Ethics and the Prisoner's Dilemma: Using Game Theory to Understand Morality

TLDR: Ethical social norms solve the prisoner dilemma and other sub-optimal equilibria. This is why evolution has made humans moralistic and why ethics are very important to a well-functioning society. Don't confuse apathy for equanimity. Ultimately, it is not rational to be apathetic.

“I don't want to sell credit to people who are going to hurt themselves with it. You should only sell products that are good for the people who use them. Some disagree with this, but I know I'm right. That is to say, you're talking to a Republican who admires Elizabeth Warren.” — Charlie Munger

I've been meaning to write about my thoughts on ethics for some time now, but this past election has compelled me to sit down and finally do so. Many "rational thinkers" that I have been following seem to think that ethics are useless and/or meaningless, and that apathy is a more rational approach. You get the impression that apathy is a badge of honour to rational thinking. Nothing could be further from the truth. It's true that equanimity is an important input for rational thinking, but be careful not to confuse apathy for equanimity. A true rational thinker should recognize that ethics are extremely important to rational thinking. I will try to explain how and why.

First, let's talk about the prisoner's dilemma, the classic game from game theory. I'm not going to explain the game in detail—that's what Wikipedia is for—but the situation can be described by the following decision table:

The Nash equilibrium—what I call the "stable outcome"—of the prisoner's dilemma is that both players lose, even though it is entirely possible for them both to win if they had strategically cooperated. (Econ wonks would say that the outcome isn't Pareto efficient.) This is important because it demonstrates a situation where two individuals both behaving selfishly will be worse off than if they both behaved selflessly. This is not what free market ideology says is supposed to happen.

Humans have evolved to follow a heuristic of selfishness—and for good reason: in most cases, behaving selfishly not only makes the individual better off, but everyone else too. Indeed, this is the standard argument in favour of free markets, and it is absolutely correct. But what the prisoner's dilemma is showing is that the selfish heuristic has a bias in some situations which we need to watch out for. Which situations? And how can we fix it?

The solution to the prisoner's dilemma is very simple: all we have to do is convince both players to behave, in a sense, selflessly. And, thankfully, this is exactly what evolution has provided for us through ethics and social norms.

Humans are naturally evolved to adopt social norms—which I like to think of as evolutionary tools for social engineering—and it is these social norms that allow humans to overcome coordination failures. What happens, exactly, to the prisoner's dilemma when "not being selfish" becomes a social norm? The individuals are compelled to not behave selfishly, essentially removing selfishness as a strategic option, and pulling them towards the more optimal outcome where they both stand to win.

Indeed, this is what I think it means to be ethical: a behaviour is ethical if, once adopted by everyone, it solves a situation resembling the prisoner's dilemma. And to get everyone to adopt an ethical behaviour, it must become a social norm. And since there is a huge advantage to adopting ethical social norms, they have been evolutionarily rewarded. I usually refer to ethical social norms as morals.

Now of course, two individuals might be able to overcome the prisoner's dilemma on their own without the need for an ethical social norm if each person recognized their impact on the other, reasonably expected the other to as well, and were thus able to cooperate in some unspoken way (through the use of a Schelling point, for example).  And sometimes, I'm sure, such cooperation occurs. At scale, however, when enough individuals are involved, behaving selfishly will always seem to be unharmful to others and not even Schelling points will save you. This is why ethics are so important: without ethical social norms, a prisoner's dilemma involving millions of people becomes nearly impossible to overcome.

Another insightful example is the ultimatum game. In that game (again, see Wikipedia), what is "rationally" expected is that the second player will accept any non-zero offer from the first player. But that doesn’t happen in real experiments with real people. Instead, the second player rejects offers that it deems unfair. But what is "unfair"? Well, that is driven by the individual's a priori ethics. When it comes to dividing the pie, it seems that we expect people to behave ethically. As we should! Because that is, in the long-run, the rational thing to do.

Ethics are closely related to what evolutionary theorists call proximate cause and ultimate cause. Proximate cause is "I eat because I am hungry", while ultimate cause is "If I didn’t get hungry, I wouldn't eat, I wouldn't survive, and neither would my gene pool". If we invoke ultimate cause, plenty of behaviour that seems irrational (locally) is actually quite rational (ultimately). Among other things, this explains morality, altruism, and ethics, and is why these traits have been so evolutionarily successful.

By now I hope you can see that it's actually quite easy to think of prisoner-dilemma-like situations: all we have to do is imagine scenarios where everyone is better off when everyone behaves ethically. Here are some:

  • Example: Why would someone take a small risk to prevent a stranger from getting hit by a bus? Because the tendency to care about the well-being of others, in the long-run, puts us all where we are today.
  • Example: Why do people vote when their votes are unlikely to change the outcome? Because the tendency to vote makes our collective decisions more representative.
  • Example: Why didn't my neighbour rob my house when he knew I was on vacation? Why didn't that passerby steal an apple from the fruit stand instead of paying for one? Because the tendency to respect property rights allows our trade-based society to function.
  • Example: Why don't people kill other people they disagree with? This one's obvious, right?

