On relationships, a game theoretic approach

Silicon Valley is as much a community as it is a place. Most of its citizens embrace that — over the long term — being a good one is best for them. But when they are overly focused on the short term, not only do they not get what they want, but can also shoot themselves in the foot.

Semil Shah (@Semil) got me thinking about this. He was talking about some unfortunate behavior he has recently seen, where people push and push to get what they want assuming it is costless.

The unfortunate thing for whoever Semil was talking about in his tweet (and I have no clue who it was) is that how they went about trying to achieve their goal was misguided. The best they could have hoped for was a pyrrhic victory. Much more likely, though, is that they were left with nothing.


If you only interact with someone once, doing whatever it takes to get your way can be rational. Inflict whatever costs are needed to drive the outcome you desire. They are simply externalities.

Whoever @Semil was interacting with likely thought of it as one time thing. But, in a community people interact with each other over-and-over. If he or she recognized this, how they interacted with Semil should have been much different.


Game Theory is the branch of economics and mathematics focused on strategic interactions between two or more parties. A favorite tool of game theorists is the Prisoner’s Dilemma where two prisoners need to decide whether or not to cooperate with each other in order to reduce the potential length of jail sentences they each face. One version of the Prisoner’s Dilemma assumes the prisoners need to make this call once while the other version has the prisoners making it over and over again (the iterated Prisoner’s Dilemma).

Classic (Single Play) Prisoner’s Dilemma — The classic version has two prisoners who are each accused of committing a crime and are being interrogated separately. The reason that the prisoner’s dilemma has its name is because when each participant makes the individually optimal choice (the technical term is defect, but it means to rat each other out), the participants collectively end up worse off. Conversely, for the two prisoners to both end up better off they each need to not say anything to the cops (the technical term is cooperate) trusting that the other is doing the same.[1]

The founder @Semil was talking about acted just like the prisoner ratting the other one out. They weren’t going to take no for an answer, plowing ahead whatever the costs. The fact that this founder destroyed their relationship with him was likely something he simply did not think about at the time.

Iterated (Multi Play) Prisoner’s Dilemma — If the two criminals were arrested day after day, and questioned continually, the incentive for one of them to throw the other under the bus goes down dramatically. The reason is simple. If you throw me under the bus today, I’ll throw you under the bus tomorrow. History — what happened the last time the game was played — matters. This is why when the Prisoner’s Dilemma is iterated participants can learn to cooperate and maximize their overall value, rather than self destructively attempting to optimize their individual well being. See: The Evolution of Cooperation and other works by Robert Axelrod along with Robert Aumann.

The most classic example involves negotiations of international accords. It would be in any one country’s best interest to not give up the right to spew CO2 into the environment. But, it is in the interest of all countries for there to be fewer greenhouse gases. Because there are constant efforts to curb greenhouse gases (and make international accords to that effect), the reputational costs of agreeing, or not agreeing, in any given round of negotiation are apparent.[2]


In life there appear to be very few single play “games”. The Silicon Valley community is one extremely complex web of overlapping, heavily iterated, “games”. In any one game it might feel better to “defect”, but doing so is incredibly short sighted.

If you want a long term career starting and financing companies, treating your interactions with other participants in this eco-system as a single play “game” is dangerous. You will run into everyone again. 7 years in Silicon Valley have made me realize this. It is better to have cooperated the last round, so they cooperate this round. And, you certainly don’t want them to defect on you.

The next time this founder wants something from Semil, or Semil is asked about him or her, what is going to happen? The founder needs to think about that question.


Thank you to Semil Shah (@semi) and Zac Townsend (@ztownsend) for their feedback.

[1] The development of game theory, and the prisoner’s dilemma, was contemporaneous to the development of nuclear weapons. This is not surprising as there was great interest in understanding the strategic logic around using nuclear weapons and how to use them “effectively” against the Soviet Union. Thermo-nuclear war is the ultimate single play game. For more on this see Brodie, Kahn, Jervis, and Waltz.

[2] This is an obviously simplified example. For more about the game theory of international environmental agreements and its relationship to iterated prisoner’s dilemmas see Porter, Wagner, Osang and Ambec.

Show your support

Clapping shows how much you appreciated Dan Kimerling’s story.