Science is a process of constructive disagreement. Alternative hypotheses make different predictions about the world, which can be tested against the empirical evidence. The results inform a new round of alternative hypotheses, resulting in an accumulation of knowledge durable enough to be called “facts” – at least until the next turn of the scientific wheel.
Darwin’s theory of evolution is a general explanation of how living things came to be as they are today and it suggests many ideas about the nature and purpose of specific characteristics. Our mission in EVOLUTION:THIS VIEW OF LIFE is to report some of the ideas being generated and the research being done to support or refute them. Often, we will just report evolution-related stories without much analysis. We know that the material that we aggregate from the internet is especially likely to include speculations and preliminary results that will eventually be judged false. Our policy is to be fairly relaxed about what we report. Not only do we lack the capacity to vet each and every story, but in many cases an effective vetting requires more turns of the scientific wheel.
For selected cases, however, we aim to review issues more thoroughly—as well or even better than the academic peer review process, in the open for all readers to see and to join by providing comments of their own. Such a case has arisen in the form of an article published in the scientific journal PLoS ONE on evolutionary policing theory. This is definitely the sort of work we should be reporting. The article suggests that by applying evolutionary theory we will gain an understanding of human behavior that might allow us to make communities safer, people more law abiding and reduce the cost of policing. But there are complications…
What is evolutionary policing theory? The question of how groups of individuals work together to function as adaptive units is a central problem in evolutionary theory. We see this all the time in humans. A platoon of soldiers fights together to defeat another platoon of soldiers. A group of neighbors spends a Saturday picking up litter on their street. It’s a problem to explain because benefitting others and one’s group as a whole typically involves costs to the individual. Group members who act in ways that benefit the group are at a disadvantage compared to members who act more selfishly. The soldiers who fight bravely are less likely to survive than the ones who cower in the background. The neighbor who pretends he has a cold will get to watch the football game on television.
A number of factors contribute to the evolution of behaviors that benefit others. One is genetic relatedness. Mothers take risks and make sacrifices to help their children, Brothers and sisters help one another (well … they often do). But this, of course, doesn’t explain the behavior of the soldiers or the neighbors. To explain this, we must consider other factors, including how the behavior is rewarded or punished by others, in addition to the costs and benefits associated with the behavior itself. Other individuals might reward bravery and punish cowardice, for example. The bad neighbor who watched the football game may find that litter tends to collect on his front lawn. The term “policing” refers to social control mechanisms that loosely bear a resemblance to policing in human terms. It is not just humans who punish group members who behave selfishly. For example, honeybee workers benefit the hive by caring for the eggs of the queen. Workers who get out of line and lay their own eggs are attacked by other workers and the eggs are eaten.
Policing is found in virtually all non-human species that function well as groups. It is even found inside the human body. Our bodies are groups of billions of individual cells working together to benefit the whole organism. With every cell division, mutant cells can arise that stop behaving for the good of the group and to start to reproduce in an uncontrolled way–cancer. One of the most important functions of our immune system is to police cancerous cells. This is why people with poorly functioning immune systems have a higher risk of cancer.
Genetic relatedness is seldom sufficient by itself to explain cooperation between individuals in a group. In the typical honeybee colony, all workers have the same mother, although they can have different fathers. Even this high degree of relatedness must be supplemented by policing, as we have just seen.
So what about humans? How did we evolve to be so cooperative? Extrapolating from our closest primate relatives, current hunter-gatherer societies, and archeological and paleontological evidence, the ancestral human social environment probably consisted of a mix of relatives and nonrelatives who coexisted in a fluid multi-group environment. Relatedness within groups was almost certainly much lower than for a honeybee colony, which means that policing was important from the very beginning of human social evolution. An idea put forward by Christopher Boehm in his book Hierarchy in the Forest, is that humans became so different from other primates because we evolved the ability to suppress fitness differences within groups (especially male dominance), so that succeeding as a group became the primary evolutionary force. In her book, Mothers and Others, Sarah Hrdy reviews the evidence of human cooperation in the raising of young and points to ways that this would have influenced human evolution. According to moral philosopher Richard Joyce in his book The Evolution of Morality, the spontaneous desire to help relatives doesn’t even come close to describing the nature of human morality, which is all about the obligation to help others and society as a whole, even when you don’t want to. Obligations must often be enforced by policing.
