Adam Curtis: The Trap

[johnwhilley]

Adam Curtis: caught in the liberal trap

“Human beings will always betray you”, proclaims the opening slogan of Adam Curtis’s new BBC series, The Trap. (11 March 2007).

It could, more helpfully, have read, “liberal journalists will often confuse you”. It’s not that I’m averse to the more lateral take on such subject matter. Intellectual theses of this sort are sadly absent from our ‘reality’-obsessed schedules these days. But, like Curtis’s Power of Nightmares before it, we have here yet another illustration of what passes for critical liberal explorations of power, and the celebratory liberal reviews that accompany them.

The production of The Trap is impressive: dark and squeamish archive, quirky montage and an ethereally arresting soundtrack. As with Power of Nightmares, Curtis also offers some useful historical insights, in this first instalment, “F… You Buddy”, showing how applied Game Theory helped define Cold War thinking.

Alas, like Power of Nightmares, The Trap gets caught up in its own contrived attempts to make large universal statements from a peculiarity of sources. And, again, as with Nightmares, it appears not to notice the corporate elephant in the room.

Curtis begins with this liberal statement of political elite motivations, at home and abroad:

[Peter Fainton]

Good review John.

Link to video of the programme(real player): http://www.indybay.org/uploads/2007/03/11/1_fuck_you_buddy.rm
http://www.indybay.org/uploads/2007/03/18/2_the_lonely_robot.rm

From: http://www.thedossier.ukonline.co.uk/video_cover-ups.htm

Cheers

[Raoul Djukanovic]

[johnwhilley]

Cheers, Peter, and thanks for the useful links.

Hi Raoul,

Nice to hear from you. Thanks for those constructive comments.

[Raoul Djukanovic]

Thanks for your considered reply John.

You write:

[Peter Fainton]

A couple of other points in relation to this programme:

Curtis claims that Nash's game theory was used to underpin the Cold War nuclear strategy, even though he was diagnosed paranoid schizophrenic. This was based on the fact that Nash believed he was involved with government agents on a project to 'save the world.' What was the Cold War if not a project to save the world? And did controlling elites use Nash's own game theory on him in order to discredit him by branding him a lunatic? So that even if Cold War spies investigated his work it would be without credibility because he'd been diagnosed 'insane'. As Curtis demonstrated by reference to Laing's work the medical establishment in America were quacks anyway, commonly admitting people for care who were not insane, as demonstrated by the 'thud' experiment. Perhaps they calculated that game theory would only establish equilibrium if the enemy did not know you were using it.

My second point is on using game theory to improve the 'efficiency' within public services, particularly health as referred to in the programme. It is questionable from a number of perspectives to apply game theory to medical care. As in manufacturing, you can be incredibly 'efficient' in making the wrong things, but if this is not 'effective' use of your resources you will go out of business. Of course if efficiency is the core motivation behind targets then you may complete more operations and get financial incentives for achieving targets but more of your patients may die because the operations were rushed and therefore not an effective use of your resources. Waiting lists may go down as mortality rates rise. Effective diagnosis and treatment is a core component in medical practice.

Denial of adequate care to the elderly or long-term disabled may improve efficiency and allow resources to be targeted more effectively to more deserving cases but in establishing equilibrium in this way those neglected will die or suffer unnecessarily. The ethos in the NHS used to be care and compassion for the sick but care and compassion is not efficient and uses resources that could be spent on military instruments of domination and control, such as renewing Trident. Sick people have been re branded 'bed-blockers' and are booted out of hospital at the earliest opportunity to free up capacity, sometimes whilst they're dying from illness they're suffering from, back into their own homes with relatives being expected to provide acute care without medical knowledge or training. This is often accompanied by demands from social services to 'asset strip' the sick person, sell their home from under them and empty any savings accounts they may have. Of course, as long as the Doctors and consultants are hitting their targets the amount of trauma suffered by the sick and dying and their immediate families is all part of the new equilibrium that Curtis refers to.

There has been a fundamental change in the provision of all health services and Curtis may well be right that the new thinking directing it is based on game theory and targets. Dead people can't complain they just become statistics in the new equilibrium. Waiting lists are the new core targets, rather than effectiveness of medical treatment, partly because this is the focus of media attention.

[johnwhilley]

Hi Raoul,

[johnwhilley]

Thanks for that, Peter.

Yes, I basically agree with all your points and sentiments concerning the assault on the NHS and deteriorating state of healthcare.

[Raoul Djukanovic]

Hi John,

Thanks again for replying. You write:

[johnwhilley]

Thanks, Raoul.

[Raoul Djukanovic]

Thanks, John. I appreciate you taking the time to share your thoughts and respect, of course, your right to critique whoever or whatever you like, however you see fit.

Nevertheless, as I keep stressing, the subtext of your critique remains the same: "that which [you] regard as problematic" is Curtis' failure to have made the film you think he should have made. Hence my suggestion that you spell out how it might look.

You write:

[johnwhilley]

Thanks, Raoul.

[Raoul Djukanovic]

Thanks, John. I agree that this exchange has probably reached its productive limits, so I'll reply to your last post below and leave you to respond to my response if you see fit.

You write:

[johnwhilley]

Thanks, Raoul. I’ll leave things there regarding your own comments.

But, just for the record, I read this Daniel Simpson critique of ML some time ago and, as you might imagine, strongly reject its viewpoint.

Best wishes,
John

[johnwhilley]

In Part 2 of The Trap (18 March 2007), Adam Curtis develops his thesis on the forces underlying the mass erosion of civil freedoms and how this crisis of political and social life can be traced to the cold war elements of game theory.

Having repeated the outline claims of the series, Curtis’s discussion continues with an analysis of John Major’s government from 1991. Against a trademark party broadcast of Major returning to his ‘working class roots’, Curtis relates how his administration had set about imposing a new kind of reformist purge of the state bureaucracy. Underlying all this, Curtis claims, was the same adoption of game theory-type target management. The market libertarian James Buchanan, who had already described politicians and public servants as “zealots”, repeats that they were never motivated by public ideals, but were really just in it for themselves.

Curtis sees the result as the triumph of “market democracy”, underpinned by the claims that only the market, not the state, can be relied upon to deliver economic rewards and individual freedoms. Intrinsic to this is the idea that self-interest and suspicion of others – the fundamentals of game theory – actually help to sustain social stability rather than generate chaos.

But, to repeat, why does Curtis place such emphasis on game theory, when the much more logical labels of ‘neoliberal ideology’ or/and ‘corporate values’ are available?

In this regard, ideology, propaganda and hegemony are all by-passed as ways of explaining how the political elite function. Obvious reference to the relationships politicians have with corporate elites and multilateral agencies are overlooked in favour of how they, supposedly, embrace this numbers-based politics.

Oddly, Major’s term in office, Curtis would have us believe, marked some kind of a significant juncture in how politicians began asserting the need for “performance targets” based on game theory. This, we are told, was also the model upon which New Labour would pursue its own regime of targets and performance-related reform of the public services.

However, before we get to the Blairite variant, Curtis seeks to illustrate how the Clinton administration, from 1992, was forced to succumb to the same logic of “market democracy”. Here, Curtis notes how “Clinton [had] promised to use the power of the state” to usher in a whole new raft of public initiatives – most notably, on healthcare. However, a crucial visit from Alan Greenspan, of the Federal Reserve, and Robin Rubin, head of Goldman Sachs bank, had persuaded Clinton that if he borrowed to fund this agenda, he would face an economic and political crisis. Indeed, they warned, Clinton would actually have to slash public spending – and dispense with his coveted healthcare reforms. They told him that there was a better way; that he should, as Curtis puts it, let the market “have its way”. Clinton “obeyed” and went on to oversee an economic boom.

Curtis discusses this apparent ‘turn of policy’ and obedience to the market with a curious sense of gravitas, as though Clinton’s speech announcing the end of “the vision of liberal politics” was more than what it obviously was: clever political rhetoric. Indeed, Curtis actually seems to entertain the fantasy that Clinton had contemplated a serious alternative to the unbridled free-market.

This is a revealing thread in Curtis’s argument, for it indicates just how ‘radical’ he thought Clinton to be, in terms of policy alternatives. Of course, it’s useful to record how the powerful private health lobby had used its political clout to stop even these moderate measures. Yet, in reality, Clinton never seriously intended any kind of neo-Keynesian, New Deal interventionism in healthcare or elsewhere.

Curtis deploys Robert Reich here, Clinton’s special policy advisor, who notes how this extreme model of laissez-faire was not what Adam Smith had understood by free-market capitalism. Yet, none of Curtis’s or Reich’s assertions about the distorted nature of market ideas are particularly original or revealing. Even if Clinton had been pursuing a more social democratic model, none of this would have represented a serious threat to the sovereignty of Wall Street. All that Curtis has been able to demonstrate is that the neoliberal elite got their way in managing to scupper even these token election-seeking pledges.

So, why can’t Curtis just state all this it in terms of the ‘Washington consensus’ - which had been in the ascendancy from the mid-1970s? Rather, Curtis has tried to convince us that the root of this ‘showdown’ lay in the ongoing rationale of game theory.

