Four perspectives on human decision making

I have been rereading Gerd Gigerenzer’s collection of essays Rationality for Mortals: How People Cope with Uncertainty. It covers most of Gigerenzer’s typical turf – ecological rationality, heuristics that make us smart, understanding risk and so on.

In the first essay, Gigerenzer provides four categories of approaches to analysing decision making – unbounded rationality, optimisation under constraints, cognitive illusions (heuristics and biases) and ecological rationality. At the end of this post, I’ll propose a fifth.

1. Unbounded rationality

Unbounded rationality is the territory of neoclassical economics. Omniscient and omnipotent people optimise. They are omniscient in that they can see the future – or at least live in a world of risk where they can assign probabilities. They are omnipotent in that they have all the calculating power they need to make perfect decisions. And with that foresight and power, they make optimal decisions.

Possibly the most important point about this model is that it is not designed to describe precisely how people make decisions, but rather to predict behaviour. And in many dimensions, it does quite well.

2. Optimisation under constraints

In this approach, people are no longer omniscient. They need to search for information. As Gigerenzer points out, however, this attempt to inject realism creates another problem. Optimisation with constraints can be even harder to solve than optimisation with unbounded rationality. As a result, the cognitive power required is even greater.

Gigerenzer is adamant that optimisation under constraints is not bounded rationality – and if we use Herbert Simon’s definition of the term, I would agree – but analysis of this type commonly attracts the “boundedly rational” label. Gigerenzer’s does not want the unrealistic nature of optimisation under constraints to tar the concept of bounded rationality.

3. Cognitive illusions – logical irrationality

The next category is the approach in much of  the behavioural sciences and behavioural economics. It is often labelled as the “heuristics and biases” program. This program looks to understand the processes under which people make judgements, and in many cases, seeks to show errors of judgment or cognitive illusions. This program has generated a long list of biases – just look at the Wikipedia page for a taste.

Gigerenzer picks two main shortcomings of this approach. First, although the program successfully shows failures of logic, it does not look at the underlying norms. Second, it tends not to produce testable theories of heuristics. As Gigerenzer states, “mere verbal labels for heuristics can be used post hoc to “explain” almost everything.”

An example is analysis of overconfidence bias. People are asked a question such as “Which city is farther north – New York or Rome?”, and asked to give their confidence that their answer is correct. When participants are 100 per cent certain of the answer, less than 100 per cent tend to be correct. That pattern of apparent overconfidence continues through lower probabilities.

There are several critiques of this analysis, but one of the common suggestions is that people are presented with questions that are unrepresentative of a typical sample. People typically use alternative cues to answer a question such as the above. In the case of latitude, temperature is a plausible cue. The overconfidence bias occurs because the selected cities are a biased sample where the cue fails more often than expected. If the cities are randomly sampled from the real world, the overconfidence disappears. The net result is that what appears to be a bias may be better explained by the nature of the environment in which the decision is made.

4. Ecological rationality

Ecological rationality departs from the heuristics and biases program by examining the relationship between mind and environment, rather than the mind and logic. Human behaviour is shaped by scissors with two blades – the cognitive capabilities of the actor, and the environment. You cannot understand human behaviour without understanding both the capabilities of the decision maker and the environment in which those capabilities are exercised. Gigerenzer would apply the bounded rationality label to this work.

On this basis, there are three goals to the ecological rationality program. The first is to understand the adaptive toolbox – the heuristics of the decision maker and their building blocks. The second is to understand the environmental structures in which different heuristics are successful. The third is to use this analysis to improve decision making through designing better heuristics or changing the environment. This can only be done once you understand the adaptive toolbox and the environments in which different tools are successful.

Gigerenzer provides a neat example of how the ecological rationality departs from the heuristics and biases program in its analysis of a problem – in this case, optimal asset allocation. Harry Markowitz, who received a Nobel Memorial Prize in Economics for his work on optimal asset allocation, did not use the results of his analysis in his own investing. Instead, he invested his money using the 1/N rule – spread your assets equally across N assets.