As human societies became larger with the advent of agriculture, the psychological policing mechanisms that evolved by genetic evolution were supplemented with other policing mechanisms that evolved by cultural evolution. In Darwin’s Cathedral, one of us (DSW) describes how the religion that became known as Calvinism caused the city of Geneva (then only about 12,000 people) to function better as a unit in the 1500’s. The history of nations such as Switzerland provides a detailed fossil record of the cultural evolution of smaller social units—which were already large by hunter-gatherer standards—coalescing into still larger units. Coalescence is not inevitable—empires fall in addition to rise, as Peter Turchin chronicles in his book War and Peace and War. And other regions of the world find it difficult to rise above the tribal level of social organization. An Evolution Institute workshop organized by Peter Turchin (who is vice president of the EI) will examine nation building and failed states from a multilevel biocultural evolutionary perspective in detail.
Against this background, we can examine the article published in PLoS ONE, by Rolf Kummerli, titled “A Test of Evolutionary Policing Theory with Data from Human Societies”. It will be helpful to describe Kummerli’s empirical study in a theory-free fashion before relating it to the evolutionary concepts outlined above. Working with publicly available information for the 26 cantons that make up Switzerland, Kummerli correlated three variables: a) crime rates; b) per capita monetary investment into policing; and c) an index of “similarity” that reflects both the size of a canton (number of citizens) and the number of foreigners.
Now for the results: The cantons with the lowest “similarity” (i.e., the cantons with the most citizens and most foreigners) have the highest crime rates and invest the most in policing. A more detailed longitudinal analysis demonstrates that within cantons, investing in policing succeeds in reducing crime rates.
These are interesting facts as far as they go, but how can they be related to evolutionary policing theory? Here’s where we think the problems begin. Kummerli makes the following two assumptions:
1. “Crime” can be considered equivalent to “defection” in a simple game theory model such as the prisoner’s dilemma. In other words, everyone has an equal opportunity to cooperate and those who commit crimes have elected to cheat.
2. The “Similarity” index is a proxy for genetic relatedness. Kummerli isn’t claiming that people in the most “similar” cantons are actually more genetically related to each other, but rather that they psychologically perceive that they are. In his words: “The reasoning here is that humans, although today mostly living in societies where relatedness is low, have likely evolved the ability to respond to cues of relatedness in the past, when cooperative interactions occurred in much smaller societies and probably preferentially among related individuals. It is likely that people have retained the ability to respond to these cues, irrespective of the current adaptive consequences.”
We regard both of these assumptions as highly problematic. The idea that all crimes are examples of cheating by people who are in a position to cooperate is a gross oversimplification, to say the least. And the idea that cues of phenotypic similarity are sufficient to get us to cooperate is also a gross oversimplification, even in the ancestral human social environment. Humans have evolved to be much more discriminating than that.
We’re especially bothered by the fact that Kummerli relied on such a narrow slice of evolutionary theory to inform his predictions. Evolutionary theory suggests other explanations of these results that aren’t so problematic.
If we had been asked to review Kummerli’s paper for PLoS ONE, we would have recommended rejection. We don’t want the reader to take our word for it, however. We have therefore asked a number of highly qualified evolutionists to comment on the paper. It is important to stress that none of us have any animosity toward the author or the subject. We regard the subject as important and we are merely acting in our capacity as scientists, evaluating the empirical claims and theoretical interpretations of a colleague. This is what goes on in the academic literature all the time, which we are showcasing for the general public.
