Thoughts on ‘Behave’ by Sapolsky

Behave: the Biology of Humans at Our Best and Worst by Robert Sapolsky is one of those books that has clarified my thinking on a lot of issues — some obvious, like biological determinism; others less obvious, like the role of science. At 700 pages it is one of the few such books which deserves its length (another recent example being Adam Tooze’s Crashed). This is not a review, just a place to collect a few of the key things the book make me think about, and as a warning they do not follow a coherent narrative — each subsection is basically self-contained.

‘As if’ theorising

The idea it is appropriate to build theories about the economy based on armchair reasoning rather than on close observation of how firms, consumers and governments actually work has always struck me as pseudo-scientific nonsense. Science is about understanding how things work and building theories that match that understanding as well as having predictive power. Theories with unrealistic assumptions but good predictive power are necessary when we don’t know how things work ,but when we find out how things work and it isn’t the same as the theory that means the theory is wrong.

The history of science is filled with examples where realism was valued as an end in itself: the shape of atoms; the Copernican theory of the solar system; and, as Sapolsky details, countless examples from biology. In one instance Sapolsky documents how it was discovered that neurons don’t actually touch each other, and the debate was about investigating this reality rather than writing 20 page papers proving that they behave ‘as if’ they do touch each other. The one counterexample I’ve heard — though I can’t recall where so I’m happy to be informed about this further — is that Euclidean geometry, despite being incorrect, is more effective than non-Euclidean geometry in some engineering and architecture. But this is rare and again, conditional on strong predictive power (which is lacking in much of economics).

Generally, theories which are based on how things work also tend to have better predictive power, which logically follows from the fact that a theory of the world which looks more like the world is going to be better at predicting the behaviour of the world. My favourite example of this is the priority heuristic from psychology, which is both more behaviourally plausible and predictively accurate theory than standard economic models of decision under risk. There is a whole class of such models, where agents don’t consult a well-defined value function, in the psychology literature.

The need for science

I’ve long complained that much of modern economics is what Piero Sraffa called “cracking a nut with a sledgehammer”: that is, extended mathematical derivations and fancy statistical techniques/research designs are used to ‘prove’ things that are completely uncontroversial. Examples include: capitalist firms sometimes rip you off; identity matters for peoples’ decisions; lack of information about a product creates problems for markets. And that’s just the work of George Akerlof.

One of the things I like about Sapolsky is that he retains perspective and humility when discussing the political implications of neuroscience. He summarises research on how war traumatises the brain and its implications for how we treat veterans. But he is careful to point out that we didn’t “need” neuroscience to “prove” they are traumatised: it’s obvious to almost everyone with a passing interest in the matter. Helping vets should not be conditional on getting the science right first, although it may improve our existing treatment.

I also couldn’t help but imagine the converse: suppose neuroscience “proved” that nothing actually changed in veteran’s brains after war. Would we conclude that they didn’t really have a problem? Of course not! In fact, we’d be more likely to comment on the limitations of neuroscience itself. A related trend that concerns me in economics is the conceit that we need Randomised Control Trials — or at least ‘plausible exogeneity’ to isolate causal effects, engage in cost-benefit analysis and therefore design policy. When it comes to the realm of ‘giving the poorest people on the planet things which are obviously good’, this becomes an unacceptable case of putting (questionable) science before ethics. Researchers’ desire to find a shiny causal effect, especially given the typically large budget of these RCTs, should not be put before the concrete needs of the poor.

Academia is Objectively Pathetic

As somebody who most certainly has taken things too far in low-stakes and esoteric academic debates, it’s funny to see these debates play out in a different field and realise just how pathetic and egotistical (mostly male) academics can be. One case recounted by Sapolsky had a senior academic declare to an (in turns out, pioneering) researcher that “[his findings] may look like neurons in New Mexico, but they don’t in New Haven”. With perspective one can only sigh at the arrogance and pomposity of this statement.

This is one reason why I’ve taken a back seat in the intra-left MMT debate recently because, despite appearances, the stakes are really quite low — we all want roughly the same thing. Furthermore, one of the things that frustrates about economics is everyone’s — mainstream or heterodox — unwillingness to accept pluralism. There are a lot of reasonable theories out there, that’s the nature of the subject.