The heuristics and biases program might look at this behaviour and note Markowitz is not following the optimal behaviour determined by himself. He is making important decisions without using all the available information. Perhaps it is due to cognitive limitations?

As Gigerenzer notes, optimisation is not always the best solution. Where the problem is computationally intractable or the optimisation solution lacks robustness due to estimation errors, heuristics may outperform. In the case of asset allocation, Gigerenzer notes work showing that 500 years of data would have been required for Markowitz’s optimisation rule to outperform his practice of 1/N. In a world of uncertainty, it can be beneficial to leave information on the table. Markowitz was using a simple heuristic for an important decision, but rightfully so as it is superior for the environment in which he is making the decision.

5. Evolutionary rationality

Gigerenzer proposes four categories, but I’ll lay out a fifth (I’m not sure about the label I’ve just given it). Evolutionary rationality develops a deeper understanding of the cognitive capabilities of the decision maker through an analysis of the adaptive basis of traits. This perspective could inform all four of the above categories of decision making. It could be used to assess what is being optimised, what the constraints might be, how biases might be due to mismatch between past and present environments, and what the heuristics are.

Gigerenzer notes the possibility of going into this territory, but deliberately holds back. In the third chapter of the book, he writes:

[H]uman psychologists are not able to utilize many of the lines of evidence that biologists apply to justify that a trait is adaptive. We can make only informed guesses about the environment in which the novel features of human brains evolved, and because most of us grow up in an environment very different to this, the cognitive traits we exhibit might not even have been expressed when our brains were evolving. …

ABC avoids the difficult issue of demonstrating adaptation in humans by defining ecological rationality as the performance, in terms of a given currency, of a given heuristic in a given environment. We emphasize that currency and environment have to be specified before the ecological rationality of a heuristic can be determined; thus, take-the-best is more ecologically rational (both more accurate and frugal) than tallying in noncompensatory environments but not more accurate in compensatory ones. Unlike claiming that a heuristic is an adaptation, a claim that it is ecologically rational deliberately omits any implication that this is why the trait originally evolved, or has current value to the organism, or that either heuristic or environment occurs for real in the present or past. Ecological rationality might then be useful as a term indicating a more attainable intermediate step on the path to a demonstration of adaptation.

There is a lot more interesting material in Chapter 3 on the link between Gigerenzer’s program and the approach taken by biologists. That will be the subject of a later post.

A week of links

Links this week:

  1. Big ideas are destroying international development. Dream smaller.
  2. Appealing to my biases – the skeptics guide to institutions Part 1 and Part 2.
  3. Most published results in finance are false.
  4. Be mean, look smarter.
  5. Constructing illusions.
  6. Predicting complex genotypes from genomic data – for those who confuse these two statements:

“The brain is complex and nonlinear and many genes interact in its construction and operation.”

“Differences in brain performance between two individuals of the same species must be due to nonlinear (non-additive) effects of genes.”

Genetics and education policy

Philip Ball has an article in the December issue of Prospect (ungated on his blog) arguing that consideration of the genetic basis to social problems is a distraction from socioeconomic causes. The strawman punchline for the Prospect article is “It’s delusional to believe that everything can be explained by genetics”.

The article has drawn a response from one of the people named in the article, Dominic Cummings. Ball suggests that Cummings presents “genetics as a fait accompli – if you don’t have the right genes, nothing much will help”, although this statement suggests Ball had not invested much effort getting across Cummings’s actual position (as contained in this now infamous essay). Ball responded in turn, with Cummings firing back (in an update at the bottom of the page), and Ball responding again.

Beyond the tit for tat – read their respective posts for that – there are some interesting points about whether genetics tells us anything about education policy.

As a start, Ball claims that “Social class remains the strongest predictor of educational achievement in the UK”, referencing this article. However, the authors of that article don’t consider the role of genetics or other potential predictors. The references that article gives for the claim are similarly devoid of relevant comparisons, which is unsurprising as they largely comprise policy positioning documents from various organisations. It’s hard to credibly claim something is a superior predictor when it is not assessed against the alternatives.