We have asked our reviewers to keep jargon to a minimum and they have also divided labor to concentrate on different aspects of the paper. In other respects, the reviews are comparable to the academic peer review process. In both cases, there is a possibility of bias in the choice of reviewers. In our case, we welcome additional reviews in the comments section of this article, including by Kummerli or those who reviewed his article for PLoS ONE. That’s what science as a process of constructive disagreement is all about.
We close with a few observations on the scientific process and how it is reported in the popular media. PLoS ONE is a highly respected scientific journal with the laudable purpose of open access–making its articles available online without charge. Yet, its peer review process did not prevent a highly problematic article such as this one from being published. This means that the peer review process is far from perfect. Just because an article is published in a highly respected scientific journal doesn’t mean that it can be accepted as true. It must be examined on its own terms, as we have done here with the help of some of our colleagues. If other qualified people think that we have judged the article unfairly, we welcome them to join the fray. We’d like nothing better than to provoke a swarm of attention by qualified people, resulting in a new round of hypotheses to test for another turn of the scientific wheel. (Note: if you want to comment, be sure to do so on the EVOLUTION: THIS VIEW OF LIFE website, rather–or in addition to–the other websites that will be carrying this article).
This is important because even though scientific journals are usually discerning about what they publish, the popular media is not so discerning – and the “blogosphere” is not discerning at all. Kummerli’s article will undoubtedly be treated as “fact” by many people, just because it has the imprimatur of science. A number of people will, unfortunately, value this work not because they have been convinced by the evidence but because they believe it supports their own views about immigration and immigrants. Everyone should know that the ideas considered by scientists begin as speculations and that many turns of the scientific wheel are required to hone them into “facts”.
None of the material reported in EVOLUTION:THIS VIEW OF LIFE or any other source should be accepted on faith, but must be evaluated on a case by case basis. That’s what science is all about, and what we’re pleased to report in detail for selected issues such as this one.
Peter Turchin (Vice President, Evolution Institute, and Professor, Dept. of Ecology and Evolutionary Biology, University of Connecticut): I applaud Kümmerli’s application of evolutionary thinking to human societies, and his brave attempt to test theories empirically. However, the execution falls far short of the declared goal.
The main problem is with the operationalization of the two key concepts of the theory: cooperation and relatedness. Unlike Wilson and Newson, I am not quite as ready to dismiss the idea of using crime as a proxy for cooperation. Some recent research (for example, see Randall Roth’s recent book American Homicide) shows that rising and falling homicides correlated with such factors as the degree with which people identify with members of their communities. Nevertheless, any proxy should be used much more carefully than the author does.
The most serious problem is the equation of the proportion of foreigners in a canton with the average genetic relatedness. This kind of naïve biologizing really hurts our attempts to bring evolutionary thinking to social sciences. It smacks of initial – and crude – attempts by sociobiologists to explain human sociality with the theories of kin selection and reciprocal altruism. In fact, on page 6 of the article the author explicitly considers only these two alternatives as possible theoretical frameworks for understanding the results. But evolutionary social science made great strides since the 1970s, and we now have much better – both conceptually and empirically tested – theories, first and foremost cultural group (or multilevel) selection advanced by Peter Richerson, Robert Boyd, David Wilson, and others.
To me this issue – equation of the proportion of foreigners with genetic relatedness – is the fatal flaw that invalidates everything else in the paper. If I were a reviewer of this manuscript I would also take issue with the more technical aspects of the analysis. Why combine canton population size and percentage of foreigners in one measure? What is the reason for using the particular – and peculiar – form (the sum of log-transformed population and untransformed percent)?
Stripping away the theoretical framework that the author imposes on the data, the one interesting and statistically robust result is simply the high correlation between the proportion of foreigners and crime rates. The relationship is very strong (see the figure). Interestingly, canton population size appears to add little, if anything to helping explain the variation in crime rates among cantons. Why this occurs is a very interesting question, but the author’s answer, employing evolutionary policing theory, is completely unpersuasive.