We then have another interesting, if tenuous, digression into genetics, as Curtis considers a possible link between the elemental logic of game theory and whether humans are actually programmed, biologically, to behave as self-interested entities engaged in rational, mathematical actions. A major anthropological study is noted here, claiming, among other things, that certain forms of tribal infighting could be explained as the greater tendency of genetically-connected individuals to intervene on behalf of each other.

Amid all this spurious argument, Curtis’s does appear to be posing an axiomatic question: are people encoded with a machine-like propensity towards self-interested behaviour? Curtis’s own apparent effort to answer this question seems vague and unsatisfactory. But, somehow, from the sense of disconnect emerges what sounds like a more usefully working proposition: the idea that people are being conditioned and tweaked to conform to the standards and categories of the system, rather than being allowed to see the fault for their disorder in the system itself.

And, for Curtis, this has come to mean a whole new numbers game for the shrinks and marketeers: “the technicians of this idea would be the psychiatrists and drug companies.”

While Curtis is still, erroneously, asserting the root influence of game theory, he is, finally, on to something substantive. He discusses how, in the 1990s, an “epidemic” of mental disorders seemed to have swept across the West, with some 50 per cent of Americans having been reported as suffering variant forms of mental disorder.

At the same time, the proliferation of SSRIs was promising “liberation from anxiety” on a mass scale. But, in the process of the Prozac revolution, people were now, more than ever, being “defined by the checklist”. It may, of course, be fairly objected that many lives have been relatively improved by such drugs. But the point is worth considering: to what extent has market society helped drive individual neurosis?

Thus, we see how corporate life creates mental discontent in the striving for commodities, then makes further profit in drugs and treatment which claim to allay those anxieties. Yet, while Curtis is on much better ground here, isn’t all this actually explainable as the product of neoliberalism and the rapacious demands of the market itself, rather than the application of game theory?

It’s also worth noting here how Michel Foucault had already charted the methods of social exclusion and confinement of individuals from the Middle Ages in Madness and Civilization. Foucault showed how the emergence of bourgeois society had brought about the need to root out inefficiency and ‘medicalise’ those not able to participate in the market. In this new world of reason, which came to be equated with capitalist rationality, many of those previously regarded as sane were now cast as insane:

[Raoul Djukanovic]

Thanks, John.

Those who missed the programme can see it here (in miniaturised realplayer form): http://www.indybay.org/uploads/2007/03/18/2_the_lonely_robot.rm

You write:

[Peter Fainton]

Thanks John. Here's a link to an earlier piece by Curtis which explores some of the same ground from a different perspective:

http://throwawayyourtv.com/2006/08/century-of-self.html

[johnwhilley]

Thanks, Raoul, Peter and the Editors for those thoughtful observations.

Peter:

You wrote to me in an exchange a while back about Edward Bernays, the ‘father of market PR’, and insightful reading it was too.

In Century of the Self, Curtis has taken Freud and the related ideas of his nephew Bernays and worked them into a treatise on how politicians and big business sought to control what they saw as human irrationality through appeals to the self. Curtis is closely concerned here with how Freud’s works on psychoanalysis were to be used to stem “dangerous instinctual drives” within human beings. We see, additionally, how Bernays was first utilised by Woodrow Wilson to help sell the post-war settlement, thus presaging how his ideas and expertise would be deployed for the purposes of “mass persuasion… in peacetime”.

Intrinsic to Curtis’s film is the belief of elites that human beings could not be trusted; that the “unconscious feelings of the masses” had to be channelled and controlled. Mass consumption would, thus, help allay these ‘irrational drives’, through what Bernays called the “engineering of consent”. This “happiness machine”, as Curtis puts it, was, thus, propagated as a means of keeping the society in a state of docile passivity.

While Curtis is documenting key historical events and influences here, there seems to be a tension between his attempts to explore whether such “hidden irrational forces” did, indeed, lie beneath the surface, threatening to summon up the “frenzied mob”, and how the claim of such was used as a propaganda tool. Thus, the preview to Part 2 claims how:

[informationist]

John: re curtis and appraisals of "free" market/capitalism/corporatism etc have you seen the Mayfair Set?

its about politicians losing power to the market. (a thread he uses a bit in this new series)

its not hugely insightful or anything but he does look at some of things you have been talking about.

one reason i think it lacks insight and perhaps a reason why curtis stays away from economics is because like jeremy paxman i dont think he understands it very well. (jeremy paxman admited on newsnight recently that he didnt understand hedge funds and derivatives and seemed a bit confused by the stock market in general, surprise? not really when you think about it)

also perhaps there are some disincentives within the corporate media system regards covering economics from a humanist (?) perspective and joining the dots etc etc,(no doubt that there are disincentives infact) but i do think in curtis` case its mainly because its not his area. (i could be wrong, he might be a professor of economics for all i know, but i doubt it)

also on an unrelated point, he needs a new aesthetic for his films. (he has been using the same sort of music, editing, footage, titles etc for years and years and its becoming self-parody, also in this last series he is recovering alot of ideas he has looked at before, politicans losing power to the stock market...mayfair set....people`s well being and social connectedness.....century of the self.....game theory, nutty egg heads making calculations about nuclear war...pandoras box)

in terms of creative ability curtis is a medium sized fish in a very small pond when you look at the mainstream tv competition, perhaps too much is expected of him?

[Raoul Djukanovic]

[johnwhilley]

Thanks, informationist, for those references.

You could have a point here regarding Curtis’s seeming circumvention of key political-economic forces – though I’m sure someone of Curtis’s intellect can’t fail to see or grasp what kind of agenda institutions like the IMF and G8 serve in maintaining the interests and ideas of corporate capitalism.

Curtis may be seeking to delineate certain novel features of this in his film, and has, in the process, said some critical things about the political class. But, as with his passing reference to Brown’s interest-rate policy-handover to the Bank of England, it’s a cursory nod to what I see as the main political-corporate dynamics underlying such relationships.

Raoul, there are three basic things to remember here:

Basic thing 1: it is highly unlikely that the BBC would ever permit the prime-time or major-slot airing of a film like The Corporation.

Basic thing 2: even if it did, it would never get the celebrated coverage, trailers and adulation bestowed on The Trap and The Power of Nightmares.

Basic thing 3: The Corporation was aired on commercial TV, as are some of Pilger’s films, but such output is (a) very occasional and (b) usually given marginal, late-night slots.

And here’s a related thought for anyone following these exchanges:

The Trap and Power of Nightmares are a kind of reference-point narrative for the thinking liberal; a sort of agenda-setting liberal statement around which the public and critics alike are invited to form an opinion.

An example of such agenda-setting is Samuel Huntington’s Clash of Civilizations text, and the ways in which this became the academic and ‘thinking media’ benchmark for addressing the immediate post-cold war world. The accompaniment to this was Francis Fukuyama’s End of History thesis, a work which was, again, invested with a central celebratory academic and media significance. Both have since been exposed as logically and ideologically flawed. Yet, they helped concentrate liberal thought and arguments at a key juncture, thus serving an important propagandist purpose.

In similar ways, Curtis’s latest works offers an agenda-setting - or zeitgeist kind of - liberal discourse. They provide a current, and visually absorbing, take on the crisis of freedom and democracy – one which reflects many of the sentiments and concerns of the liberal classes and people at large as they witness daily political spin, hear the constant threat of terrorism and sense the alienation of consumer-driven life.

The Trap and Power of Nightmares offer something of a thinking explanation of all this. Yet, with Curtis's dark warnings of mendacious forces also comes a kind of palliative reassurance: you are being manipulated and lied to, but it is dysfunctional and cynical elements within the system which are at fault, not the basic ideal of liberal capitalist democracy itself.

This is a key nuance of how liberal propaganda works. It is not so much that one has to agree or disagree with such output. It is, rather, that one assumes in it the status of ‘landmark’ narrative; a focal point context around which the crises of freedom and democracy might be understood and discussed. As with the kind of agenda-setting output we see in Newsnight over Iraq, this is a subtle way of setting the permissible lines of critical comment; a means of defining how the issues are framed and held up for public and peer inspection.

John

[informationist]

no doubt curtis has the intellect, but does he have the will? (same goes for jeremy paxman, who is oxbridge presumably, i wonder if curtis is too?, he seems the sort, im not making a class statement here, im making a point about classical education, one being more comfortable reading cicero or homer than reading warren buffet or milton friedman et al)

perhaps Curtis should make a series about the stock market as it exists today worldwide (rather than just post war britain), and look at people like Jim Cramer...(although he would have to put some time into working out what Cramer is talking about here, decode the market speak, its shocking how much Cramer comes across as an untrustworthy second hand car salesman in the interview below, would you like him to manage your pension fund?, is he typical of wall street traders?, i would interested to know)

http://www.youtube.com/watch?v=708wDFX28lc

I agree that "market forces" and corporate structure etc are very important topics but they are also complicated and difficult ones with alot of history and context to include*.