So, what is the evidence on this point? For one, we have twin and adoption studies. As a sample, Bruce Sacerdote studied Korean adoptees into the United States (admittedly, not the UK as per the quote) and found that shared environment (which would include socioeconomic status) explained 16 per cent of the variation in educational attainment. Genetic factors explained 44 per cent. This is a consistent finding in adoption studies, with children more closely resembling their biological parents than their adopted parents. For twin studies, an Australian analysis found a 57 per cent genetic and 24 per cent shared environment contribution to variation in education. A meta-analysis of heritability estimates of educational attainment found that, in the majority of samples, genetic variation explained more of the variation in educational attainment than shared environment.

Of course, we don’t have the genetic data or understanding at hand just yet, but there are other factors such as IQ that are better predictors of education than social class. This territory is also complicated – there are genetic effects on both IQ and social class – but IQ tends to outperform. This meta-analysis shows that IQ is a better predictor of education, income and occupation than socioeconomic status – not overwhelmingly so, but superior nonetheless.

Then there is the link between genetic factors and socioeconomic status, with a long line of studies finding a relationship. One of the more recent was by Daniel Benjamin and friends (ungated pdf). They found heritability of permanent income (20-year average) of 0.58 for men and 0.46 for women. Part of the predictive power of socioeconomic status comes from its genetic basis. Gregory Clark’s hypothesis of low social mobility being a result of genetic factors reflects this body of work.

Turning next to Ball’s pessimism of the future of genetics, he states:

In September an international consortium led by Daniel Benjamin of Cornell University in New York reported on a search for genes linked to cognitive ability using a new statistical method that overcomes the weaknesses of traditional surveys. The method cross-checks such putative associations against a “proxy phenotype” – a trait that can ‘stand in’ for the one being probed. In this case the proxy for cognitive performance was the number of years that the tens of thousands of test subjects spent in education.

From several intelligence-linked genes claimed in previous work, only three survived this scrutiny. More to the point, those three were able to account for only a tiny fraction of the inheritable differences in IQ. Someone blessed with two copies of all three of the “favourable” gene variants could expect a boost of just 1.8 IQ points relative to someone with none of these variants. As the authors themselves admitted, the three gene variants are “not useful for predicting any particular individual’s performance because the effect sizes are far too small”.

This, however, is only part of the picture. If we look at another study in which Benjamin was involved, three SNPs (single nucleotide polymorphisms – single base changes in the DNA code) were found to affect educational attainment. In total, they explained 0.02 per cent of the variation in educational attainment – practically nothing. But combine all the SNPs in the 100,000 person sample, and you edge up to 2.5 per cent. But even more interesting, they calculated that with a large enough sample they could explain over 20 per cent of the variation. Co-author Philipp Koellinger explains this in a video I recently linked. Although this study found variants with low explanatory power, it also points to the potential to explain much more with larger samples.

For more on the background to the feasibility of identifying the causal genetic variants for traits such as IQ, its worth looking at this paper by Steve Hsu. Possibly the most important point is that the causal variants for traits such as cognitive ability and height are additive in their effect. In his final response, Ball states that “And that might be because we are thinking the wrong way – too linearly – about how many if not most genes actually operate.” But the evidence shows that is how they largely work. Although a few years old now, this paper’s theoretical and empirical argument that genetic effects are largely additive has generally been affirmed in later research. This considerably simplifies the task of predicting outcomes based on someone’s genome. In fact, this is one reason selective breeding has been so successful and genetic data is already being used successfully in cattle breeding (There’s an example of the gap between entrepreneurship and policy development – while some of us are arguing whether this stuff is possible, someone else is already doing it).

Now, supposing you have this genetic data, how might this change education? Returning to the article I linked above (ungated pdf), Benjamin and friends suggested this genetic information could be used to better target interventions. They propose early identification of dyslexia as an example.