The irrational thing to think is that a person behaving ethically in any of the above examples is "irrational". Get your head out of the sand.

So how should we think about ethics and elections? Well, this is how I think about it:

Here we have a prisoner's dilemma (voter's dilemma?) that is inherent to any democratic election. It is simply not in Voter A's interest (locally) to vote out of concern for Voter B, and vice versa. But ultimately, it is in each of their interests: both would be much better off if both voted with concern for the other. And since voting happens at scale, this voter's dilemma is nearly impossible to overcome without some sort of ethical social norms. And that is why, you see, it is unethical to be apathetic. Not only that, if you're not behaving ethically, then you're free-riding off everyone that is.

I have often said that understanding society can be reduced to understanding the "initial value problem" of morality. In other words, morality plays an important part in determining where society is headed. In a sense, morality is the ultimate engineering problem. Given the morality we've seen in this past election, what can we say about where US society is headed? Are you apathetic and, therefore, behaving unethically?

Why are Founder CEO's so Effective? Arrow and the Behaviour of Firms

The maxim in Silicon Valley is that founders make the most effective CEOs. Assuming that's true, can we understand a deeper reason for it? I think we can. To do so, let's talk about how groups make decisions.

  • Famously, Arrow's impossibility theorem shows that no voting system will be "rational" in the usual sense.
  • A rational voting system means that:
    • If every individual prefers option A over option B, then A should win over B. (Pareto Condition)
    • If A wins over B, and B wins over C, then A should win over C. (Transitivity)
    • If A wins out of the choice of {A,B}, then B should not win out of the choice of {A,B,C}. Either A wins or C wins. (Independence of Irrelevant Alternatives)
    • All preferences are allowed. There are no restrictions on what any individual can prefer. (Unrestricted Domain)
    • No single voter can choose the election outcome. (No Dictator)
  • Arrow's theorem says that it is impossible to have a voting system that satisfies all of these conditions all the time.
    • Note: The voting system is abstracted away. Arrow's theorem applies to any voting system. The only assumption is that individuals vote based on their ordinal (and not cardinal) preference.
  • Another way of describing Arrow’s theorem is: Groups of people cannot reliably make rationally decisions.
    • Or: Groups of rational agents tend to be irrational.
    • Or: A single individual making decisions will always be more consistent than a group.
  • In other words, for a group to always behave rationally, its voters must be irrational or it must be run by a rational dictator.
    • That is: Either the voters are irrational or the voting system is irrational.
  • Why does this relate to firms? Think of a firm as a collection of individuals whose daily actions (i.e. votes) determine the direction of the firm.
    • If this analogy holds, Arrow’s theorem applies: A firm will only reliably behave rationally when it is run by a dictator.
    • More loosely: Firms will behave more rationally when the power of the CEO is more concentrated.
  • This is why large bureaucratic organizations, where power is dispersed throughout the organization, will often seemingly behave irrationally, lacking clear direction and regularly making illogical decisions.
  • And hence the preference for founder CEOs who have the moral authority to wield concentrated power within the organization.

So yes there is a theoretical case to be made for founder CEOs.

The Nature of the Firm: Why do firms exist?

