The Origins of Large Scale Cooperation III

This is the third in my three part series about origins of large-scale cooperation in human societies. There is much more to say about the topic, and perhaps I will return to it later, but next time I plan on moving to a different topic.

Today’s focus will be the development of cooperation in the Neolithic, and we will explore three hypotheses: that food surpluses, that institutions, and that common threats such as warfare are major drivers of cooperation. As usual, this work is supported by Coefficient Giving, though all conclusions here are my own.

In the previous two posts, we explored how groups of human societies in the Paleolithic (the Paleolithic precedes the Neolithic, which began around 11,700 years ago in the Near East and at later times in other parts of the world) cooperated on a large scale. We saw that the nature and extent of large-scale cooperation is debated. Two recent papers in particular—Boyd and Richerson (2022) and Singh and Glowacki (2022)—suggest that cooperation between distinct groups of hunter-gatherers was common, challenging the conventional view that such cooperation was rare and not a significant part of the human evolutionary milieu. In particular, the latter paper proposes the “diverse histories” model, which holds that both large-scale cooperation and group separation, as well as nomadism and semi-sedentarism were both common. It follows that no particular lifeway is “natural”, and humans are prepared for either one.

However, we also saw that the evidence behind these two papers is patchy. While the conventional view does deserve to be challenged, it would be a mistake to conclude the opposite and find too much similarity between Paleolithic societies and the settled farming societies of the Neolithic.

In any case, it is clear that the Neolithic brought about major increases in the scale of cooperation, whatever the prevailing situation was before that.

A leading hypothesis about the expansion of cooperation is that it is tied to food surplus. Vásárhelyi and Scheuring (2018) build a model of how division of labor, a more narrow form of cooperation, may have developed. In their model, individuals have both a personal proficiency and a genetic fitness for a task, both of which improve with performance. In a subsistence society, individuals need to remain generalists, as the ecology punishes gaps in skillsets more than it rewards expertise in a particular area. With a food surplus, the situation reverses and specialization is rewarded more than generalization. The authors suggest that their model may explain the “Sapient Paradox”, as described by Renfrew (2008)—if humans had achieved anatomical and cognitive modernity by about 60,000 years ago, then why did it take until 10,000 years ago for technology to begin advancing rapidly? The explanation of Vásárhelyi and Scheuring (2018) is that the missing ingredient, fulfilled around 10,000 years ago by favorable ecological conditions, was food surplus.

Shavit and Sharon (2023) approach the question of cooperation and the Neolithic revolution from a different angle. A central puzzle of early agricultural societies is that, especially during a bad wet season, it appears rational for individual farmers to revert to hunting and gathering, even though it is best for the group as a whole for them to remain farmers. The authors posit that social pressures, such as from a “shaman”, would have prevented reversion. This is evidence, according to Shavit and Sharon (2023), that cooperation preceded agriculture and was thus a cause, rather than an effect as in the model of Vásárhelyi and Scheuring (2018), of agriculture.

Turchin et al. (2022) also find that surplus has been a major driver of large-scale cooperation. Combining cultural evolution theory, as described by Richerson and Christiansen (2013) and data in Seshat: Global History Databank, the authors find that agricultural productivity, as well as the adoption of military technologies—particularly iron weapons and cavalry in the first millennium BC in Eurasia—are the best predictors of social complexity. “Social complexity” is defined as a composite of three measures: social scale, including metrics related to population and geographic area; the number of levels of hierarchy; and metrics related to government sophistication.

Robinson and Barker (2017) examine cooperation more broadly, particularly among humans and ants. In both cases, they find that resource sharing is a major driver of inter-group cooperation, again showing how a food surplus may be a driver of large scale cooperation.

As discussed in previous posts, Boyd and Richerson (2022) and Singh and Glowacki (2022) both challenge the notion that surplus is a decisive factor in the emergence of large scale cooperation, noting that such cooperation preceded the surplus that would be generated from agriculture. Popular books such as James Scott’s Against the Grain: A Deep History of the Earliest States (Scott 2017) invert the relationship and argue that the emergence of large, social hierarchies led to surplus which early states coercively extracted from farmers, rather than a surplus allowing the emergence of hierarchies. Nevertheless, while surplus is certainly not the only factor to explain cooperation, the bulk of research suggests that it was, at least, a major factor.