*(its complicated and difficult if you want to try and give a rounded view, its not complicated or difficult to throw together a specious polemic, ala` durkin, climate change)

The more i think on it the more i think maybe Curtis is not the man for the job given his lack of experience in the area?. (also a documentary which attempts to demystify economics is never going to be the next "super-size me" in audience/ratings terms)

it does seem as if there is an economics blind spot in terms of mainstream TV documentary, but its not all due to propaganda filters imo. (as i said, its a complicated issue, and not popular amongst the public, which is ironic of course as its the stuff they/we really need to know and to take a stand on, perhaps someone should make a docu about why economics is important and how insane it is that the public know so little about it, as a primer for a documentary on the subject itself)

EDIT: Jim Cramer seems to have kicked up a bit of stink in that interview, here is a Reuters story about it.. http://today.reuters.com/news/articleinvesting.aspx?view=CN&symbol=TSCM.O&storyID=2007-03-20T225520Z_01_N20362926_RTRIDST_0_CRAMER-INTERVIEW.XML&pageNumber=0&WTModLoc=InvArt-C1-ArticlePage3&sz=13

[johnwhilley]

Thanks, Info.

It’s not so much, I suspect, that Curtis evades or has a specialist problem with economics, per se, rather that he doesn’t appear willing to look closely enough at how the assaults on freedom and democracy are the logical outcome of political-corporate relations, sui generis.

And, on that note of structural neglect, here’s another key point to consider. While many reviewers have been absorbed and seduced by The Trap and other of Curtis’s films, it’s worth noting one of the most basic omissions in these ‘forensic’ explorations of power, politics and propaganda: discussion of the media itself.

It’s a remarkable feat for a writer supposedly concerned with mass indoctrination and the propagation of ideas. Again, Curtis might retort that he is not making a film about the media. Yet, how is it possible to construct a serious and searching thesis on the crisis of freedoms without asking crucial questions about the media’s role in the process of political and cultural indoctrination?

This, of course, is part of the same liberal mindset that sees the BBC as a sentinel of truth and the Guardian as a radical voice of dissent. But it’s also, paradoxically, because Curtis is preoccupied with the ‘power of politics’ – or his version of such - rather than the politics of power. He looks, almost one-dimensionally, at how leaders and governments come to manipulate the public through arcane ideas like game theory, yet assumes the corporate media – an internal part of the capitalist system - to be somehow subsidiary to that process of control; as if the media holds a detached or neutral place in this game of political, economic and cultural power.

As noted by the Editors, it’s the acid test of the ‘critical’ writer: how prepared are you to question the institution which employs and protects you? To what extent are you likely to see or seek causal links in the medium that you serve? It’s the same kind of question that’s been put to other critical writers like Monbiot. Yes, we see the often valuable critiques they offer on key issues. But why this reluctance of enquiry when it comes to the media’s own role in filtering political deceit and protecting the elite – including themselves?

Again, one can’t help but conclude that such deference to the real centres of power is why Curtis enjoys the journalistic latitude he gets at the BBC.

Best wishes,
John

[Danny]

I've been a huge fan of RD Laing since I was a teenager. And when I was a teenager in the 80's I was a design engineer learning how to code from very smart people in an electronics company. Instead of having chess competitions or whatever we'd run our algorithms up against our competitors.
The 'which algorithm could eat up the most memory' (much) later became the staple of early virus writers but there were other games. I'm no mathematician but I am fond of games so I learned a lot.

I mean no disrespect to anyon on this thread but you really have to code to understand game-theory. Otherwise it's like a football commentator who has never played football, they might be entertaining to an audience but they won't convince anyone who has played.

That's my impression of every 'game-theory' commentator I've read, Adam Curtis' new series included, to a lesser degree Philip Ball though I would recommend his book 'Critical Mass' to anyone interested in a +grounding+ in the subject.

I don't even think Robert Axelrod explains the theory well but I'd be a damn fool if I attempted to criticise him without 20 years preparation and lots of Omega fish oil capsules. My main point is there is no way in hell the politicians who claim this theory as justification for their policy understand a jot of it, they are told what to believe by charlatans and they use it to their own best personal advantage like the simplest of algorithms.

Tat for tits you could call it.

[Danny]

THE GRAND TOURNAMENT
You are on a train, and as you take your seat you find a wallet. It is fat with money. What do you do? At face value this is a simple case of binary choice,. You can either try to return the wallet to its owner, for example by looking inside for an address or handing it to a member of staff, or you can place it quietly in your pocket.
The behavioural models we have looked at in earlier chapters have tended to assume that agents in a multi-agent scenario respond to the actions of their neighbours in what we might regard as a knee-jerk or at least a somewhat mechanical way. That is to say, stimulus A induces response B, either invariably or with a certain probability. But in a situation like this, choices aren't made so simply. True, some people are invariably honest and some invariably dishonest. But the terrain in between is not negotiated by the random throw of a die. What flashes through our mind, perhaps involuntarily, is the thought, 'Who would know?' And then, maybe, 'What if I'd lost my wallet, how would I feel?' In such cases, we weigh up our options according to some moral code - but that code is bedevilled by temptation.
Temptation is arguably the fundamental problem for human societies. It sometimes pays not to be the good, kind, considerate citizen but to rebel, to cheat, to fight, to do the dirty. If my neighbours are all meek and law-abiding, what is to stop me from appropriating some of their land, or goods, or cattle? A Hobbesian individual in a Hobbesian world is as miserable as everyone else. But a Hobbesian in Eden can run riot, amass a fortune, gorge himself, and fear no reprisal (unless he believes in God). Temptation is a part of the human condition, and that is the problem for all utopias: not everyone is nice, because sometimes crime pays.
It is not obvious how to devise a 'particle' that can be led into temptation. But in the 1950S Merrill Flood and Melvin Dresher at the RAND Corporation in California did more or less just that. They developed a simple mathematical model which incorporated the element of temptation into an interaction between two agents. The model was presented as a kind of game. Flood and Dresher were exploring the theory of games devised by the mathematical physicist John von Neumann in the 19:WS. One of the most formidable math­ematicians of the twentieth century, von Neumann helped to establish the theoretical basis of the computer and made crucial contributions to the Manhattan Project during the Second World War. He cultivated something of a reputation as a playboy genius, which his passion for gambling and poker did much to enhance. But von Neumann didn't just want to play these games - he wanted to understand them.
For sheer complexity, a mathematician can do no better than to study the game of chess. There is a sense, however, in which poker is much more challenging, for it incorporates the psychological element ofblufIing. The question is not, as in chess, what the next best move a choice either to 'cooperate' or to 'defect'. The best outcome - the maximal payoff - for one agent comes ifhe defects (testifies) while the other cooperates (that is, refuses to testifY - the cooperation here is with the other prisoner, not the authorities). In that case, the other player is the sucker and gets the worst outcome. But if the agents play rationally, they get neither this optimal payoff nor the next best thing, which is the payoff from mutual cooperation. Instead, they get the meagre rewards of mutual defection, which are only a little better than the sucker's payoff.
To recast this dilemma in terms of individuals living in a society, we can regard cooperation as being law-abiding and defection as breaking the law for one's own gain at another's expense. The basic dilemma ­ that cooperation is good but defection can be even better - was recognized by Rousseau and Spinoza. In his Discourse on the Origin of Inequality, Rousseau imagined five men from pre-civilized times agreeing to cooperate on a stag hunt, on the understanding that each will get a fifth of the spoils. When a hare comes within reach of one of the men, he grabs it - but without his help, the stag escapes. The 'defector' has the immediate gratification of stewed hare, rather than sharing the difficulties and dangers of catching a stag - but his fellows have nothing.
At face value, the Prisoner's Dilemma seems to confirm Hobbes's pessimism: egoistic individuals guided by logic will always seek to exploit one another. What did it seem to say to the Cold Warriors who drew on the advice of the RAND Corporation? Get the first fist in, for your enemy will try to do the same. Build up your nuclear arsenal with all the resources you can muster, for your enemy is planning to defect, and so you must be ready to do so too. Indeed, you should even consider defecting first - making the first strike. If the other side 'cooperates' even to the extent of not launching a strike immediately, you can make them the sucker by doing so at once. You win; they lose.