  • In his seminal paper The Nature of the Firm, Ronald Coase attempted to answer the question: Why are there firms?
  • This question of "Why are there firms?" could instead be phrased as "Why do people have jobs?"
  • Or invert the question: "Why isn't everyone self-employed?"
  • To approach this question, let's take a step back and think of a production system comprising of many individuals represented as nodes.
  • Each individual contributes to a different part of the production process, but also requires inputs from other individuals throughout the system.
  • The coordination between these individuals occurs through price signals in the free market — i.e. market transactions.
  • Through price signals in the free market, the individuals collectively reach an equilibrium of what they produce, how much they produce, and their rewards for producing. For econ wonks, this is called the Walrasian Equilibrium.
  • But in the real world, that's not the whole story. It turns out that groups of individuals often join to work together — that is, as a firm.
  • And what is most interesting is that the firm organizes within itself without price signals. Instead, the firm is centrally planned by the entrepreneur.
  • As Coase says, "the distinguishing mark of the firm is the super-session of the price mechanism."
  • But why do these firms emerge? Why don't those individuals continue to act independently and coordinate through price signals in the free market?
  • Coase says the reason that firms emerge is because of transaction costs. These are costs that naturally emerge from using the free market.
  • Transaction costs are things like: information asymmetry, escrow fees, contract writing and negotiation, regulation compliance, the lack of an established market, brokerage fees, sales taxes, the risk of trade secrets being leaked, etc.
  • Transaction costs act as friction to market transactions, making it costly and difficult to use the price mechanism of the free market.
  • Coase says that the creation of the firm can reduce these friction by bringing people "under one tent" and removing the need for individuals to trade with each other in the free market.
  • Instead, individuals provide their services to other individuals inside the firm according to the firm's organizational structure. This removes transaction costs that would have otherwise been encountered in the free market.
  • Examples:
    • Instead of contracting a lawyer every time a legal question arises, a firm will hire the lawyer as an employee and form a legal department.
    • Instead of contracting a designer every time a new prototype is required, a firm will hire the designer and form a design team.
    • Instead of contracting an engineer, a firm will....
  • That is why firms exist: To reduce transaction costs associated with using the free market.
    • And in what world would firms not exist at all? In a world with zero transaction costs.
    • Without transaction costs, the free market equilibrium (Walrasian Equilibrium) would prevail.
  • In this sense, the firm is the anti-thesis of the market. Unlike the market, the firm is centrally planned, removing the need for price signals between individuals.
    • Firms can then trade with other firms and individuals using price signals in the free market.
    • Firms are thus blobs of centrally planned economies operating within a free market ecosystem.
    • Firms essentially take groups of individuals and make them resemble individuals in the marketplace. (Note that an individual is also centrally planned, in a sense.)
  • The firm centralizes what was once decentralized when the cost of decentralization is too high.
  • The creation of firms helps the ecosystem resemble the idealized free market equilibrium (Walrasian Equilibrium) that would arise if there were no transaction costs.
    • Outside the firm, production is directed by price signals through market transactions.
    • Within the firm, production is directed by a central planner (the entrepreneur), removing the need for market transactions.
  • Coase predicts that ecosystems with larger transaction costs will tend to create larger firms. On the other hand, smaller transaction costs tend to create smaller firms.
  • Coase also predicts that more individuals will tend to create more firms and larger firms. On the other hand, fewer individuals tend to create fewer, smaller firms.
  • In the limit where transaction costs go to zero, the firm will go extinct and everyone will be self-employed.
  • In reality, an ecosystem will have many firms of many sizes and structures, producing many kinds of goods, and include many self-employed individuals navigating the free market.

Bank-runs are simply the collapse of the network effects surrounding credit

The bank runs in Greece have been on my mind the past few days and in order to properly think about what is happening, I have been gathering my thoughts on the nature of bank runs in general.

  • Network effects are powerful because they have momentum.
  • This momentum works in both directions: both when the network is growing as well as when the network is collapsing.
  • Since network effects can collapse on themselves just as easily as they can grow upon themselves, people often label the collapse of any network effect as a “bubble.”
  • But bubbles are actually the collapse of unsustainable network effects. Not all collapsed networks are bubbles.
  • Money and, more broadly, credit are subject to network effects.
    • (Indeed, modern money derives all its value from the network effect.)
  • Banks are systemically dependent on the network effects of money and, in particular, credit.
    • All money is credit, and credit is just another word for trust.
    • In other words: The more people trust a bank, the more trustworthy that bank becomes.
  • Bank runs are simply the collapse of the network effects surrounding credit, as people lose trust that their deposits can be withdrawn.
  • This is why, like all other network effects, bank runs are self-fulfilling, contagious, and very difficult to stop once they have started.
    • Because if everyone does withdraw their cash, there will not be enough cash to go around.
  • And this is why the role of the central bank (which can create as much cash as it wants) as the lender-of-last-resort is so crucial to financial stability.

The European Central Bank has said that it will not act as the lender-of-last-resort to halt the Greek bank run. The interesting question now is whether the bank run will spread into Italy and Spain. And what the European Central Bank will do about it.


What is 21 Inc. trying to do?

Ben Thompson today published an article on Bitcoin and the start-up company 21 Inc. I then somehow found myself in a twitter debate about Bitcoin and 21 Inc.'s business model with the notorious Walt French.

It turns out that it's very hard to express your ideas on Twitter. So below I've outlined my thesis:

  • Bitcoin is valuable for internet-of-things devices, allowing devices to algorithmically exchange property rights.
  • Ideally, IOT devices should be convenient to use with a "plug and go" design, just as normal devices exist today.
  • One solution to achieve this "plug and go" is for IOT devices to have access to bitcoins somehow embedded into its design, without the need for the user to "set it up" and link it to their own Bitcoin address. But how?
  • 21 Inc. aims to solve this issue by embedding IOT devices with its bitcoin-mining chips.
  • 21 Inc. will provide bitcoins on demand to the IOT devices that use its chips and in return, 21 Inc. will keep all bitcoins mined from those devices.
  • This works because the amount of bitcoins each IOT device needs is small and with high frequency. On the other hand, the amount of bitcoins each IOT device earns is large but with very low frequency.
  • Similar to a bank that borrows short-term and lends long-term, 21 Inc. will profit from the difference in bitcoins mined and consumed over the life of all the devices in its portfolio. This is also analogous to how insurance works by pooling risk.
  • In the end, 21 Inc. will be providing users a way to "pre-pay" for bitcoins that their IOT device will use over its lifetime without the hassle of set-up.

I think that's pretty much 21 Inc.'s plan at the moment.