Surplus many enable cooperation, but cooperation also requires coordination, which in turns requires institutions, which we examine next.

I briefly touched on institutions in my last post, but they are an important part of the story that may deserve at least a full post of its own.

Today, I am using the definition of an “institution” from Hurwicz (1996), which is that an institution is the set of possible game forms in a society. What this actually means is perhaps better illustrated by examples. Property rights, criminal law, the United States Constitution, the patent system, and money are examples of institutions. These are structures that define how the “players” of the game (e.g. individuals and organizations) are to behave.

A more conventional idea of an “institution” would entail organizations, such as university, a government, a central bank, or a church. These are not institutions under the definition of Hurwicz (1996). They are players in the game, whereas an institution is the game itself.

Powers, van Schaik, and Lehmann (2016) distinguish economic games and political games. Economic games are played on a regular basis by the individuals and organizations within a system, while political games are played less frequently and govern how the economic games are structured. Elinor Ostrom (see e.g. Ostrom 2009), who won a Nobel Prize in Economics for her work with institutions, further considers constitutional games, which determine how political games are played, and meta-constitutional games, which determine how constitutional games are played. The process can only be regressed so far before one reaches immutable laws of human nature.

In the last post, I discussed Powers and Lehmann (2017), which discusses several mechanisms that are posited to drive large-scale cooperation and argues that they show diminishing returns with scale, and thus they are unable to explain cooperation on a very large scale. I briefly mentioned that they conclude that institutions are the one mechanism which can plausibly extend to arbitrarily large scales.

Powers, van Schaik, and Lehmann (2016) explore this hypothesis more thoroughly, regarding institutions as central to the emergence of social complexity. Examples of key institutions were rules for food sharing, rules governing the usage of the irrigation system in an agricultural society; property rights; and later on, rules for sharing reputation among medieval traders. They argue that cooperative breeding—a system by which individuals who are not among the breeding pair help take care of the young—selected for individuals with prosocial orientations and thus helped the capacity of institution formation to appear among Homo sapiens.

Powers, van Schaik, and Lehmann (2016) also assert the importance of language and that nonlinguistic animals are incapable of forming institutions. This position is plausible but not entirely uncontested. Tomasello et al. (2005), for instance, argue that great apes, which are nonlinguistic primates, show behavioral regularities, which in some ways superficially resemble institutions, but are not institutions. Behavioral regularities, such as a flock of birds flying in unison, are widespread in the animal kingdom. According to Tomasello et al. (2005), great apes show awareness of each other’s intentions, but they do not have the shared intentions that are the basis of institutions. Precisely distinguishing between institutions and behavioral regularities is a complex and not settled question.

Casari (2007) provides an account of how institutions can develop, working specifically in the context of property rights and communal governance in the Italian Alps in the late Middle Ages and early modern period. Another valuable paper is Milgrom, North, and Weingast (1990), which discusses by theoretical model the emergence of institutions governing trade in medieval Europe. The challenge for these authors is to explain how norms governing trader behavior may arise when traders are numerous. If one trader cheats another, the two are unlikely to interact again, and so the cheater might not appear to have a strong disincentive to cheat. The model of Milgrom, North, and Weingast (1990) includes the emergence of judges, or law merchants, who spread information about cheating traders. Under the model, a trader pays a fee to a law merchant to gain reputational information about a possible counterparty, and the trader can also pay a fee to file a complaint against a counterparty. The law merchant then levies penalties. The penalty is reputational rather than legal; failure to pay a fine for cheating leaves a trader with a bad reputation and therefore an inability to keep trading. The model is robust against cheating among law merchants. Milgrom, North, and Weingast (1990) present the model as a voluntary system that spontaneously arises among traders and suggests it as a model for how institutions may have developed in the early Neolithic period.

Finally, I’ll mention Powers and Lehmann (2013), who have had quite a lot to say about institutions, about how demography and institutions have co-evolved. They build a model in which a patch of resources can support social and asocial populations. Under the model, individuals from social populations can invade asocial populations and create institutions, which thereby enable investment in common goods that raise carrying capacity. Under this model, it would be too simplistic to say that institutions enabled agriculture, or vice versa; the two would have been mutually reinforcing. This adds to the simplistic narrative that institutions caused agriculture in a one-way fashion.

