Early medieval consensus and the prisoner’s dilemma

I’ve been thinking for a while about a post by Guy Halsall in which he expressed scepticism about the recent emphasis on the ‘consensus model’ in early medieval politics, and dived briefly into game theory to argue that the struggle for power in the early Middle Ages was in material terms, pretty much a zero-sum game: if you had land/animals, someone else didn’t. I don’t want to get into arguments about Guy’s wider point on whether or not an early medieval state existed (which relies crucially on how you define ‘the state’). But I do want to think a bit harder about the extent to which the early medieval struggle for power is a zero-sum game, because I’ve been arguing in terms of attitudes towards Carolingian masculinity that there is a lot of consensus and co-operation within the elite.

As a technical point, if you’re going to argue that it’s really a zero-sum game, you’re saying that there was no growth in resources at all in the early medieval period, which is a very strong statement to make. The material evidence suggests that there is population growth, increasing wealth (as seen in coinage etc). I don’t, however, think that this invalidates Guy’s argument. If economic growth is very slow (and even today, annual GDP growth in the UK can be on the order of 1-2%), for practical purposes, Guy is right that you do have a zero-sum game. You can’t easily and quickly recover from serious material losses in battle, for example.

But you can accept that a zero-sum game is a reasonable approximation of the overall situation in one society without that implying that kings and aristocrats necessarily have opposing interests. Partly this is because ‘aristocrats’ as a group are not a player in this game – particular aristocratic individuals/families/collectives, etc are. And also because kings and magnates aren’t the only players in this game – so are individual peasants and peasant collectives. That’s why someone like Chris Wickham, who certainly doesn’t believe the early Middle Ages is all sweetness and light can accept the model of consensus politics. It’s just that he sees this as a conspiracy between members of the upper classes in order to oppress the peasants more effectively. (It’s certainly noticeable that whenever there’s a peasant rebellion later in the Middle Ages, or the early modern period, aristocrats previously at each others’ throats start to co-operate in putting the lower orders back in their place).

Guy, however, is arguing that this alliance between local elites and the central power isn’t necessary. Whether or not it’s necessary, I think in some cases we do have evidence that it existed, at least by the Carolingian period. Einhard’s letters, for example, show how a courtier is able to get favours for his friends from the king, and his letters end up being used as a formulary, which suggests that this is fairly standard practice. But I want to look at consensus from a rather different angle, looking specifically at Guy’s argument about warfare as a zero-sum game.

Guy makes his point with a simple example:

By way of a hypothetical illustration let us suppose that territory X, a component of Kingdom A, can support 200 warriors from the surplus generated by its land. If access to Territory X is lost by the King of A, either through conquest by Kingdom B or the secession of the local Governor of X, then the King of A loses 200 warriors, whose control is gained either by the King of B or the Governor of X.

This is accurate in as far as it goes: King B gains, King A loses. But in practice, this isn’t really how things works. There is always collateral damage in war: so B probably loses at least 5 of his own warriors gaining this land, which may well have been harried enough that its productivity has declined temporarily. In other words, any war is likely to lead to an overall loss of resources.

If you like at it like this, what you have is a classic prisoner’s dilemma game. The best overall solution (the one that preserves most resources) is if A and B don’t fight. But the best solution for either A or B is a successful attack on the other. However, this risks the worst overall solution, which is that you end up with dozens of warriors dead on both sides and the land you’re fighting over rendered useless. When you have warfare between two relatively evenly-matched parties, it’s possible that everyone loses: in that sense it’s not a zero-sum game.

This doesn’t simply apply to warfare, either. In any game with rules, it’s always preferable if you’re allowed to ‘cheat’ and the other players aren’t. People are naturally anti-Kantians: they want rules and also want personal exemptions from them. (My ideal library, for example, is neither a reference library, nor a normal lending one, but one from which I and only I am allowed to borrow). But competition between powerful men, is always particularly prone to escalate in destructive ways, whether it’s potlatches, or spending on political campaigns, or revenge killing.