Figure: the correlation between the crime rate (number of registered crimes per 1,000 of canton population) and the percentage of foreigners in the canton. Data from Table 1 in Kümmerli (2011).
Peter Richerson (Distinguished Professor Emeritus, Dept. of Environmental Science and Policy, UC Davis): This author is looking at a very complex problem – and has used a very complex (and yet incomplete) set of data to try to explain differences in the categories. He has proposed a simple hypothesis based on evolutionary policing theory, which makes certain predictions. He reports correlations which match those predictions. This, unfortunately, is a pattern commonly found in papers which attempt to explain human behavior from an evolutionary perspective. If I were reviewing the paper for publication, I would recommend that the editor ask the author to make substantial revisions before acceptance.
A much more complete analysis should be performed in order to compare the model suggested by this hypothesis with models suggested by other hypotheses. May I suggest that the author consult the highly cited book, Model Selection and Multi-Model Inference by Burnham and Anderson for advice on this?
The author of this paper appears to be aware of the work criminologists because he cites papers by criminologists. Criminologists have many plausible ideas about crime, which also suggest models. Most of them are what I call ecological explanations – factors in the social or physical environment (such as inequality or urban anonymity) cause crime. Of course, from an evolutionary perspective such ecological factors are plausibly components of evolutionary explanations as well. Criminology is a hard field because many of these ecological variables are correlated with one another. Ethnic diversity is likely to be correlated with degree of urbanization for example. If this paper had been submitted to a good criminology journal, reviewers would probably insist at least acknowledging such alternatives or they would suggest, as I have, that a more sophisticated statistical analysis be performed to see if the ethnic diversity hypothesis has any independent power to explain crime and policing.
Patterns of crime in Switzerland (see http://en.wikipedia.org/wiki/Crime_in_Switzerland) are very familiar. Non-Swiss people often commit more crimes than native Swiss in a pattern that is roughly correlated with the GDP/capita of their countries of origin. This may be correlated with the wealth and earning power of the immigrants themselves. The GDP/capital of a country is correlated with the amount of time that has passed since the population of the country of origin began to adopt modern norms, values and economically important skills. It also correlates with the ethnic distinctiveness of the group relative to Swiss. Sub-Saharan Africans have high rates of crime relative to Swiss, while Austrians, French and Germans have lower rates. Given that most criminal activity is low paying and dangerous, it will attract low-skill individuals who lack the ability to compete for better paying jobs. Fewer Swiss and Germans will be attracted to crime than sub-Saharan Africans. The covariance of “racial” distinctiveness and income differences is liable to be high. Teasing out which effect is more important is hard work. High crime rate groups have proportionately more young males than the Swiss and young males are more likely to commit crime. This, however, appears to explain only a relatively small part of the overall effect.
Ethnocentrism is a component of many theories of crime. In-group norms are often silent about exploiting out-groups or might even encourage such exploitation. For example, Felicia Pratto’s book “Social Dominance” develops a lot of evidence that policing in the US has an element of ethnic domination that in turn results in the evolution of cultures of resistance that glorify criminal behavior (think gangsta rap) that in turn serves to justify rough policing and high rates of incarceration in a sort of sick coevolutionary arms race.
In summary, lots more work to do before this hypothesis is ready for publication.