* In these repeated games a 'score' is kept of how well each player did on each round. A player scores highest for defecting when the other cooperates; moderately for mutual cooperation; poorly for mutual defection; and worst of-all for cooperating when the other defects (the 'sucker's payoff'). One might imagine each player accumulating points or money, for example, rather than prisoners accumulating years of incarceration.
So in the repeated or iterated Prisoner's Dilemma game, the players have the chance to learn from their mistakes and to build up a relationship of mutual trust. Cooperation can evolve.
Is this how real people play the game? Psychologists have studied that question extensively in controlled tests, and found that co­operation does develop - but to a degree that varies widely, depending on the character of the players, the nature of the payoffs and the circumstances of the interaction. One can imagine, for example, that it is easier to defect anonymously than face to face. And let's not forget the element of temptation. If you think you are facing a nice player who will do their best to cooperate, you might be tempted to throw in an occasional defection, thereby boosting your own score at your oppo­nent's expense. If they are forgiving, you might get away with it if you don't try it too often. Sadly, in a cooperative world, defection pays. This then raises the question of what is the best way to play the iterated Prisoner's Dilemma. If you know nothing about your opponent, which strategy should you adopt?
In the late 1970s Robert Axelrod devised an experiment to try to answer this question. He asked professional game theorists to submit strategies for playing the iterated Prisoner's Dilemma, and then put each strategy to the test in a round-robin tournament conducted on a computer. Each strategy was played one-to-one against each of the others for many rounds; the winner was the strategy with the highest aggregate score. The fourteen entries came from psychologists, mathematicians, economists, sociologists and political scientists. Each strategy consisted of a set of rules for determining the choice of cooperation or defection. For example, one might simply choose always to cooperate. (This is obviously a bad choice, since it always comes off worst unless everyone else is also an unconditional cooperator. So no one chose this strategy.) Or one might cooperate on the whole but defect every fourth round. Many of the submitted strategies were more complex than this. But the tournament was won by the simplest of them all. It was submitted by Anatol Rapaport, who called it Tit For Tat. Its sole, rule was to begin by cooperating, and thenceforth to do whatever its opponent did in the previous round.
Playing against an unconditional cooperator, Tit For Tat (TFT) cooperates throughout the entire encounter. So both players do equally well. Against an unconditional defector, TFT gets the sucker's payoff in the first round (where it cooperates), but then it defects consistently, as though determined not to be taken advantage of again. Because of the first round, TFT comes away slightly worse off than the inveterate defector - but only just. Both of them, in any event, do much worse than if they'd cooperated. By mirroring its opponent, TFT can adapt to whatever the situation calls for. Against cooperators it is nice; against defectors it is tough. When faced with a mixture of cooperation and defection, it gives as good as it gets. So TFT reaps the benefits of cooperation where possible, but cannot be exploited .. Neithe1does it exploit: it never achieves a higher total payoff than its opponent. Some of the other strategies did well against those with a tendency to cooperate; others could hold their own against defectors. But by making the best of both situations, TFT came out on top in a diverse mixture of strategies. It was a modest, simple-minded victory.
Following the success of his first computer tournament, Axelrod decided to hold a second, with essentially the same rules as before. News had travelled, and this time there were sixty-two entries, from six countries. Some came again from professional scientists and academics, but others were from computer hobbyists, including a ten year old boy. All knew the outcome of the first round and so had a chance to consider the reasons for Tit For Tat's success. Anyone could submit any strategy, but only one person chose TFT: Rapaport. All the others decided that they could outdo TFT with something more sophisticated. They couldn't. TFT was again the winner.
Does this mean that Tit For Tat is the best way to play the iterated Prisoner's Dilemma? Not exactly. There is in fact no best way to play, for it depends who you're playing. It is very easy to illustrate that this is so. If you are playing against a colony of unconditional cooperators, you will do best to be an unconditional defector - that strategy will fare better than TFT, which would behave like one of the cooperators (except when it plays you). But the message of Axelrod's tournaments seemed to be that if you don't know who you're up against, TFT is the best default strategy.
What, then, makes Tit For Tat so special? For one thing, it is flexible: able to cooperate but not open to exploitation. Cooperation from the other player will immediately elicit cooperation from TFT in the next round. But defection is met unhesitatingly with defection. This sends out a clear message: TFT will do as it is done by. It is a strategy from the Old Testament, not the New: an eye for an eye, not turning the other cheek. This clarity of response is in itself a factor in TFT's favour. In the second tournament, one entry was a strategy designed to try to figure out the rules the other player is using, in order to find some way of exploiting them. This often happens in real life: one person will check out the other, weighing up how much he or she can get away with. If you know beyond doubt that you can never defect without being treated the same way, you have a good incentive to cooperate. If you have reasons to doubt that the retaliation will be relentless, you might be tempted to try your luck. TFT, in contrast, guilelessly encourages cooperation and discourages defection.
But there is another telling aspect of Tit For Tat that contributed to its success: it is never the first to defect. All strategies can be broadly divided into two camps by this criterion: will they defect first or not?
TFT takes the other player's actions into account too, but only to the extent of mirroring them - it does not seek any deeper understanding of the opponent's strategy. Those that will not are generally called 'nice' strategies. (There is no consensus on what to call the others; but 'nasty' will do.) Axelrod found that nice strategies do consistently better than nasty ones. Indeed, in the first tournament the ranking produced a clear distinction: the eight top-scoring strategies were all nice, and the others, separated from the nice ones by a substantial gap in points scored, were all nasty. So the Prisoner's Dilemma starts to look less grim when it is iterated: niceness and cooperation fare better than nastiness and exploitation. Even individual selfishness need be no barrier to fair play. But being cooperative does not in itself guarantee success; Tit For Tat plays a much tougher game than that. Axelrod has identified four characteristics of a successful strategy:
Don't be the first to defect (be nice). Always reciprocate.
Don't be too clever.
Don't be envious.
What is envy, in this context? It means not trying to do better than the other players, but simply doing as well as you can for yourself. The Prisoner's Dilemma is not what is called a zero-sum game: someone else's gain does not have to come at the expense of your loss. If you both cooperate, you can both do well (even if not as well as you would if there is an option to exploit the other players). Axelrod confesses that real players seem to find it hard to relinquish competitiveness and envy. In his tests of the iterated Prisoner's Dilemma with student volunteers, he finds that they tend to measure their performance against that of others, in which case any advantage gained can tempt the others into defecting to try to redress the balance. This can trigger recriminatory outbursts of defection.
The live-and-let-live behaviour in the trenches of the First World War can be seen as an example of cooperation arising from a Tit For Tat strategy. Belton Cobb's remark makes it clear that there was no compunction about retaliating in a lethal manner to hostilities from either side - mutual ceasefire did not depend on good feelings between enemies. (Yet, significantly, it seemed that self-interested cooperation allowed these feelings gradually to develop.) And on the whole both sides followed a 'nice' strategy, declining to fire first.
In case we should still wonder whether human sentiments, rather than the mathematical exigencies of game theory, led to this reciprocal cooperation, we might bear in mind that Tit For Tat strategies are also found in the natural world. There is evidence that vampire bats, stickleback fish, monkeys and even viruses behave according to the rules of TFT. No one can reasonably attribute altruism to viruses: their behaviour is purely the result of genetic selection. That is to say, those organisms with a genetic predisposition to show TFT -like behaviour gain an evolutionary advantage, and so natural selection works to their benefit - ensuring that this genetic trait becomes more widespread.
This implies that we too may be genetically hard-wired to co­operate, perhaps in a TFT -like manner. Indeed, it would be astonishing and puzzling if we were not. Edward O. Wilson argues that as civilization evolved, such modes of human behaviour will have become converted from instinctive impulses to social norms, then to legal imperatives, and ultimately to moral principles.
One could even argue that the case for a genetic embodiment of the lessons of the Prisoner's Dilemma is supported by the readiness with which we greet their optimistic aspects. We would be sorely dismayed if game theory were not capable of producing cooperative behaviour; in fact we might then be tempted to dismiss it as nonsense or as mendacious. We are, it seems, predisposed to look favourably upon altruism and to frown upon apparently selfish behaviour. That this might be a learnt response does not evade the issue; we learn it because those are the cultural norms of our society - and where did they come from?
Here, then, is a possible resolution to the divergent views of human nature evinced by Hobbes and Locke, which led them. to such differing conclusions about systems of government. People do not, in the absence of a higher authority, necessarily seek to exploit one another in the way Hobbes envisaged. But neither do they desist from it because of a 'reason' instilled in them by God. The 'reason' can come from nature alone: from the inexorable mathematics of interaction coupled to the winnowing effect of natural selection. The Prisoner's Dilemma is actually implicit in Hobbes's analysis, since he acknowledges the miseries of mutual defection and argues that men are better off cooperating if this can somehow be arranged:
That a man be willing, when others are so too, as farre-forth, as for Peace, and defence of himselfe he shall think it necessary, to lay down this right to all things; and be contented with so much liberty against other men, as he would allow other men against himselfe. 22
This connection between Leviathan's State of Nature and the Prisoner's Dilemma was pointed out in 1969 by political scientist David Gauthier. Without a contract to cooperate, says Hobbes, a man would 'expose himselfe to Prey,.23 But such a contract is liable to dissolve unless there is some authority that can enforce it, since men's appetites will make them liable to defect the moment they see advantage in doing so. Thus, says Gauthier, Hobbes's omnipotent sovereign provides an escape from the Prisoner's Dilemma that men face in the State of Nature, since in a sovereignty defection no longer brings potential rewards but only certain punishment. Even if, as has been argued, it is somewhat misleading to cast Hobbes's scheme in game-theoretic terms when he had no interest in deducing the psychology of people faced with such behavioural dilemmas, it seems clear that Hobbes recognized the underlying problem that arises when antisocial actions offer potential rewards.
But game theory suggests that Hobbes's rather extreme solution - a capitulation of all individual powers and rights beyond self-preservation - may not be necessary. His error, if we may call it that, was to treat people as blind animals who cannot learn from experience - an 'experience' that can be handed down from previous generations as a genetic predisposition towards cooperation.
By the same token, we might expect to find other implications of game theory hard-wired into human experience. The tendency to form tribal groups increases the likelihood of repeated interactions with other group members and so enables cooperation to develop. Robert Axelrod endorses the notion of prolonged interaction - he calls it 'enlarging the shadow of the future' - as a way of promoting and nurturing cooperative behaviour. The flip side of this principle is distrust of strangers, since it takes time to establish the mutual trust on which cooperation depends. But this apparent biological predisposition for xenophobia should be moderated by the realization that 'nice' strategies do best: even on the first encounter it is preferable to cooperate.
So by arranging to make future exchanges more probable, we can guide two parties towards the benefits of mutual trust. This could entail making a relationship more durable - it was the long-term confrontation of forces at the Western Front that made the tacit ceasefires possible. Or we might increase the rate of interactions: in small communities, the same people deal with one another day after day both socially and economically, and so trust is easier to establish than in large cities where interactions are more occasional and impersonal. Increased demographic mobility reduces the durability of interactions and so reduces the incentive for cooperation: transient neighbourhoods are rarely cohesive and 'neighbourly'.
It is clearly not news to businesses that their interests are served by developing good long-term relationships with clients. But the way in which such relations can break down gives us some reason to suspect that the reciprocity does indeed stem from a Prisoner's Dilemma style of exchange. A study in 1963 indicated that one of the commonest reasons for court action between businesses is the complaint of wrongful termination of a franchise. It is only when relations between businesses are about to cease - when the iterations of the 'game' are about to end that one or both players decides it is worth their while to start a legal battle and risk bitter recrimination rather than to .find a 'peaceful' way to resolve differences. In psychological tests, people who play the Prisoner's Dilemma will often sacrifice mutually established cooperation for a few rounds of defection when they know the game is about to finish. In the same way, companies about to go bust are at greatest risk from non-paying clients, and are themselves more likely to default on debts.
The durability of interactions has implications for modes of government. Karl Popper considers that the most important attribute of a true democracy is not what it does, but that it 'should keep open the possibility of getting rid of the government without bloodshed, if it should fail to respect its rights and duties, but also if we consider its policy to be bad or wrong.' For as Pericles put it in democratic Athens, 'Even if only a few of us are capable of devising a policy or putting it into practice, all of us are capable of judging it. ' In a democracy, unpopular governments can be removed by elections ­ which seems incontestably proper. But the termination of any government carries some risks, for a departing government no longer has anything to lose from acting with blatant self-interest. Bill Clinton's outgoing US administration of 2000 demonstrated this with a display of political backhanding that the president would never have risked in mid-term.
To some extent this situation can be remedied by the existence of political parties, which carry long-term accountability for the short-termism of its members. There is no doubt that in 2002 the British Conservative Party was still paying the price for its deeply unpopular policies, its arrogance and its corruption while in office five years before, even though the perpetrators had largely disappeared from the political scene (several of them ignominiously). The US Republican Party paid the same long-term price for the Watergate affair in the 1970s. Thus a political system with a durable party structure might be expected to be less susceptible to corruption than one with more ephemeral kinds of political organization. Karl Popper called the party system 'horrible', since it makes parliamentarians tend primarily to serve their party rather than their constituents. 'I think', he wrote, 'that we should, if possible, go back to a state where MPs say: lam your representative, I belong to no party.027 But this could be a recipe for eliminating the accountability that political systems need if they are not to be plagued by abuses of power.
FOR YOUR OWN GOOD
It would be a foolish evolutionary psychologist who tried to argue that Tit For Tat is all there is to altruism. For one thing, the self-sacrifice we are capable of showing towards our own kin has biological roots which owe nothing to game theory - it seems to be an aspect of the 'selfish gene' idea, benefiting individuals who share a close genetic similarity to ourselves.
Moreover, behavioural economist Ernst Fehr and his colleague Simon Gachter in Switzerland have conducted experiments with human subjects which suggest that cooperation can arise in groups even when the individuals do not encounter one another repeatedly. Fehr and Gachter divided 240 students into groups offour, gave each of them an equal sum of money, and invited them to invest it (or not) in a group project. The project produced returns in proportion to the degree of investment. If all four members invested all their money, they all got a return that exceeded the investment. So it was in the group's interest for everyone to invest everything. But because each member received less than one monetary unit for investing each unit of his or her own, it was in each individual's interest not to invest but to freeload, relying on the contributions of the others.
This is analogous to a Prisoner's Dilemma insofar as it presents players with benefits for mutual cooperation but temptations towards individual defection. But Fehr and Gachter mixed up the groups after every round of investment and return, giving them no opportunity to establish mutual trust. They found that cooperation could never-the-less flourish if the rules included some provision for punishing defectors (those who invested little). Players would mete out such punishments even when they were charged a fee for doing so. Without the threat of punishment, cooperation was low; when this threat was introduced, cooperation increased sharply. The researchers call this 'altruistic punishment', since it is likely to be of no immediate and direct benefit to the punisher - even if punishment reforms the defector, the punisher is unlikely to encounter him or her again. It is nevertheless altruistic because it may benefit those others who will be grouped in later rounds with the player who defected.
This sort of behaviour suggests that the possibility of Tit-For-Tat-style retaliation or punishment may have a part to play in enforcing cooperation in society, even when encounters are not repeated. Fehr and Gachter reported that players seemed moved by a sense of injustice: they were simply angry at defectors, and acted on that anger irrespective of whether they stood to gain from it. Moreover, players reported that punishment was a deterrent to their own inclinations to defect. It is worth noting, however, that this is precisely the kind of behaviour one might expect from players predisposed towards a TFT strategy. By allowing punishment to be meted out after the game has been played, the researchers were in effect enacting a kind of tworound game, in which first-round defectors are themselves treated to defection from other players in the second round. The results could imply that we are all imbued with a desire for justice which we will exercise, if necessary, at our own expense.
In all these games, mutual cooperation pays best in the long term. If indeed cooperators are more 'successful', we would expect a predisposition towards cooperation to become an irreducible element of our neural circuitry. Sadly, however, this does not mean that Kropotkin was right: that no government is needed because people can be trusted to organize themselves. History shows us what people are capable of, and it does not look much like Eden. Human nature is diverse; it is also mutable, for better or worse. And it is influenced not just by one-to-one interactions, but by the multitudinous society in which each of us is embedded. To deal with that, game theory needs to get more sophisticated.
18 Pavlov's Victory Is reciprocity good for us?
The fact is that a man who wants to act virtuously in every way necessarily comes to grief among so many who are not virtuous.
Niccolo Machiavelli (1513)
Can one wait with calm confidence for the day when the despotic states that have made wars in the past have been turned, by the . social and economic forces of history, into peace-loving democ­racies? Are the forces of evolution moving fast enough? Are they even moving in the right direction?
Kenneth Waltz (1954)
In the interests of peace I am opposed to the so-called peace movement.
Karl Popper (1988)