Epigenetics, Social Statistics, IQ and All That

Speaking of low stakes internet arguments, I recently tweeted that Quillette is full of “racist pseudoscience” because I am increasingly convinced that methods which attempt to ascribe complex phenomena like IQ or preferences to traits like race are pseudoscience. As Sapolsky documents, in a similar but less scathing fashion than Cosma Shalizi, key concepts like ‘heritability’ don’t actually measure the extent to which traits are the product of genes rather than environment. Instead, they are statistical artefacts of said environment, so using them to discern between environment and genes and make causal claims regarding one or the other is bunk. I agree with Stephen Jay Gould in that I am happy for IQ tests to be one way of identifying intellectually gifted — or disabled — individuals for the purposes of education and health, but not for much beyond that.

A major theme of Behave is that environment and genes simply cannot be separated, both due to limitations on our knowledge but also because they are intrinsically linked. Some genes only activate if someone has been exposed to childhood abuse; testosterone only fosters aggression if aggression is necessary to attain or maintain social status; in a more widely known example, diabetes doesn’t kill you, the combination of diabetes and certain foods does. In none of these examples can you isolate the influence of environment versus genes and nor does it even make sense to speak of them separately.

These types of interactions are increasingly grouped under the banner ‘epigenetics’ which has been proposed as the scientific link between history and genes. When dealing with historic environmental influences as persistent, horrific and multifaceted as the historic treatment of black people in the United States the potential for epigenetic influences cannot be ignored. Something similar goes for differences between men and women, which simply cannot be attributed to pure genetic influences given the long history of culturally defined gender roles.

Again, Sapolsky is excellent on the humility that should be induced by science: we simply do not know that much about the human body, and current techniques are unable to isolate the overwhelming majority of genes. The more we find out the more mistaken we realise we were — it’s only 50 years since we discovered epilepsy, and previously an epileptic person would have been tried as a criminal for, say, crashing a car during a fit. In one of the final chapters Sapolsky uses this point (as well as witch hunts — another past ‘crime’) to suggest huge revisions to how we view free will and what this implies for the criminal justice system. Because his proposals are not much fleshed out beyond these principles I can’t say whether they’re a great idea, but I’d like to see more.

Virtue Ethics FTW

I’ve long been strangely sympathetic to the antiquated moral theory of virtue ethics, which emphasises moral character as opposed to rules or consequentialism. One advantage of virtue ethics is that unlike consequentialism or deontology it is an actually practicable theory of moral ethics. Promoting certain values (honesty, reliability, kindness) into people is basically the only way humanity has managed to instill morality into everyday behaviour and transmit it across generations.

Sapolsky notes there is an unexpected neuroscientific basis for virtue ethics: if somebody accepts a set of values about what it means to do ‘good’ then they will behave according to them using the more automatic parts of the brain. For example, people who engage in heroic acts often report doing so unthinkingly — just because it was ‘the right thing to do’. Consequentalism and deontology tend to involve more considered thinking and deliberate restraint (respectively), both of which engage the slower acting prefrontal cortex which makes following them more cognitively demanding and ultimately less likely.

Moreover, Sapolsky seems to imply another point I’ve thought myself: virtue ethics is an indirect way we have evolved to deal with the tricky details of consequentialism (long vs short-term, unpredictability, slippery slopes). Virtue ethics rehabilitates the undeniable appeal of consequentialism to some extent, while avoiding any ‘murder 10 people to improve the lives of 10,000’ implications. Thus, there are good theoretical and practical reasons to favour virtue ethics.

Pinker on Violence

Sapolsky is inherently good natured but in my opinion he quietly demolishes Stephen Pinker’s famous ‘decline of violence’ thesis. He notes that most of the examples of high violence among ancient tribes are flawed, often because such instances are almost always related to modernity and contact with non-tribes. Even more alarmingly, Sapolsky points out that in his ranking of the most violent events in history Pinker categorises both “twelve centuries of the Mideast Slave trade” and 6 years of World War 2 as ‘events’. Once you correct for time, modern wars and massacres are prominently featured at the top of the list of the most violent events in history.

I’ve not read Pinker’s books (despite owning two of them) so YMMV. But this, in combination with Pinker’s cluelessness about the poverty data he makes such extensive use of; cherry picking of environment statistics; bizarre misinterpretations of art and literature; and bad-natured responses to people who point these things out, means I’m no longer inclined to bother.

Dodgy Research

I said this wasn’t a review but I should note a couple of issues with the book. It was written a few years ago and things move fast these days so I don’t really blame him, but he does use some dodgy research in places: the bookdraws heavily from the notoriously underpowered and hard to replicate ‘priming’ studies in psychology (eg tell people they’re poor and see how they do on a test). He also repeats Stephen Levitt’s discredited abortion and crime thesis (spoiler: it’s all about the lead). No doubt there are other examples throughout, but the book is strong and even-handed enough that I doubt its main messages would be affected by a revamp.

Next time: epistemology and economics (maybe)