This is where consensus politics come in. It’s not a fluffy, bunny-hugging hippy idea, but one of a number of tools used to stop powerful men with retinues going for each others’ throats and ripping society apart. The preferred mechanisms vary, according to the polity and the scale involved. ‘Stateless’ societies (such as early medieval Iceland or Patrick Geary’s C11 France) could still have some complex mechanisms for preventing resort to customary vengeance and for dispute settlement more generally. In polities with rulers, other patterns are also visible, such as the ruler as a supposedly neutral arbiter between disputing parties, and there are pre-emptive attempts at bonding, encouraging magnates/warlords to co-operate rather than slaughter one another. The same patterns can sometimes be seen between kingdoms as well, reaching its extreme in the interminable struggles between Carolingian rulers in the second half of the ninth century. Agreement and co-operation between rival rulers wasn’t something that could be expected spontaneously, but instead was something deliberately and laboriously created.

Going back to the game theory aspect, the outcome of the prisoner’s dilemma is very different if you have repeated games. The most effective strategy in any one game is to ‘defect’ (to cheat or resort to violence), but in a repeated game, co-operation (and a limited willingness to forgive others) is more effective. Looked at in this way, consensus politics is an attempt to encourage some co-operation. It’s not surprising that such an attempt wasn’t always successful, given the short-term temptations for individuals of resorting to violence, but in the slightly long term, such a strategy could benefit the ruler and the aristocracy.

Not only did such consensus politics potentially prevent damaging internal conflicts, but co-operation developed within a polity could also provide the military basis for expansion. Agobard of Lyons in 833 (Liber Apologeticus 1-3), criticising Louis the Pious in 833, deplores the preparations being made for civil war:

when the army ought to have been sent against external peoples, and the emperor himself should have fought against barbarian nations, so that he might have subjected them to the faith and extended the boundary of the kingdom of the faithful

Consensus politics in the early Middle Ages, I would argue, never excluded the use of violence. It was just a question of who that violence was used against.


9 thoughts on “Early medieval consensus and the prisoner’s dilemma

  1. There are many theories regarding reciprocal altruism (Prisoners’ dilemma) and the growth of the state and political order. See, for example, Francis Fukuyama’s latest, The Origins of Political Order. Bottom line (from my reading of it, anyway): authority derives, ultimately, from trustworthiness, i.e. proven ability to cooperate. Also, of course, from nepotism/kin selection.

    The book paints its arguments with broad strokes but is admirably clear. I’m enjoying it.


  2. How does this account for people who simply change sides? I’m exhausted right now and a little fuzzy, but aren’t there several cases just in the Annales Fuldensis that mention shifts of allegiance between Charles the Bald (or maybe Lothar) and louis the German? It seems to me that the knowledge that nobles/leading men *can* sometimes switch sides makes consensus even more sensible whenever possible.


  3. Thanks for a very thoughtful and convincing illustration of how we should be envisaging early medieval consensus. A lot of what you are saying sounds very sensible and comes close to what I tend to think of in terms fo how consensus worked at assemblies. As you point out, consesus is not ‘fluffy’ at all and consensus rule does not need to involved everyone getting along or getting their own way. You may already be acquainted with this, but Steffen Patzold has a very good recent article on the topic in which he argues that consensus always went hand-in-hand with competition: ‘Konsens und Konkurrenz. Überlegungen zu einem aktuellen Forschungskonzept der Mediävistik’, Frühmittelalterliche Studien 41 (2007), pp. 75–103.


  4. Thanks for all your comments. A few specific reactions:

    Cosma – the possibility of using mathematical models in the humanities is one that keeps on looking interesting, but as you’ve shown many times, the results can be distressing. But I’m not sure that anyone can do a better job on a lot of medieval topics, because we’re missing too much data. We’d just have to guess at suitable parameters, and probably make a hash of it. So I can do a plausible hand-waving suggestion of why warfare is a prisoner’s dilemma, but there is no sensible way to produce a genuine payoff matrix, which would allow you actually to model situations.

    In contrast, social network theory might be possible to apply to early medieval situations, because we have thousands of charters with witness lists to work with. And at a very quick look at Greif’s work, he’s focusing on a part of the Middle Ages (the thirteenth century onwards) where we actually have some useful data about trade – some aggregated trade statistics (mostly from England) and a decent number of individual merchants’ contracts and law court reports.