If the Middle East today bears witness to harsh words and harsh deeds, it has done so before:
You must purge the evil from among you. The rest of the people will hear of this and be afraid, and never again will such an evil thing be done among you. Show no pity: life for life, eye for eye, tooth for tooth, hand for hand, foot for foot.!
The uncompromising reciprocity of a Tit For Tat rule may have worked for the children of Moses (although even for them it was not the lawful response to all insults). But can this really form the baseline of a civilized society? Game theory seems best suited to investigating the State of Nature that Hobbes feared as barbaric and Locke idealized as beneficent; it delivers the hopeful message that goodness can arise out of barbarism. But only at the cost, it seems, of uncompromising retaliation to all aggression. The whole point of government, says Locke, is to eliminate the necessity for every man to be his own judge and enforcer. But does that oblige a government to enforce cooperation among its subjects and its neighbours alike with the same policy of swift and unquestioning retribution? Where in this social calculus might we find room for negotiation, conciliation, mediation, even forgiveness?
To understand the implications of game theory at more than a superficial level, we need to subject Tit For Tat to a rigorous examination. That is my objective in this chapter. We shall circumscribe this strategy's advantages and probe its weaknesses. We shall release it into a community and watch the consequences. We shall ask - as we always must in social physics - not just what the models tell us but what we consider desirable, and whether the two can be reconciled. And so we shall come back to the enduring question: what choices do we have?