There is one more hypothesis to explore today about the emergence of large-scale cooperation in the Holocene: warfare. Stanish and Levine (2011) explore this hypothesis in the context of first millennium BC states around Lake Titicaca in modern-day Peru. The authors find warfare from about 500 BC and a burn site in the first century AD that is most likely the site of major violence. Following the burn, there is evidence of political integration among the states in the region. The authors conclude that warfare is a causal driver of integration.

The aforementioned Robinson and Barker (2017) also discusses common threats as a driver of cooperation. Among many animals, group threats such as predators and harsh environmental conditions are drivers of group formation, and they cite several references on how threats lead to group fusioning, or the merging of several groups into one that can cooperate against the common threat. Warfare would plausibly be such a common threat. Note, however, that group fusioning is different from large-scale cooperation. In the case of cooperation, the groups remain distinct, while under fusioning, the groups become one single group.

The nature and extent of large-scale cooperation in the Paleolithic is unclear, but it is clear that cooperation expanded greatly after the advent of farming. We examined three hypotheses on why this happened. First, we saw that large-scale cooperation, a latent capability of humans, was enabled by the food surplus that favorable climatic conditions of the early Holocene allowed. Second, we saw cooperation was enabled by the uniquely human ability to build institutions, which was further enabled by those same climatic conditions. Third, we saw more briefly that shared threats such as warfare may have been a driver of cooperation.

The evidence in the papers discussed today does leave something to be desired. Many of the papers build theoretical models, which are sound in their own rights, but it is not clear how well the models describe actual societies. Regarding Stanish and Levine (2011), the authors demonstrated warfare and state formation around Lake Titicaca in the late first millennium BC and early first millennium AD, but the causal connection between the two is not clear. Nevertheless, the story I have told today has emerged as a conventional narrative in anthropology of how civilization as we know it emerged from our hunter-gatherer prehistory, though by no means is it universally accepted.

In my mind, this narrative still leaves many open questions, some of which I may explore in the future. How and why did language develop? Given the connection of language with institutions, this is necessary to understand the human story. There is also a common narrative that the Holocene brought about the favorable climate needed to allow human civilization to develop; previously civilization was a latent capacity. Is this story really accurate? This, again, is the Sapient Paradox.

Comments and corrections about the posts in this series, especially from people with expertise in anthropology, would be greatly welcome. I have greatly enjoyed researching these posts and am hoping to return to the subject some day.

Last month, Paul Ehrlich died at the age of 93. He is best known for his work on what he believed to be overpopulation. I wrote critically about his most famous work, The Population Bomb, last year. Among the people I follow, the reaction to Ehrlich’s work has been overwhelmingly negative, and justifiably so. Nevertheless, there remains a strong audience for overpopulation worries and for various forms of population control. As I wrote in 2024, there is more of a link than most people realize between population control and immigration restriction, and in the United States right now, anxiety about immigration is at a fevered pitch. As recounted in an article at The Conversation, Ehrlich served on the board of advisors of the Federation for American Immigration Reform (FAIR) until 2002. FAIR was at the time, and still is, one of the leading anti-immigration organizations in the country.

This is my first blog post for which I have used a large language model, specifically Claude Pro, and so in the interest of transparency, and also because it is interesting, I would like to clarify how I use it. Claude has assisted me in finding papers on particular subjects. About a third of the papers discussed today were found from queries to Claude, and the rest were found through other means. I have reviewed all sources discussed to insure that I am representing them accurately, and though Claude’s summaries helped me identify which papers to read, I have not found the summaries reliable enough to assume they are accurate. Claude fact-checked an earlier draft of this post and also reviewed the post for spelling and grammar mistakes, though I do not claim that all mistakes were removed. Claude did not write any of this post directly, nor did I copy and paste any material into the post.

I tried ChatGPT briefly in 2023, and at the time, I found it to be a fun toy but not something that I would use for serious work. The picture has greatly changed in three years. If the quality of large language models advances in the next three years like it did in the last three years, the world might look very different in 2029.