Daniel O’Brien (Radcliffe Institute for Advanced Study, Harvard University): There is something in Kummerli’s attempt to apply evolutionary policing theory to crime in human societies that is inspiring. It takes a well-reasoned theoretical model and applies it to a field of study that very rarely has been explored from an evolutionary perspective. Such bold attempts at interdisciplinarity can help to reorient the field, but I fear, instead, that the paper tries to reinvent, with the proverbial wheel in this case being the field of criminology. Through decades of empirical study and theoretical debate, criminologists have developed a sophisticated set of tools that are well-suited to explore their topic of interest: methods for data collection, techniques for analysis, and models for interpretation and prediction. The point of introducing Darwinian theory should be to add to this existing tool kit, but Kummerli does not embrace the rich literature dedicated to the study of crime and policing. This weakens the overall impact of the study, while also making it unclear what the findings contribute to existing knowledge. I’d like to briefly list a few items that stand out me in this regard:
What is the geographical scale of crime? This study analyzes crime as a function of cantons, municipalities that have a median population of 200,000 people, and range as large as a million (Zurich). It is virtually a consensus among criminologists, however, that crime is predominantly a function of smaller communities. It is typically studied at the level of Census tracts (pop. ≈ 4,000), and recent work has argued that crime is generally concentrated in “hotspots” that can be as small as a single intersection (Braga, Hureau & Papachristos, 2011; Braga, Papachristos & Hureau, 2010). If diversity were to encourage criminality as a competitive strategy, it would do so at a much finer scale of analysis than the canton.
How to measure similarity? Criminologists, as well as most other fields engaged in neighborhood-based research, would generally agree that diversity is an important characteristic of a community. Most studies operationalize this variable more broadly, however, with a measure of ethnic heterogeneity. The most common approach is an Herfindahl index, which essentially measures the probability that an interaction between two randomly selected members of a community will be cross-ethnic. In America this calculation is based on the categories of black, white, Hispanic, Asian, and sometimes native American or Pacific Islander when appropriate. In Switzerland I imagine that one could do something similar. Estimating the number of immigrants in a community, however, is its own measure that is distinct from diversity.
Why do diversity and crime correlate? As Pete Richerson has pointed out in another review of this paper, diversity and crime are known to correlate, and this correlation may be attributed to one or more of the other factors that often accompany diversity—population density, socioeconomic disadvantage, the anonymity of residential instability. One of the most compelling and well-validated theories on the patterning of crime, the systemic theory, has its own explanation for this correlation (see Shaw & McKay, 1942/1969 for the original formulation). The basic argument is that, before considering the role of public enforcement, communities police themselves in informal ways, using mechanisms that would be familiar to the anthropological arm of evolutionary psychology, like public scolding, shaming, and even direct intervention. When a community’s members are not cohesive enough to make these mechanisms work, crime starts to emerge, particularly in young males. Certain demographic variables make it hard to generate this cohesiveness, one of them being diversity. Proponents of this field would argue that crime is higher in diverse neighborhoods not because disparate ethnicities living side by side are busy competing with each other, but because they will have a difficult time creating a community that will be able to manage public behavior. They have the data to support this case (Sampson & Groves, 1989; Sampson, Raudenbush & Earls, 1997).
I end with this last point because the systemic theory of crime might be a better place for those who want to pursue evolutionary policing to start. It is a sophisticated ecological model that explains the social processes that underlie the generation and prevention of crime. A synergistic combination of evolutionary policing theory with the systemic theory—or any other well-supported theory of crime, for that matter—could be revolutionary in the way that Kummerli’s paper aims to be.
Braga, A. A., Hureau, D. M., & Papachristos, A. V. (2011). The relevance of micro places to citywide robbery trents: A longitudinal analysis of robbery incidents at street corners and block faces in boston. Journal of Research in Crime and Delinquency, 48, 7-32.
Braga, A. A., Papachristos, A. V., & Hureau, D. M. (2010). The concentration and stability of gun violence at micro places in boston, 1980-2008. Journal of Quantitative Criminology, 26, 33-53.
Sampson, R. J., & Groves, W. B. (1989). Community structure and crime: Testing social disorganization theory. The American Journal of Sociology, 94, 774-802.
Sampson, R. J., Raudenbush, S. W., & Earls, F. (1997). Neighborhoods and violent crime: A multilevel study of collective efficacy. Science, 277, 918-924.
Shaw, C., & McKay, H. (1942/1969). Juvenile delinquency and urban areas. Chicago: University of Chicago Press.
