'Where quirky beliefs and preferences come from'

During the twentieth century successive economists conceived of human action, first as highly rational and subsequently as highly irrational as behavioural economists such as Daniel Kahneman identified apparently illogical responses in our decision-making. In Hidden Games, MIT behavioural economists Moshe Hoffman and Erez Yoeli attempt to explode both of these extreme views of human behaviour and instead replace them with a middle-ground perspective.

In order to achieve this, they adopt a game theory perspective. If you're not sure what it is, in the words of the authors, game theory is 'a mathematical tool kit designed to help us figure out how people, firms, countries and so on will behave in interacting settings'.

Importantly, Hoffman and Yoeli consider how such 'games' operate at a subconscious level to guide our decision-making. Hence, the title Hidden Games. In a nutshell, they suggest that in order to comprehend the human condition, we need to realise that much of our decision-making occurs at this unconscious level and that understanding this helps us to reveal our rational, but hidden, motives. In contrast, when we make decisions consciously, these are more likely to be bad ones leading to bad outcomes.

This sounds counterintuitive but their argument is based on the notion that unconscious decision-making draws on responses which helped our ancestors to survive and thrive. Such an evolutionary analysis will not appeal to all. In fact, as Steven Pinker has suggested, there appears to be a whole industry devoted to criticising the evolutionary approach.

But despite detractors, evolutionary approaches continue to have a growing influence on the behavioural sciences. Throughout Hidden Games, Hoffman and Yoeli guide the reader through three important levels of explanation that are used to help us understand why our unconscious games are played out: ultimate (rather than proximate), etic (rather than emic) and primary (rather than secondary).

I'll let the authors themselves explain each of these concepts (see our Q&A online) as, having read Hidden Games, I know they will do a better job of it than me. Their writing incorporates a large number of real-world examples such as Stockholm syndrome and unconscious bias, as well as anecdotes including why Eastern children savour spicy foods that Western ones would reject, and how John Nash was misportrayed in A Beautiful Mind.

This style does not, however, skimp on the science, but rather I feel aids in its assimilation. Personally, I found it to be a real page-turner, which is not a term normally associated with texts designed to introduce a new theoretical framework. Since Hidden Games did such a good job of whetting my appetite, I wanted to get a feel for the authors' main objectives in writing it – read on…

Reviewed by Lance Workman
Visiting Professor of Psychology at The University of South Wales

I enjoyed Hidden Games. Can you outline in a nutshell what you hope it might achieve?

We wanted to demonstrate another way of looking at and utilising game theory: to understand our quirky beliefs and preferences. We wanted people to see that game theory wasn't obviated by the 'behavioural' critique that people are often 'irrational'. And we wanted them to see how far a handful of simple, stylised models can take us in better understanding otherwise puzzling aspects of human behaviour.

Can you explain game theory to readers not familiar with it?

Game theory is just a set of models which specify the way incentives and information work and a set of analytical tools which tell us what to expect different agents to do. The models all have in common that they specify potential actions and the payoffs for every combination of actions. 

What does game theory add?

Sometimes the effects of incentives are pretty straightforward. If the price of milk goes down, people will buy more of it. If the job market rewards programming skills, more people will learn how to program.

But the effects of incentives are not always this straightforward. One way the effects of incentives can be hard to disentangle is when what you will get rewarded for doing depends on what others do, and vice versa. For instance, when two birds of the same species each want to nest in the same location, does it make sense to fight over this patch of land or vacate if you notice the other was there first? This really depends on what each bird can expect the other to do. There is no 'universally' optimal thing to do here.

Or what about if you are trying to sell your prized tulips? There are a handful of potential buyers. You are thinking of setting up an auction. Are there specific rules to your auction you should announce? This is, perhaps even less obvious.

Game theory can help in these cases.

In the case of the birds, game theory tells us that it's optimal for each bird to defend the territory aggressively, if she was the first to arrive there, if and only if, you are a member of a species where other birds do just that. In that case, the bird is said to be territorial – that's how biologists explain animal territoriality. But game theory can tell us even more. It tells us what conditions are necessary for territoriality to work. 