They also suggest using genetic data as controls. This could provide more precision in studies of whether interventions to target socioeconomically disadvantaged children are effective. The genetic controls allow you to hone in on what you are interested in. In the question and answer session of a video of talk by Jason Fletcher I recently linked, Benjamin pointed to the famous Perry PreSchool Project and noted that additional precision through the use of genetic data would have been of great value.

Ball also indirectly alludes to another reason to learn about genetic factors. In his last response, he writes:

Personally, I find a little chilling the idea that we might try to predict children’s attainments by reading their genome, and gear their education accordingly – not least because we know (and Plomin would of course acknowledge this) that genes operate in conjunction with their environment, and so whatever genetic hand you have been dealt, its outcomes are contingent on experience.

This argument runs both ways. Supposing there are large gene-environment interactions, how can you understand the effects of changing the environment without looking at the way that environment affects people via their genome? As an example of this, Jason Fletcher examined how variation in a gene changed the response to tobacco taxation policy (he talks about this in a video I recently linked). Those with a certain allele responded to taxation and reduced smoking. Others didn’t. Too be honest, I’m not sold on the results of this particular study, but it illustrates that genetic factors that need to be considered if these gene-environment interactions are as large as people such as Ball believe.

[I should admit at this point that G is for Genes: The Impact of Genetics on Education and Achievement is sitting unread in my reading pile….]

Putting it together, Ball is off track in his suggestion that learning about and targeting genetic factors distracts from dealing with socioeconomic issues. Understanding of genetic and socioeconomic factors are complements, and by disentangling their effects, we could better tailor education to address each.

That is not to say that the genetic enterprise is guaranteed to be successful. But there is plenty of evidence that our genes are relevant and, on that basis, should be considered.

Further, there are changes we can make today. Ball asks what genetics can add beyond recognition that some children are more talented than others. The thing is, much schooling is still structured as though we are blank slates. Maybe it is an understanding of genetics that will finally get us to a point where education is better designed for people with different capacities, improving the experience across the full range of abilities and backgrounds.

The beauty of self interest

In my review of E.O. Wilson’s The Social Conquest of Earth, I quoted this passage which captures Wilson’s conception of the origin of cooperation in humans.

Selection at the individual level tends to create competitiveness and selfish behaviour among group members – in status, mating, and the securing of resources. In opposition, selection between groups tends to create selfless behavior, expressed in greater generosity and altruism, which in turn promote stronger cohesion and strength of the group as a whole.

This passage from Matt Ridley strikes at the heart of Wilson’s dichotomy between selfishness and generosity:

“Group selection” has always been portrayed as a more politically correct idea, implying that there is an evolutionary tendency to general altruism in people. Gene selection has generally seemed to be more of a right-wing idea, in which individuals are at the mercy of the harsh calculus of the genes.

Actually, this folk understanding is about as misleading as it can be. Society is not built on one-sided altruism but on mutually beneficial co-operation.

Nearly all the kind things people do in the world are done in the name of enlightened self-interest. Think of the people who sold you coffee, drove your train, even wrote your newspaper today. They were paid to do so but they did things for you (and you for them). Likewise, gene selection clearly drives the evolution of a co-operative instinct in the human breast, and not just towards close kin.

You can read the full article here.

E.O. Wilson’s The Social Conquest of Earth

The Social Conquest of EarthThe re-eruption of the war of words between E.O. Wilson and Richard Dawkins has occurred just as I have come around to reading Wilson’s 2012 book The Social Conquest of Earth. In an interview on BBC2 (watch it at the bottom of this post), Wilson stated:

There is no dispute between me and Richard Dawkins and there never has been, because he’s a journalist, and journalists are people that report what the scientists have found and the arguments I’ve had have actually been with scientists doing research.

It is an interesting call to authority that Wilson himself challenged in his reply to Dawkins’s stinging review of the book in Prospect magazine (Wilson’s reply is at the bottom of the Dawkins piece) – or even Wilson’s statement at the end of the book that:

Science belongs to everybody. Its constituent parts can be challenged by anybody in the world who has sufficient information to do so.

