Monday, January 07, 2013

Steven Pinker: The Better Angels of Our Nature

Steven Pinker's The Better Angels of our Nature argues that violence has been declining over the last several centuries, and continues to decline, even though the common wisdom seems to say the opposite. Pinker marshalls an enormous quantity of data to buttress his story, and fills in with enough explanation to make a very convincing case. At the end, he tries to explain this long term trend, and comes up with several mechanisms, but since this section is less data driven, it's less convincing than the fact that the change is broad, pervasive, and has continued for a very long time.

Pinker starts out with wars, genocide, and other large scale killings. He amasses a dataset of all known mass killings before the modern age. The data is spotty, so it's hard to draw many conclusions, but it is clear that the distant past included its share of conflicts resulting in lots of deaths, and the twentieth century's war are memorable more for their recency than their scale. In addition, there's a pretty clear trend that the great power conflicts of the early twentieth have disappeared. At the end of his chapter on the Long Peace, Pinker points out that since the end of WWII there have been zero:

  1. nuclear weapons used
  2. battlefield fights between great powers
  3. armies crossing the Rhine (longest interval since 200 BCE)
  4. wars between european states
  5. wars between developed countries anywhere in the world
  6. territorial expansions by conquest for a developed country
As we're getting close to 50 or 60 years without a conflict between major powers, it becomes more plausible that it's a trend rather than an aberration.

From large-scale conflicts, Pinker moves on to socially-approved violence, and then to individual violence. Socially-approved violence includes things like slavery and wide-spread repression as of jews and gays as well as public execution and public torture and punishment. All these have gone from common to unacceptable over the long term. Pinker shows that, in parallel with granting rights to more and more groups the statistics on personal violence in a very broad range of contexts have been declining. Historic attitudes toward blacks, women, gays, ethnic groups, children, and animals have all changed dramatically.

Finally, Pinker tries to figure out what's been driving this change. He starts out by describing some broad trends: The Long Peace, and The Rights Revolution, but he admits they're just names, not a description of causes. From there he looks for factors that could have caused these trends. Empathy may have been increased because of the spread of literacy and mass entertainment that give us more access to other points of view. Self control, likewise may have been improved by the promulgation of personal habits that enable people to make their short-term desires subservient to their longer-range goals. He considers biological evolution, but concludes that while it would have been capable of producing a change, we don't have any evidence for the hypothesis. Next Pinker discusses whether humanity's moral sense or rationality has improved in some way to produce the improvement. He accepts that people are getting smarter (i.e. "the Flynn effect") and argues that once we reach a certain level, we can use reason to see that cooperation is more in our interest than violence, first at the personal level, and gradually at broader levels on interaction.

Finally, Pinker presents a framework for thinking about how various changes have impacted people's incentive structures, and what consequences they have for interactions. It's all based on the basic prisoner's dilemma payoff matrix, showing the options two parties face when they can make independent choices as they interact. The first version is called the Pacifist's dilemma, and shows that wars and fights are costly, but it's better to be the aggressor than the defeated. A second shows that "Leviathan" (a government) can change everyone's incentives by penalizing agression. Next is a chart showing that trade ("Gentle Commerce") improves things for everyone by improving the payoffs as long as agression is avoided. His final chart assumes that empathy and reason are added in, and everyone feels not only their own gains and losses, but those of the other party as well. At that point only positive sum outcomes make sense, since each player gains no advantage by imposing costs that are felt by both sides.

This model is plausible, but not compelling. Something like this might be going on, but it's hard to say that he's actually found the mechanism driving things. An interesting postscript is provided by Nassim Nicholas Taleb, who argues that Pinker doesn't understand the statistics of the fat tail that Taleb has been writing about for a while. If the proper curve is not a normal distribution, but instead a fat-tailed curve with most of the weight in the extremes, then the stats demonstrating that there is an effect to be explained are worthless. Taleb gives a bunch of (not well-explained) possible mechanisms for supposing that Pinker might have missed something, but he doesn't analyze whether the historical data looks more like a normal distribution or a fat-tail distribution. I suspect that the near-constant level of violence in the past makes Pinker's position more believable that violence has in fact gone down. It's possible that large-scale conflicts will occasionally arise with enormous body counts, but the drop in violence on all lesser scales doesn't seem likely to be reversed, and that doesn't seem consistent with Taleb's models of financial system variability.

Overall, the book provides good news, and more fodder I can use to try to convince people that what appears in the news is exceptions rather than trends.