ACCIDENTS HAPPEN
After Axelrod's second computer tournament, Tit for Tat looked invulnerable. But it isn't. In the real world it has a fatal flaw: communications are imperfect. Mistakes are made; intentions are misunderstood. In 1983 the Soviet Union shot down a South Korean civilian aeroplane which had mistakenly strayed into Soviet air space, in the belief that it was a military craft. All 269 passengers, including several Americans, were killed. A Tit For Tat policy, rigidly applied, would dictate that this error could be avenged only with Russian blood. Fortunately it was not, although the incident did heighten Cold War tensions. The NATO bombing of the Chinese embassy in Belgrade during the attack on Serbian forces in Igg8 was at face value another apparent 'defection' resulting from a mistake. (There is still debate about whether it was truly unintentional.)
The hair-trigger status of American and Soviet nuclear arsenals during the Cold War highlighted the awful risks of a retaliatory policy in the face of potential mistakes. Taken to its extreme, this creates the scenario gloriously and chillingly lampooned in Stanley Kubrick's film Dr Strangelove, in which a rogue US army general launches a preemptive strike on the Soviet Union. All but one of the BS2s is recalled in time; but the one that cannot be contacted releases its warheads. This triggers global nuclear war even though the Soviets know the bombs were dropped 'by mistake', because they have automated their missile system with the Doomsday Machine, which retaliates to any nuclear attack without the option of human intervention. They believed that the absolute certainty of retaliation would enforce cooperation, but the system did not allow for errors.
The problem with mistakes, as far as Tit For Tat is concerned, is not simply that a lone, erroneous defection provokes the same in return. Tit For Tat's simplicity means that, if this happens between two players who are using this same strategy, they get locked into a cycle of mutual recrimination. One defects by mistake; in the next round it returns to cooperation (since that is what its opponent did in the last round), but the other player returns the defection. This causes the first player to defect in the round after that, and so on: the mistake echoes back and forth for the rest of the match, so that mutual cooperation is never restored. (In the Strangelove scenario, of course, a single round of defection from each side is enough to end the game once and for all.)
This sort of behaviour arises in many cultures and societies. Axelrod points to the example of family feuds in Albania and the Middle East, which can continue with mutual reprisals for many generations - even after the original incident has been long forgotten. Agonizing vicious cycles of mutual slaughter have plagued the Protestant and Catholic communities in Northern Ireland for decades, and currently seem to be destroying all hopes of a peaceful settlement between Israel and Palestine. Clearly, Tit For Tat does not guarantee a harmonious world.
Nor, amid the mess and confusion of reality, is it always the best strategy. This became apparent when Axelrod's tournament was repeated, this time allowing for the possibility that the players may make errors. The players would occasionally choose their response at random rather than according to the rules of their strategy. For an error rate (a 'noise' level) of 10 per cent - one in ten random choices- TFT is no longer the winner. In fact, TFT then fares even worse when playing against other TFT players than when playing a mixed bag of strategies, since occasional errors create ample scope for fruitless cycles of reprisal.
In such a situation, TFT needs modification if it is to score highly.
One alternative, called Generous Tit For Tat (GTFT), lets a certain fraction of defections go unpunished. Another, Contrite Tit For Tat (CTFT), declines to retaliate to a defection that follows a defection of its own - it 'accepts' that it got what it deserved. GTFT outperforms all the other entrants in Axe1rod's second tournament when there is 1 per cent 'noise'; CTFT comes sixth. For higher noise levels, CTFT outstrips GTFT. Tit For Two Tats (TFTT) is a strategy that retaliates only after suffering two consecutive defections: it waits to see whether a defection implies that the other player's intentions really are bad, rather than being simply a mistake (that is, noise). TFTT was devised by the evolutionary biologist John Maynard Smith, and it came tWenty-fourth in Axelrod's second tournament. Maynard Smith did not enter it in the first tournament; ifhe had, it would have won, because that first mixture of strategies contained some which impaired TFT's performance by getting locked into mutual retaliation (even without errors). This reinforces the point that there is no best way to play the game.
Another strategy which copes well with noise is less benevoknt.
Pavlov is a strategy based on pure opportunism, and was named in 1988 (although invented earlier) by David Kraines of Duke University and Vivian Kraines of Meredith College, both in North Carolina. Its philosophy can be summarized as 'win-stay, lose-shift'. Like TFT, it bases its choice of action on what happened in the previous round. If it did well, it makes the same choice again; if it did poorly, it switches. 'Well' here means either the reward for mutual cooperation or the best payoff of all, that for unilateral defection. In short, Pavlov sustains behaviour that brings rewards but changes behaviour that brings punishment. This recalls the simple, conditioned responses of Russian physiologist Ivan Petrovich Pavlov's dogs.
With a tough customer like Tit For Tat, Pavlov is quite happy to cooperate. It doesn't fare well with an incorrigible defector - it will try to cooperate every other round. But Pavlov will mercilessly exploit a habitual cooperator once it realizes it can get away with it, whereas TFT would nobly cooperate. Pavlov performs poorly against the con­testants ofAxe1rod's original tournament - in 1965 Anatol Rapaport gave this strategy the dismissive label of'simpleton'. And it doesn't do a great deal better even in the presence of noise. But it has the virtue of being able to recover quickly from an isolated error, and if the circumstances are right it can come into its own, as we shall see.
DARWIN'S ALGORITHMS
That history is a sound guide to policy is a cliche, although Friedrich Hegel doubted that nations and governments were so guided. Yet people, businesses, institutions and even countries surely do sometimes change their behaviour in the light of experience - just as the British and German troops whose job it was to eliminate one another on the Western Front ended up entering into unspoken truces of mutual self-preservation. Some lawbreakers can be reformed. It is this capacity for change that makes international relations both complex and worth arguing over. Some observers believed that Saddam Hussein's government in Iraq might not have been given inevitably to unconditional defection but could have been transformed into a more cooperative regime had it been engaged in dialogue rather than isolated with sanctions and then barraged with bombs.
One of the most interesting and important questions we can ask of the Prisoner's Dilemma is what kind of behaviour emerges when the players can evolve - when they are allowed to change their strategies. In reality, people apply all sorts of moral, ideological, habitual and whimsical criteria in deciding how to behave. But in the spirit of game theory, it is useful to begin by asking what players will do if they are purely pragmatic: that is, merely seeking to optimize their gains. It seems reasonable to assume that players will tend to adopt those strategies that are more successful. This can be simulated in Axelrod-style tournaments by including an evolutionary dimension. At the end of one full round-robin, for example, we might allow players to adopt a new strategy with a probability proportional to that strategy's overall score. In this way the more successful strategies will multiply, while ones that perform poorly will die out. It's not hard to see that this is a Darwinian 'survival of the fittest' scenario. It Inimics the way in which genetic mutations spread in a population: those carrying a mutation that conveys a reproductive advantage generate more offspring, en­hancing the prevalence of that 'adaptive' mutation.
Martin Nowak at the University of Oxford and Karl Sigmund at the University of Vienna conducted just such an experiment in game theory in 1992, with salutary results. They set up a diverse population of strategies, in all of which the choice of whether to cooperate or defect was determined by what the opponent did in the previous round. Some strategies were more inclined towards defection, others tended to cooperation. Nowak and Sigmund let them all compete against one another, and then altered the proportions of each of them in line with their relative successes. Naively, we might expect that this evolutionary model will be ruled by Tit For Tat, which appears generally to fare best in a mixed population. And this did at first seem to be the outcome. Early in the game the defectors had the upper hand; cooperative strategies died out, and the average payoff of the population fell towards the low payoff gained by mutual defection. But at some point a tiny band of TFT players began to grow rapidly until they dominated the population (Figure 18.1). This takeover was accompanied by an upsurge in cooperation and a rise in the average payoff.
The abruptness of this change is reminiscent of (although not strictly equivalent to) a phase transition. The rise ofTFT is a collective effect, a result of many mutual interactions between players. Mutual defection eventually becomes so self-defeating that a small group of

TFT players gains more from their mutual cooperation than defectors do from exploiting the TFT players' initial attempts to cooperate. At this point the tables are turned, and it pays to cooperate with the doughty TFT group. Their presence helps seed the spread of cooperation throughout the population. But Nowak and Sigmund found that the triumph of TFT is short-lived. Once it has established a culture of cooperation, TFT starts to suffer from its Achilles' heel: unforgivingness. These simulations contained an inherent amount of noise in the way the strategies worked, and this meant that TFT was gradually superseded by its more tolerant sibling, Generous Tit For Tat. In the end only GTFT remained.