Rafe Sagarin (ENVIRONMENT Editor and Assistant Research Scientist, Institute of the Environment, U. Arizona): I share most of the concerns of my colleagues with the Kummerli paper, but the focus of my critique is more pointedly at the brief section at the end where the author attempts to put any kind of qualification on his conclusions. He notes, “comparative approaches are based on correlational analysis, which preclude making firm conclusions on the causalities between correlating variables.” This is a slight spin on the tired old platitude, “correlation does not imply causation” which has been reinforced by the purveyors of the kind of reductionist experimental approaches that dominated the life sciences in the 20th century. Yet this canard has been refuted time and again by some of the most influential works of science. For example, the robust theory that a large extraterrestrial impact was the cause of the K-T extinction was built on a series of elegant correlations. The view that correlational studies are unable to isolate underlying mechanisms will become increasingly inaccurate as larger data sets across more spatial and temporal scales are brought to bear on complex 21st century challenges at the social-ecological nexus.
At the same time, because we are now breaking the barriers between life sciences (which were once safely confined to the laboratory or scientific field preserve) and the world of human affairs, the bar for correlational studies must be set very high. Multiple hypotheses must be tested using multiple layers of data taken at multiple scales and analyzed in (appropriately enough) a multi-variate framework. This paper falls short not, as the author would like us to believe, because it is merely correlational, but because it is not correlational enough.
The correlations he produces fail all of the high bar tests that a complex world requires of us. At each level of this paper there is an opportunity to enter the world of multiple causality, but the author doesn’t go there. Using only percentage of immigrants as a positively correlated measure for relatedness, for example, fails to capture the demographics, economics, and social pressures that may lead to high or low relatedness in a given geography. Using only a bivariate relationship between (supposed) relatedness and policing likewise ignores multiple drivers of increased vigilance. And using only a bivariate relationship between policing and crime is a disservice to the existence of an extensive, multi-disciplinary, and continually developing science of crime.
These multiple disconnections between analysis and reality, forced by flattening the complexity so severely, can be most clearly seen in the interpretation of the relationship between crime and between-year changes in policing effort. The author claims that because simple regression models show a (barely) negative and (barely) significant result, the third indicator of evolutionary policing theory is identified and thus he “found full support of all three predictions”. A careful look at Figure 1d, where these data are plotted, shows that the curve is attracted by the tension between two primary forces: first, by one outlying point in which a huge expenditure in policing led to no change in crime (violating the author’s prediction) and second, by the vast majority of the remaining points, lined up nearly vertically, which show that small expenditures in policing lead to either a decrease, an increase, or no change in crime. For the small subset of six places that decreased spending on policing, it appears from the data that two found decreases in crime, two found increases, and two showed virtually no change. Yet the author eschews this observable interpretation of his data in favor of the statistical interpretations.
Ultimately the failure of this paper is not wholly the author’s. I believe his work is an admirable first foray into what should be a series of tests that might get at the core questions of evolutionary policing theory. The failing is in the peer review process. It is very disturbing to see in the acknowledgments “an anonymous referee” thanked, suggesting that either only one peer reviewer was used, or that others gave their comments but the author and editors chose to ignore them (and thus they weren’t thanked in the acknowledgments). I suspect the case is the former, as other authors of papers published in PLoS ONE have told me they only had one outside reviewer on their papers. If so, this is a disturbing finding. If there are faith-based beliefs in the scientific community, the most commonly held is a faith in the peer review process. Yet even in my own reviews for over a dozen different journals I have been amazed and humbled at the insights provided by my fellow reviewers that I did not come to myself. In other words, it is essential to have more than one independent peer reviewer on every paper. PLoS ONE has done a great service to the scientific community by producing a top quality open access journal with a quick turnaround time that respects the multi-disciplinary nature of 21st century science. But if their published products erode our faith in the peer review process, the status and usefulness of this journal will likewise erode.