The birds, for instance, can't be too desperate for land, or too different in fighting ability. Game theory also tells us what can dictate which animal will fight: first arrival works, as does whether the animal is currently in possession, or whether the animal invested in building up the site, or what sex the animal is, but who is bigger or older or hungrier won't generally work.

The game theory also makes clear how to 'test' this explanation: randomly assign animals to who arrived first, and see if this affects how hard they fight or whether they run away, even if they've had no time to get used to the site or build up defences. Maybe even fool two into each thinking they arrived first, and see if an actual fight ensues. Ecologists have run just these experiments on territorial species, which is part of why they really buy into Maynard Smith's explanation in terms of 'the Hawk-dove game'.

When it comes to tulips, it turns out, various types of auction rules all yield the same expected revenue – for example, if you let people bid in sealed envelopes; give to the highest bidder and charge them the second highest bid; have a public posted price slowly descend until someone raises their hand, at which point give them the tulip at the posted price. But surprisingly, the way to maximise one's revenue is to pre-commit to a 'floor' –announce that if no one bids above a certain price, or if the publicly announced price falls below a certain level, then the item will be destroyed. And in fact, this is what tulip auctioneers did during the Dutch tulip mania. These highly prized flowers were crushed, if the price dropped below a certain pre-announced level.

Game theory is able to help in situations like these because the effect of incentives is really hard to disentangle without the help of a bit of math. But with a bit of math, not only can you decipher what's optimal for everyone to do, but also under what conditions everyone behaving optimally will yield which perverse effects. That's what makes it powerful for scientists.

Of course, the scientist still needs to be assuming people are responding to their incentives, that they are in some sense optimising, which may or may not be a reasonable assumption, depending on the setting, or may be justified in very different ways, depending on the application. But game theory can at least tell you what to expect if everyone is optimising. Which can be quite useful to know, and often hard to figure out without the game theory.

Your focus is on ultimate rather than proximate, etic rather than emic, and primary rather than secondary (rewards) explanations of human behaviour. Could you briefly outline how you use these three terms for our readers?

An 'ultimate' (a.k.a. 'functional') perspective helps answer why something got to be the way it is. This perspective is particularly useful when that thing is 'designed', sometimes by an actual designer, but often by processes like biological or cultural evolution. Chicks evolved to chirp to signal to their parents that they are hungry and healthy. This is an ultimate explanation. Notice, it doesn't rely on what's going through the chicks' minds – what they are 'thinking' – when they chirp. That would be a proximate explanation.

Likewise, suppose we are trying to understand people's taste in food. In the Middle Ages, contrary to popular belief, Europeans had a taste for foods that were rich in spices, and quite sweet – perhaps a bit like modern Moroccan food. Today, Europeans have banished spices and sugar to the dessert table, but spice cocktails remain in heavy use in other parts of the world – India, Ethiopia, Sichuan and Mexico, to name a few. 

A proximate, and not-so-satisfying explanation for these tastes would be that this is simply what people enjoy(ed) – what's tasty. An ultimate explanation might consider the costs of spices and sugar in the late Middle Ages, and the desire to signal wealth and status, or the need to suppress food-borne pathogens, especially in warmer climates without access to refrigeration.

Emic is analogous to proximate, but instead of focusing on the thoughts and feelings going through people's minds, it focuses on the kinds of explanations people within a culture tend to give. Moshe was raised an Orthodox Jew, and when he, like many young Orthodox boys, asked why Jews had to wear a kippah on his head, the answer he was given by his Rabbi was 'to remind us that God is above'. 

An impartial, outside observer, though, would give a different (etic) answer, which might go something like: at various points in history, Jews were forced to wear head coverings to demarcate them as outsiders, then Jews eventually came to wear these even when not required to, as a signal of commitment to the community.

As for primary versus secondary rewards, this distinction comes from animal trainers and the animal learning literature. Dogs like food, company, and a good belly rub. To the extent that these likings are innate and don't need to be learned, we would call them 'primary'. But dogs can also learn to like other things. Maybe they start to like the sound their leash or bag-o-treats makes when you pick it up. That learned excitement is a secondary reward. Your voice, saying 'good boy', might likewise do the trick. 