Regardless, the debate between Wilson and Dawkins is a continuation of the group selection debate that has been running since the 1960s, with Wilson now on the group selection side, and Dawkins on that of the selfish gene. But despite this framing of the debate as a confrontation between two apparently diametrically opposed views, The Social Conquest of Earth suggests that Wilson’s view is somewhat more complicated, and possibly confused.

The old and new group selection

As background, it is worth defining three concepts: group selection, a newer conception of group selection called multilevel selection, and inclusive fitness.

The older form of group selection is a process where the differential survival of groups leads to the evolution of traits that benefit the group. This type of group selection, pushed in the 1960s by V.C Wynne-Edwards in particular, might involve members of a group restraining reproduction during times of scarcity so that the group does not experience resource shortages.

This concept received many harsh critiques, most famously by George Williams and in popular form by Dawkins in The Selfish Gene. The basic problem is that if someone cheats and does not restrain reproduction when others do, they will have more offspring and come to dominate the group. The altruistic trait will only emerge if groups with more altruists have a large enough advantage over other groups to compensate for their disadvantage within their groups. These conditions are generally considered to be met in limited circumstances, and most evolutionary biologists would say that the evolution of group adaptations in this way is a theoretical possibility, occurs in some circumstances, but is a practical rarity.

Group selection was somewhat reinvigorated in the late 1970s by David Sloan Wilson and friends under a reworking that is commonly called multilevel selection. The first distinguishing feature of multilevel selection is that the definition of “group” can include transitory groupings that regularly remix. You could consider two individuals who briefly trade to be a group. The second feature of multilevel selection is that selection is decomposed across multiple levels. The analysis would look at the fitness of the two trading individuals with respect to each other, which is the individual level selection, and of the fitness of their group relative to other groups.

Multilevel selection has received a largely muted response, with inclusive fitness the alternative framework preferred by Dawkins and friends – not to mention the dominant paradigm in evolutionary biology. Inclusive fitness combines the direct effects of a trait on an individual with the indirect effects of the trait on other individuals who possess that trait. Kin selection, a strategy of favouring relatives, maximises inclusive fitness.

Inclusive fitness is famously captured by Hamilton’s rule, which states that an altruistic trait will spread if rb>c. c is the cost to the altruist of the trait, b the benefits to others, and r the relatedness between the altruist and beneficiaries. A trait to favour your brother will spread if the benefits to the brother, who is 0.5 related to you, are double the costs to you. Or as J.B.S. Haldane put it, he would give his life for two brothers or eight cousins.

While apparently opposing perspectives, inclusive fitness and multilevel selection are two sides of the same coin. If you can describe an evolutionary dynamic in terms of multilevel selection, you can also give an inclusive fitness story (many suggest the two approaches are mathematically equivalent, although this is debated). They are simply different accounting methods, or languages. The intuitive explanation for the link is that higher levels of selection (the level of groups) can favour the spread of a trait because the members of that group have a degree of relatedness.

Wilson’s critique of kin selection

Wilson’s core argument through The Social Conquest of Earth is that the concept of inclusive fitness has been discredited. This claim stems from the infamous 2010 Nature paper by Martin Nowak, Corina Tarnita and Wilson on eusociality.

An E.O. Wilson drawn ant on the title page to my book

An E.O. Wilson drawn ant on the title page to my book

Eusociality involves a division of reproductive labour, such as that which occurs in the bees, ants and wasps. Eusociality and kin selection are closely linked as the higher relatedness between sisters in the bees, ants and wasps has been used to explain the willingness of most females to forgo their reproductive success for one of their sisters, the queen.

Nowak, Tarnita and Wilson’s argument was that the evolution of eusociality could be explained through simple individual selection and did not require the framework of inclusive fitness. They presented a model in which eusociality evolved without any reference to relatedness.