    It’s possible there’s some early medieval Byzantine/Islamic material that I don’t know about, but we simply don’t have that kind of trade information for earlier in Western Europe. Between 500-1000, for example, we’re effectively reconstructing trade patterns from archaeological finds, which can show us broad upward and downward trends of consumption/production and exchange networks, but doesn’t give us any absolute quantitative data to work with, or enough chronological accuracy to connect changes to historical events. I suspect the most we can do for my period is take patterns that are seen in areas with better data and say ‘this sounds psychologically plausible and isn’t directly contradicted by the available evidence’.

    Nicola – how far back historically does Fukuyama go? Most analysis of ‘the state’ tends to focus on the world post-1500, perhaps because there’s all this argument about to what extent there are ‘states’ earlier (at least in the West). What I find interesting is how stable the idea of a kingdom is, long before that. Even in places where kings have no real power, the formal structure of kingship/rulership often seems to survive and have real prestige: such as in late Carolingian Catalonia, (as discussed by Jonathan Jarrett), the Japanese shogunate, or some of the Roman emperors. How many oligarchies or republics or democracies can show that kind of institutional continuity, with the exception of the Venetian republic? Medieval polities didn’t and couldn’t do a lot of things that modern states did, but I think this kind of mediating/focal point role was very important, and explains why magnates were surprisingly often prepared to remain at least nominally as part of some objectively fairly hopeless polities.

    ADM – the Carolingian rhetoric of consensus is already visible under Charlemagne, so before magnates had much opportunity to change sides (unless they supported Pippin the Retrospective Hunchback). But yes, that new option makes it all the more necessary. Stuart Airlie and Mayke de Jong argue that part of Louis the Pious’ problems with his sons and magnates was that he created so much uncertainty about his favour. Constantly switching around which son was going to inherit which territory, and hence who the magnates needed to ingratiate themselves with, was inherently destabilising. In contrast, consensus is part of a series of tools that are intended to emphasise that all of the elite matter. It’s for the same reason that Hincmar says in De Ordine Palatii that the king should have courtiers from all regions of the kingdom. Much of the art of kingship was persuading magnates that they had enough of a chance of benefiting from associating with the king for it to be worth them staying loyal, and consensus-building was a key part of that.

    Levi – thanks for the Patzold reference. I have actually got a copy of that article, but hadn’t got round to reading it yet (!), so it’s a useful reminder.


  5. Magistra — One can certainly see that Charlemagne is doing that with the Mainz magnates. We’ve got charters there for e.g., Otacar during Pippin’s reign (I think the first is 756) and then later, he is referred to as Charlemagne’s fideles. Coopting locals is a useful thing.

    PS — have sent an email to your personal addy re visiting.


  6. I suspect the most we can do for my period is take patterns that are seen in areas with better data and say ‘this sounds psychologically plausible and isn’t directly contradicted by the available evidence’.

    That’s fair enough; I was mostly agreeing, and pointing at someone else who found cooperative solutions in game theory useful for thinking about medieval history. But I suspect you are right that it will be easier and more productive to look at social networks than to try applying game theory*. It would be really fascinating to know how network structure has varied over time and space, even if only for elites, and even if the data aren’t as complete as what Padgett and Ansell found for Renaissance Florence.

    *: On a hyper-technical note, John Sutton has some great books [1,2] on extracting quantitative results which hold across all games of a given form, without knowing the exact pay-off matrix. This means he can (e.g.) make predictions about how the concentration of an industrial sector relates to how R&D-intensive it is, without having a fully worked-out model in the usual sense. Of course even there, he’s relying on having data of a sort which may just not exist for earlier periods.


  7. Magistra, it’s a two-volume text. The first goes all the way back to early hominins (!) and ends in the 18th C (I think; I’m only at the simplistic band–>tribe–>state explanation stage right now).

    Jonathan, I’m enjoying it as an Evolution of Human Politics 101 text, an introduction of concepts. The way I read such things is to believe everything as I’m reading, then put it down and think about it. Given the errors I’ve spotted so far (grammatical, for example) I don’t think I’ll be using it as an authority–more like an excellent Wikipedia article, a jumping off point. It’s very readable (though if he confuses ‘climactic’ for ‘climatic’ one more time, I think I’ll kick something).


Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s