The crucial role of TFT -type strategies in bringing about this change is emphasized if the evolutionary game is replayed with no TFT players to start with. Then the prognosis is a gloomy one: cooperators die out and we are left with a colony of selfish defectors that plough their Hobbesian furrow to eternity. rather than the aim, of an evolution towards cooperation. In other words, it is needed to establish cooperation in a diverse population, but once that has been achieved, 'softer' cooperative strategies will take over. In fact, since even GTFT will occasionally get caught up in unproductive recrimination caused by mistakes, a better strategy in a universally cooperative environment is unconditional cooperation: complete forgiveness. This sounds all very nice and inspirational. But the best strategy of all in a population of unconditional cooperators is unconditional defection: ruthless exploitation of the meek. Pitted like against like, cooperators do better than defectors, but cooperators are highly vulnerable to rogue defectors. A small band of defectors can wreak havoc in a cooperative culture. Tit For Tat can prevent this from happening, for it treats defectors severely while rewarding cooperation. It can be regarded as the police force of game theory, imposing cooperation with a firm hand. It models an ideal policing strategy in some respects, for (in the absence of noise) it only ever -and invariably - punishes defection, and never exploits cooperation. The implication seems to be that if we accept some level of defection as inevitable, we have to concede that a society needs at least some TFT players in order to maintain a general culture of cooperation.
Even that, however, may not guarantee a fair society. In 1993 Nowak and Sigmund discovered that TFT's implacable sense of justice does not always come out on top. In their earlier evolutionary games, players based their strategies for their next move on their opponent's previous move. But Pavlov, the opportunistic win-stay lose-shift strategy, does more than this: it takes into account the player's own last move too. When the two researchers pitched their earlier strategies against Pavlov, they found that Pavlov's opportunism triumphs. Pavlov does poorly against defectors and lacks TFT's ability to 'invade' a defecting population and spread cooperation. But in a (slighdy noisy) community imbued with a spirit of cooperation, Pavlov thrives. N owak and Sigmund found that in such a circumstance Pavlov emerges victorious, even outstripping GTFT.
Both these strategies, Pavlov and GTFT, are somewhat tolerant to errors, unlike TFT. But Pavlov has another advantage. If in the model we allow strategies to randOInly mutate into new forms, GTFT comes to share some ofTFT's transience, becoming 'softened' by a gradual drift towards more unconditionally cooperative strategies. Pavlov, however, retains a hard edge. If it discovers by chance that it can get away with unilateral defection, it will continue to do so. So it is a wolf in sheep's clothing: it behaves well while cooperation becomes the norm under the firm authority ofTFT, but it remains quite capable of exploiting a cooperative population once the TFT police have been transformed to unconditional cooperators. The motto of a Pavlovian society is no longer 'Do as you would be done by', but 'Never give a sucker an even break.'
The simulations from which Pavlov emerged as victor reveal a fascinating history. Because they involved the interplay of strategies which all based their next move in some way on the previous moves of both players, they took a more complex course than the earlier simulations. Most strikingly, there was far less of a sense of inevitability about the changes that took place over time. Each run of the simulation produced a different sequence of events. In the sample history shown here, there is an early attempt to establish cooperation: after a turbulent period this fails, and unconditional defectors reign for a long time. Then, after about 92,000 generations, the cooperators gain the upper hand.
This victory is short-lived and soon collapses into defection. Close inspection of the breakdown reveals that it is caused by the drift of TFT towards GTFT and thence to more forgiving strategies, creating a nation of 'softies' which is ultimately destroyed by rogue defectors. But this time the defectors are not quite unconditional: the dominant strategy is instead one called Grim Trigger, which meets cooperation with cooperation until it encounters a defection (as, in a noisy game, it inevitably must). Thereafter, Grim Trigger defects unconditionally. It is rather like Dr Strangelove's Doomsday Machine. After about 220,000 generations, however, there is another resurgence of cooperation, which - after some initial adjustments - proves long-lasting. This switch is again triggered by TFT, but gradually it drifts towards a predominance of GTFT players before ultimately being taken over by mostly Pavlov agents, or close variants of it. This population is cooperative but potentially opportunistic, and is robust against invasion by defectors. It is not such a bad place to live - but the more virtuous of its citizens are not entirely safe from the threat of exploita­tion by superficially 'nice' Pavlovians. These simulations display a mixture of chance and certainty.
Cooperation always wins out if you wait long enough - there is always a happy(ish) ending. And Pavlov is not always the final dominant strategy, although it triumphs in about four cases out of five. Most remarkable, though, are the sudden revolutions that punctuate the course of events: we see 'good' and 'evil' empires rise and fall, and uprisings that falter and fail. Even in periods of apparent stability (for better or worse), the tally of strategies (and a detailed inspection of their characteristics) shows a certain amount of variation and a shifting of norms.
It is hard not to see in all of this an allegory of human history. Marx believed that the socialist revolution was inevitable. Game theory seems to be saying that nothing is so certain, since even if things are going to end up a particular way, we can't be sure just where along the evolutionary path we are at the moment. Did the players in the Second Cooperator Uprising (generation 92,000) think that the Age of Perpetual Cooperation, long forecast by the martyred philosophers of the failed First Uprising, had finally arrived? Were the commentators of the Third Cooperator Uprising (gen. 220,000) right to conclude that this was the 'end of history'?
Empires rise and fall not only in time but in space. Rome once ruled from Portugal to the Black Sea, from the Scottish borders to North Africa. Charlemagne's Frankish realm reached deep into Germany, Italy and the Balkans. The Ottoman Empire overran lands from Transylvania to Egypt. The history of the world is a patchwork of borders, growing and shrinking. Imperialism seems mercifully to be a thing of the past, but the borders of NATO and of Europe are still liable to shift, and the maps of the Central and East European nations have changed more in the past ten years than in any other decade since the end of the Second World War. Does the Prisoner's Dilemma have anything to tell us about the way national and international boundaries move?
Introducing the element of space into the contests of game theory is no trivial matter, since it places constraints on the interactions that each player can have, and can therefore strongly influence the outcomes. In a tournament where everyone plays against everyone else, cooperators have the chance of deriving mutual benefit from their interactions; if they are on opposite sides of a map, they can no longer draw strength from the group and may be overwhelmed by defectors. So isolation can militate against cooperation. Israel's geographical location, surrounded by predominandy Islamic states, surely contributes to its self-perception as an embatded nation, and its supporters might argue that this situation prevents the country from being able to adopt the kind of conciliatory policies that European nations can afford to pursue. On the other hand, players in fixed locations may find a greater incentive to cooperate than do itinerant players, since they are compelled to interact repeatedly with their neighbours rather than moving on after 'one exchange. One of the difficulties faced by travelling people is that they have litde opportunity to establish mutually trusting (and trustworthy) relations with those they encounter on their journeys - the 'shadow of the future' is not long enough.
Axelrod began to explore the notion of territoriality in the Prisoner's Dilemma in the 1980s. He considered a chessboard-like world in which each player occupies one grid space and interacts with the four neighbours with who it shares an edge. Colonization of the board by successful strategies can take place by means of an 'evolu­tionary' mechanism. In each round of the game, all players play against all four of their neighbours. If one or more neighbours of a particular player gets a better score, the player converts to the strategy that was most successful.
Axelrod was primarily interested in how cooperation can spread in an exploitative society by means of Tit For Tat - or conversely, how a cooperative society can be undermined by defectors. He found that, for certain values of the payoffs, a single defector can spread its baleful influence throughout a community of Tit For Tat players, providing the seed from which defecting strategies gradually expand. Curiously, however, the growing colony of defectors is not so much like a spreading stain as like a snowflake: it sends out branches which split and rejoin to form a complex tapestry.
Strictly speaking, the 'spreading of defectors' is really the spread of the use of defection strategies among players that stay immobile at each grid point. But one could equally choose to regard this as the replacement of cooperative players by defecting players, as though the 'defectors' kill off the 'cooperators' and colonize their grid spaces. The two perspectives are entirely equivalent. If defection is more lucrative, however, it becomes the :ommonest mode of behaviour. Yet small islands of cooperation constantly appear and vanish across the grid. The fraction of cooperators now quickly evolves to a stable average level, irrespective of the starting configuration - there is a kind of 'irrepressible' tendency to cooperate even in a relatively selfish environment. Thus defection md cooperation need not inevitably annihilate the other, but can coexist indefinitely in patterns which are unpredictable in detail but entirely predictable in terms of averages.
The spread of defection and the spread of cooperation are not equivalent. Cooperators do best in dense clusters, where they can benefit from their mutual support. As Edmund Burke put it in 1770, When bad men combine, the good must associate; else they will fall me by one, an unpitied sacrifice in a contemptible struggle.' But defectors in the midst of cooperators do best on their own, since they do far better to interact with cooperators than with other defectors. So although defection breeds more defection, the exploiters tend to repel me another, which leads to the formation of the thin threads of defection. Nowak and May found that a single defector spreads through a cooperative colony in much the same kind of snowflake permutations as Axelrod had seen. Again, these ramified magic carpet' patterns can be regarded as the result of the 'repulsion' between defectors, which militates against their forming a dense colony. You might say that each defector prefers to find its own patch, as isolated as possible from its rivals.

GOVERNED BY REASON?

