Moralizing gods update: Seshat still searching for something that isn’t there

Last week, Turchin et al posted a reply to critiques of their Nature paper that argued that social complexity precedes moralizing gods. This goes against the big gods hypothesis that the presence of moralizing gods in human social networks allowed the growth of larger and more complex communities. Instead, Turchin et al argue that larger groups come first and that moralizing religions evolve as conflict mitigating adaptations to help those larger groups stay cohesive.

The alternative hypothesis is supported by what Turchin et al’s Seshat team describe as a team of “rival researchers”, apparently led by Edward Slingerland. Seeing these two groups going at it on social media, one could easily conclude that while moralizing gods may or may not precede large, complex societies, they definitely don’t precede archaeological big data research teams.

The problem of absence of evidence is still there

Bret Beheim, one of the more level-headed critics of the original paper, was quick to point out (See the comments in Turchin’s latest blog post) that Turchin et al have still not addressed the central concern that the Seshat database, on which the analysis is based, treats absence of evidence for big gods as evidence of their absence. If there is no historical or archaeological evidence of moralizing gods for a certain time and place, Seshat assumes that they are absent.

Since the original critiques, lots of data has been added to Seshat. Lots of data has been validated. The time scales and regional coverages have been refined. But the analysis apparently still assumes that if there is no evidence of big gods for a region and period, they don’t exist.

I’ve argued before that in archaeology, we have to be extremely careful when looking at an absence of evidence. We really learn something about the past when we find something in the archaeological record that we weren’t expecting. However, because the archaeological record is fragmentary, because we don’t fully understand how some things are preserved, how some aren’t, and how some are transformed over time, we can’t be sure that we can learn something useful from absence of evidence. If something isn’t there now, it might still have been there in the past.

Beheim is right. But the objection goes even beyond the specific way that the Seshat database treats absence of evidence. The fundamental problem remains that the Seshat team’s approach to testing their hypothesis depends not merely on not finding evidence of moralizing religions in the past, but it depends on demonstrating that they don’t exist at certain times and in certain places. That’s an archaeological bridge too far.

The difficulty is compounded by the fact that what they are trying to not find in the past, does not leave any immediately obvious material traces in the first place. Regardless of methodological sophistication, I struggle to envision an approach that would allow them to find that any particular belief is absent at a certain time and place in the past, including belief in moralizing supernatural agents or principles. I can see how we could conclude that they are present, but how do we know they aren’t?

On mere intuition, we are more likely to find evidence of moralizing religion in complex societies with writing, than in others. We would have to somehow reject that intuition before we could even start thinking about whether the record can tell us anything about the presence of moralizing religions in other kinds of societies.

What are the archaeological correlates of moralizing gods? Moralizing gods only know. Crucially, even figuring this out wouldn’t tell us whether moralizing religion was really absent if we find no evidence of it.

The Seshat team developed a tool to test a hypothesis they aren’t evaluating

The “powerful new way of testing theories” on cultural evolution that the Seshat team proposes might indeed work wonders if they were trying to see if metallurgy precedes urban centers, for example. While there are legitimate definitional issues to resolve with both concepts, they both leave more obvious traces in the archaeological record than either moralizing religion or social complexity. 

Every time we fail to find metal objects or evidence of metal processing in a certain archaeological context, we increase our level of confidence that it was actually absent in that time and place. I can’t say the same for moralizing religion, or social complexity, even if we could clearly define them. 

Through careful examination of the archaeological record, using a sophisticated database and the statistical approaches of the Seshat team, I could see getting to a high degree of confidence that metallurgy precedes or doesn’t precede urban centres. I wouldn’t be absolutely certain, but I would be confident. I don’t see how, at the moment, I would achieve any sort of confidence that moralizing religions are not present before large, socially complex societies.

The problem is not with the specific statistical approach of the Seshat team. Rather, they have developed a powerful approach and data infrastructure for testing a completely different hypothesis than the one they are evaluating. They have obviously worked very hard since the critique of their original paper, which Beheim acknowledges. However, what they have done is a bit like adding very elaborate decoration to an arch that doesn’t have a keystone, in response to the critique that it is structurally unsound. The arch will eventually be gorgeous, but it still won’t have a keystone. It won’t even be a proper arch.