I recently watched a presentation by Steven Pinker on the history of violence on YouTube. He makes the case that over the course of history, human beings have been steadily becoming less and less violent, contrary to popular claims that the opposite is true. He presents evidence to support his case and, on a macro-scale, I didn't find his presentation to be very controversial. It syncs with most people's intuitive sense of people from earlier times being uncivilized savages and steadily becoming more civilized over time. And the evidence supports this conclusion on large time scales, specifically on the millennial scale. In order to illustrate this fact he produces a graph that compares the likelihood of a man being killed by another man's hand in Europe or the United States in the 20th century--a notoriously violent century--with other 'societies' from earlier times.
And while I agree with his conclusion that on the millennial scale we have become much less violent, there is an inherent fallacy in this chart; he is not comparing apples to apples. Selecting the most violent enclaves of humanity at certain times in the past to compare to the entirety of Western Civilization--which spans nearly half the globe--is a comparison that tells us nothing about the relative levels of violence between time periods because you're comparing unalike entities in their respective periods. An interesting, relevant comparison might be between the most violent enclaves within 20th Century Europe and the United States--like maybe Compton and Auschwitz--with those earlier enclaves of humanity. The earlier savage 'societies' might still prove to be more brutal, but it would be a more honest and illustrative comparison at least.
None of which is to contest his general conclusion on the millennial-scale, but this illegitimate comparison bleeds over into his case for the century scale, which is much less convincing. In his presentation on the century scale he notes the gradual, consistent decrease over time in socially sanctioned forms of violence. And while this is certainly true, it is immediately glaring that he should choose to focus on one type of violence while omitting other types on the century scale. Why would he do this? Because on the century scale, if we start at the Enlightenment like he chooses to do (bizarrely), we see the decrease in all types of violence until the 20th century. Where only 'socially sanctioned' forms of violence and 'one-on-one murder' continued to decrease, but an unprecedented level of mass murder occurred, creating what could rightly be considered a whole new category of violence, representing a quick and drastic reversal of the general downward trend of total violence. But if you omit this monstrosity of the 20th century, then yes, you can make it seem as if the downward trend in violence continues uninterrupted on the century scale.
The graph he presented for his millennial case could also be adapted for the century scale, and made to be interesting and relevant by comparing 20th century Europe and the U.S. to 19th, 18th, 17th et. al. century Europe and the U.S. This would tell us something about the 20th century specifically and whether or not it was an anomoly in the general downward trend of violence, or whether it too was actually a century of decreased violence. The latter seems to be an utterly incredible, indefensible claim, and that is likely why Pinker chooses not to make it. And I suspect that such a graph would hardly congeal well with his thesis, with levels of violence not changing much for a few centuries after the Enlightenment, or trending slightly downward, and then jumping sharply upward.
His presentation seems to suggest that he believes not only that humanity trends--generally, over the long run--towards less violence, but that there are no blips along the way in this downward trend. That the trend downward is steady and virtually unceasing, without interruption. But the evidence of the 20th century defies this suggestion, and his misconstrual and avoidance of the pertinent comparisons and data speaks as loudly as the data he actually does present.