The Tit For Tat policy and its more generous variants have defined most thinking about how cooperation evolves. Although TFT can yield to softer 'nice' strategies or to the opportunistic Pavlov once cooperation is established, there is still no better way to initiate cooperation in a Hobbesian world of exploitation than to meet tit with tat. On this basis, some argue that firm and immediate reprisals for bad behaviour are the only way to make 'rogue states' act responsibly: hence the recent bombings of Serbian Belgrade and the invasions of Afghanistan and Iraq. Popper put this conclusion into words which sound almost shockingly brutal, coming from one who had a reputation as a liberal: 'What is happening in Bosnia is proof of the failure, the cowardice, the blindness, of us in the West. It shows we do not want to learn what this century should have taught us: that war is prevented with war., The idea of'preventive war' goes back at least as far as Kant, who advocated it in his essay 'Perpetual peace' - not that kings and princes in times past needed any philosophical endorsement for taking up arms. It is certainly true that strategies in the Prisoner's Dilemma which make softer or delayed reprisals, such as Tit For Two Tats, fare less well in a diverse population. By itself this seems to argue for air strikes rather than sanctions.
But as we've seen, TFT has its drawbacks too. It is painfully evident in both the Israel-Palestine conflict and the troubles in Northern Ireland how reprisals can hold back progress towards cooperation and peace. They can simply serve to undermine the establishment of trust. There is clearly a need for mechanisms that can 'damp out' the echoing cycle of retaliation, if good relations are ever to be resumed between TFT players who, for whatever reason, have broken rank. The Contrite or Generous Tit For Tat strategies offer some solutions; another is to operate a Partial Tit for Tat approach, where the reprisal is slightly less severe than the event that provoked it.
Human nature seems likely to complicate attempts to generate cooperation via TFT exchanges. It would be foolish indeed to ignore the strength of passions or the longevity of resentments when human lives are lost in episodes of 'defection', while TFT remembers nothing beyond the previous round and will 'forgive' at the first sign of cooperation. And consider the proposal in the United States that the custodial parent in separated couples who have children might be given the right to deny visits if the other parent does not keep up maintenance payments. This scope for retaliation could be regarded as an incentive for the maintenance-payer to co­operate. But quite aside from the fact that this scheme would logically have to allow the converse (payments legally withheld if access to the child was denied), it would be short-sighted not to take into account the irrational behaviour that can arise in partnership break­down and arguments over child care responsibilities, which may override the ability of the 'players' to decide dispassionately what course of action is most advantageous to them in the long run. More pertinent still is the matter of whether there can possibly be any justification for letting children become bargaining chips in such exchanges.
The threat of a Tit For Tat response has been advanced, both implicitly and explicitly, as a theoretical rationale for the policy of nuclear deterrence. The argument here is that, even if the awful prospect of an exchange between nuclear powers does not materialize, the evident capacity and stated readiness of a nation to retaliate to such an attack is an essential element of a peaceful status quo. Popper again supported the notion: 'We should have learnt by now that peace on earth needs to be backed up with weapons ... You could never get peace inside a country by reaching a compromise with the criminals.,5 This might be valid for war in general; but for nuclear war - and in particular the Cold War concept of mutually assured destruction - the iterative process that is essential for TFT's superiority (indeed, for its very definition) is not an option.
Strategist Hermann Kahn describes with compelling clarity the confused and irrational thinking he often encountered at the RAND Corporation during the early years of the Cold War:
One Gedanken [thought] experiment that I have used many times and in many variations over the last twenty-five or thirty years begins with the statement: 'Let us assume that the President of the United States has just been informed that a multi-megaton bomb has been dropped on New York City. What do you think that he would do?' When this was first asked in the mid-I950s, the usual answer was 'Press every button for launching nuclear forces and go home.' The dialogue between the audience and myself continued more or less as follows:
KAHN: What happens next?
AUDIENCE: The Soviets do the same!
KAHN: And then what happens?
AUDIENCE: Nothing. Both sides have been destroyed.
KAHN: Why then did the American President do this?
A general rethinking of the issue would follow, and the audience would conclude that perhaps the President should not launch an immediate all-out retaliatory attack.
In other words, according to political theorist Brian Skyrms, 'A strategy that includes a threat that would not be in the agent's interest to carry out were she called upon to do so, and which she would have the option of not carrying out, is a defective strategy.
In any event, it would be naive in the extreme to assume that real players in the international politics game will display the rigorous rationality of the idealized players in game theory. Supporters of an 'assured Tit For Tat' deterrent, for example, must grapple with the strong possibility (as Popper admitted) that Khrushchev had every intention of using his missiles in 1962, had he been able to get enough of them into Cuba secretly - deterrent or no deterrent. Looked at in this light, TFT simply creates the climate and the conditions for such a crisis. Moreover, the spectacle of many long-term adherents of the deterrence argument for American nuclear proliferation who are now advocating a missile defence system which undermines the very basis of this argument - what is mutually assured destruction if it is not mutual? - should remind us that the formal logic of game theory is no more than an expendable tool in the face of political ideology and expediency.
Another of the less attractive implications of Tit For Tat is that in effect it compels cooperation from defectors only if they 'know' how implacable it is. This implies that one needs to acquire a reputation for being ready to go on the offensive, which can become manifest as hypersensitivity - refusing to tolerate the slightest perceived threat or insult - or as a bullying propensity to adopt gunboat diplomacy. In the 1960s the United States was prepared to fight a bitter war on the other side of the world primarily to maintain, during the height of the Cold War, its reputation for toughness. This was admitted in a memo sent by John McNaughton, Assistant Secretary for International Security Affairs, to Secretary of De fen se Robert McNamara, outlining US aims in Vietnam: they were '70 percent - To avoid a humiliating US defeat (to our reputation as a guarantor)' (my italics), and only 'IQ percent ­To permit the people of [South Vietnam] to enjoy a better, freer way oflife.'s
In other words, the success of Tit For Tat could be regarded as an incentive to act belligerently. After all, the disastrous consequences of facing an opponent who does not appreciate one's ruthlessness were brought home in a gloriously sardonic manner in Kubrick's classic study of Cold War diplomacy, when Dr Strangelove explodes down the phone at his Soviet counterpart, 'You fools! A Doomsday Machine isn't any good if you don't tell anyone you have it!'
Anyone who considers using the Prisoner's Dilemma as a basis for deciding policy should feel duty-bound first to enumerate all the factors t neglects. Most obviously, as I have indicated, it takes a highly simplistic riew of human nature: the assumption that people act rationally to seek he best gain for themselves neglects not only irrational passions, the faIlibility of our reasoning powers and sheer foolishness, but also the positive influence of moral codes of conduct. Both experience and :volutionary biology lead us to expect that many people have an innate nstinct to cooperate with their fellow beings, and do not have to learn that his serves their best interests before they will do it. On the other hand, a ew people are probably pathologically inclined to defect within society, lometimes even when they see it does them no good in the long run. Moreover, the Prisoner's Dilemma provides no scope for negotiation: the prisoners" remember, are not allowed to collude, but must deduce each }ther's motives only from the way they play the game. Suspicion thrives n such circumstances, and in real life we generally seek to quell it by :onducting our transactions more collaboratively.
All the same, there is no denying the strong implication from game :heory that an unswervingly retaliatory strategy is the best way to bring about cooperation. As game theorist Karl Sigmund says:
It would of course be rather silly to attempt to reduce all human interactions to the Iterated Prisoner's Dilemma, or to negate the role of superior authority In civilized communities. But with all due restraint, it is worth pointing out :hat the brutally simple principle of paying back in kind leads to cooperation In a society of egoists, while the apparently higher summons to dispense with reprisals undermines such cooperation ... The harsh law of retaliation seems to have been the foundation stone of many, possibly all, stable societies.
At the risk of posing a question which evolutionary psychologists might regard as tautologous, one surely has the obligation then to ask: is this a moral way to behave?


The notion of Tit For Tat sits uncomfortably with liberal sensibilities. 'I came to this project', said Robert Axelrod in 1984, 'believing one should be slow to anger. The results of the Computer Tournament for the Prisoner's Dilemma demonstrate that it is actually better to respond quickly to a provocation.'IO But pacifist thinkers from St Francis to Gandhi have asserted that to meet violence with violence is self-defeating. That is surely the message of the New Testament: love thine enemy, for it is the meek, not the vengeful, who shall inherit the earth. Many pacifists would argue that non-violence is a choice based not on cold logic but on higher moral imperatives, such as 'Thou shalt not kill.' When faced with behaviour that can only be considered as exploitative, even murderously so, this can induce agonies of doubt which, if honesdy confronted, belie any suggestion that pacifism is a soft option. David Jones, a conscientious objector in the Second Wodd War, explained the dilemma:
The pain of being a conscientious objector was the increasing knowledge of the enormity of what the Germans were doing. So that was the really challenging thing, not the war, so much as how one can justify not somehow trying to do something about it.
Cecil Davies, a kindred spirit in the same conflict, found the question irresolvable:
Wilfred Owen said he was a conscientious objector with a bad conscience, and I think lots of COs often had a bad conscience, and while I still think I was right to do what I did when I did, I suppose if! had known about the Holocaust it might have been different ... Life isn't simple.
Indeed not - and the Prisoner's Dilemma should not let us pretend otherwise. But one cannot avoid the conclusion that, on its own terms, game theory implies that a retaliatory policy to defection might truly be more 'moral' insofar as it serves the greater good. A Tit For Tat strategy not only protects oneself from exploitation but also helps to safeguard the entire community. For it is only when TFT is eroded by softer cooperative strategies that the community becomes endangered by exploitative defectors or Pavlovian opportunists. Unconditional cooperation might seem more noble and kindly but it also places the burden of policing on the rest of the community. A utilitarian would be hard pressed to find an objection to TFT.
All of this seems like knowledge worth having. But we must beware of how easily it might be twisted or misconstrued. Naively, one might argue that TFT justifies the death penalty for murder - even if we allow that the state, not the individual, should implement the punishment. But the whole point of the iterated Prisoner's Dilemma is that it teaches cooperation through experience and adaptation: defectors convert to nice strategies because they are better off for it. Capital punishment simply eliminates the player, so that no further iterations take place. There is nothing in the Prisoner's Dilemma which suggests that one player learns from the mistakes of another - that the death of one defector warns others away from defecting. This could happen, of course, but game theory is silent about it - so the success of TFT is irrelevant to debates about the death penalty. The same surely applies to nuclear stockpiling during the height of the Cold War. Game theory and the Prisoner's Dilemma were very popular at the Pentagon, but one could hardly draw lessons from TFT when the game was, by its very nature, one that would be over in a single reciprocal exchange.
John Locke had the wisdom to see that, while he advocated the 'tooth for a tooth' form of Tit F