But dogs didn't evolve to like the sound of a leash or the words 'good boy'. They learned to like them because of their association with primary rewards like treats. We can apply this distinction to humans, too. We all evolved to like food, sex, companions, shelter, and status. But what about a good stamp collection? The reward that an avid collector gets from finding and acquiring a rare stamp isn't primary. It's learned; it's secondary. Likewise the warm feeling we get when we give our favourite charity some of our hard-earned cash.  

Keeping ultimate, etic, and primary explanations distinct from proximate, emic, and secondary ones makes it clearer how game theory can be applied, and what puzzles it can explain. Naturally, people have a tendency to 'explain' their quirky tastes, beliefs, and behaviours with recourse to proximate, emic, and secondary arguments. Game theory helps us dig past these – to understand the primary rewards at play, and offer etic, ultimate explanation for tastes, beliefs, and behaviours.  

You suggest that we are more likely to make good decisions when they are made unconsciously. This sounds counterintuitive. Why do you both feel decision-making works best this way?

This is an interesting question. It's not that non-deliberative decision-making is inherently better. We really do need to consciously deliberate to figure out how much change we should get when we hand the teller a fiver for a chocolate bar that costs 3.83.  

Our point in Hidden Games is that a lot of the optimising that we do – and, consequently, a lot of what we can explain with game theory – happens outside of our awareness, and has little to do with conscious deliberation. Just as chicks haven't the faintest clue about ultimate explanations, no one consciously calculated what kinds of spice cocktails to use in a Malabar curry to maximise their potency against local foodborne pathogens.  

One advantage to focusing on non-conscious optimisation of this sort is that conscious deliberation is notoriously buggy. People are terrible at understanding probabilities, they can't take an expectation, and they certainly can't get statistical laws like Bayes' Rule right – which they'd need to if they were consciously calculating payoffs in a game. Fortunately, we're not depending on them to do these things.

That said, non-conscious optimisation processes also have their bugs. Indians who live in the UK and the US like spicy food even though refrigeration, regulation and latitude render the climate in their adopted lands far less hospitable (to foodborne pathogens). Such cultural lags are par for the course when it comes to non-conscious forms of optimisation. To understand beliefs, tastes and behaviours like these, it's important to keep in mind: what were they optimised for?

You (politely) disagree with veteran behavioural economist Daniel Kahneman's views on decision-making. I'm sure that none of us wants to teach our grandmothers to suck eggs – but where do you think he has gone wrong?

We are huge fans of Danny Kahneman's work. Danny and the field he helped found have done a tremendous job documenting the many ways our more deliberative decision-making can go astray.

Of course, like anything, the behavioural approach can be taken too far. Behavioural researchers are sometimes quick to resort to 'irrationality' when there is a perfectly rational explanation. A great example of this is in the results of anonymous, one-shot experimental games. Subjects are often quite generous in these experiments, yet if you solve the game literally, the game theory says they shouldn't be. AHA!!! PEOPLE ARE IRRATIONAL!!! Or maybe, outside these experiments, subjects have plenty of good reasons to be generous (reasons game theory can help us uncover), and they're applying their generous instincts in the experiment even though those reasons are, due to the experimenters' devious machinations, absent.

Likewise, behavioural folks can sometimes be too quick to dismiss tools like game theory that make heavy use of 'rationality' assumptions. Yes, the behavioural critique has bite when people are relying on conscious deliberation, but less so when it comes to understanding beliefs, tastes and behaviours that are shaped by biological and cultural evolution.  

Since John Maynard Smith was one of my personal mentors, I'm really pleased to see how you've made use of game theory. But my impression is that it hasn't been in vogue for some time. Are you hoping that Hidden Games will lead to a resurgence of interest in Game Theory?

That would be great. We definitely hope the book helps researchers see the power of game theory to explain where quirky beliefs and preferences come from. And if it helps researchers better understand when the behavioural critique of game theory applies (and, ehem, doesn't) that'd be fantastic.

Finally, what's next for Moshe Hoffman and Erez Yoeli?

Moshe is planning to write a book that generalises this approach beyond game theory. What sorts of other tools can help us understand the primary rewards at play in shaping our beliefs and preferences? Erez is planning to write a book that applies these ideas to practical problems, and, particularly, to motivating people to do more good.