in the early days of GT, when people were mostly interested in zero-sum games. I shall presently state my case-and, in the process, review a related mathematical tool whose importance I simply cannot overemphasize: linear programming Making the connection First, I need to define a decision problem. Definition 1 A decision problem is a tuple (@, C, Fl, where Q2 is a set of possible states of the world, C is a set of consequences, and FCC is a set of acts The idea is that, if the decision-maker(dm) chooses act f E F, and the prevailing state is w E Q, then the consequence f(w)E C obtains. Consequences are supposed to represent all the dM cares about The dm does not observe the prevailing state of the world: that is, there is uncertainty about it. Still, she might be assumed to exhibit preferences(represented by a binary relation 2C FX F)among available acts. The classical"representation result exhibits a set of joint assumptions about >(and F)which are equivalent to the existence of a probability measure H E A(Q2)and a utility function u: C-R (unique up to a positive affine transformation)which represents x V,g∈F,f≥9兮/a(f(u)(du)≥/(g(a)(d)However, in my mind, (2) is much more important than (1). The reason is that, in my perhaps-not-so-humble opinion, GT is best viewed not as a “bag of tricks”, but as a methodology—a select collection of internally consistent, general principles which helps you structure your own reasoning about strategic interaction in a given setting. Games as Multiperson Decision Problems Decision theory enjoys an enviable status in economics. It is, often seen as the prototypical example of a rigorous, well-founded theory, which is easy to apply to the analysis of a wide variety of relevant decision situations in the social sciences: from Econ 101-type models to GE and finance. The link with GT might not be apparent. Indeed, it has been somewhat down￾played in the past, because certain conclusions one can draw from strategic reasoning seem to be so much at odds with decision theory that the very existence of such a link might appear questionable (think about: “more information can hurt you.”) However, the link is definitely out there, and recent research has brought it to the fore. What might be surprising is that the link was quite evident in the early days of GT, when people were mostly interested in zero-sum games. I shall presently state my case—and, in the process, review a related mathematical tool whose importance I simply cannot overemphasize: linear programming. Making the connection First, I need to define a decision problem. Definition 1 A decision problem is a tuple {Ω, C, F}, where Ω is a set of possible states of the world, C is a set of consequences, and F ⊂ C Ω is a set of acts. The idea is that, if the decision-maker (DM) chooses act f ∈ F, and the prevailing state is ω ∈ Ω, then the consequence f(ω) ∈ C obtains. Consequences are supposed to represent all the DM cares about. The DM does not observe the prevailing state of the world: that is, there is uncertainty about it. Still, she might be assumed to exhibit preferences (represented by a binary relation ⊂ F × F) among available acts. The classical “representation” result exhibits a set of joint assumptions about  (and F) which are equivalent to the existence of a probability measure µ ∈ ∆(Ω) and a utility function u : C → R (unique up to a positive affine transformation) which represents : ∀f, g ∈ F, f  g ⇔ Z Ω u(f(ω))µ(dω) ≥ Z Ω u(g(ω))µ(dω) 2