The model itself was interesting, but it was sandwiched between a not particularly well thought-out or supported claim that “the production of inclusive fitness theory must be considered meagre” and that it “does not provide additional insight or information” to standard natural selection theory. I will let the many responses to the paper speak for themselves, including the main response (with the 130 odd signatures – ungated version here), which contains a table indicating the contributions of inclusive fitness. But if I were to single one paper out, it is this one by Garnder, West and Wild, which addresses many of the mathematical arguments. Its main point, in short, is that Nowak, Tarnita and Wilson fail to distinguish between general kin selection theory and the kin selection methodology used to address specific problems. Their criticisms do not apply to the general theory.

Coming back to Wilson’s book, however, Wilson seems to take an even stronger stance than in the paper. For example, he states that:

Martin Nowak, Corina Tarnita, and I demonstrated that inclusive-fitness theory, often called kin selection theory, is both mathematically and biologically incorrect.

Through the book, Wilson’s characterisation of the paper’s reception has to be described as either deceptive or oblivious. Gems such as “The beautiful theory [inclusive fitness] never worked anyway, and now it has collapsed” contrasts with what even a cursory glance at the responses suggest. Nowak, Tarnita and Wilson’s critique has not generally been accepted, although reading the book gives no impression of the slightest opposition to Wilson’s position. The interview that triggered this latest spat suggests Wilson is still singing a deceptive tune. He states:

I have abandoned it [the notion of the selfish gene] and I think most serious scientists working on it have abandoned it. Some defenders may be out there, but they have been relatively or almost totally silenced since our major paper came out.

Given the paper, it is no surprise that Wilson argues throughout The Social Conquest of Earth that individual level and group selection is all that is required to explain the evolution of eusociality in insects. Wilson argues that, after the emergence of eusociality in a single colony through individual level selection, “between-colony selection” leads to the wider spread of the eusocial trait. Its selection at the individual and group levels without a multi-level selection framework. As Wilson states:

But multilevel selection, in which colonial evolution is regarded as the interests of the individual worker pitted against the interests of its colony, may no longer be a useful concept on which to build models of genetic evolution is social insects.

I have no idea why the preferred model isn’t simply a multilevel selection framework with alternative assumptions, and the confusion only increases from here.

Eusociality in humans

Where things get truly confusing is Wilson’s consideration of humans. Try as I could, I could not conceive of a sympathetic reading that would allow Wilson’s position to be seen as coherent.

First, his branding of humans as eusocial is a stretch under any definition, although he is not alone in attempting that.

But more confusingly, his argument that eusociality arose in humans due to multilevel selection is hard to understand because I have no clear idea of what he actually means. As a start, its not multilevel selection in the traditional sense, as Wilson has rejected the other side of the multilevel selection coin, inclusive fitness. Initially, I put it down to his error, but when I hit the last chapter, I realised he was using the term “multilevel selection” to mean something different. When Wilson speaks of multilevel selection, he is generally referring to individual level and group selection occurring in tandem, the “groups” being as we would traditionally define them. But then why isn’t his dynamic in eusocial insects multilevel selection under his definition?

Part of my confusion (and initial assumption) also stemmed from the contrast between Wilson’s past statements and what he wrote in the book. Compare these two paragraphs – the first from a 2007 paper co-authored with David Sloan Wilson, and the second from the last chapter of the book.

The theories that were originally regarded as alternatives, such that one might be right and another wrong, are now seen as equivalent in the sense that they all correctly predict what evolves in the total population. They differ, however, in how they partition selection into component vectors along the way.

Theorists of inclusive fitness themselves have argued that kin selection can be translated into group selection, even though that belief has now been disproven mathematically.

Based on this, it seems that E.O. Wilson is no longer on the same page as the number one champion of multi-level selection, David Sloan Wilson. It is particularly strange in that the two Wilsons characterise what multilevel selection means for humans in almost the same way. As E.O. Wilson writes, and I expect David Sloan Wilson would agree:

Selection at the individual level tends to create competitiveness and selfish behaviour among group members – in status, mating, and the securing of resources. In opposition, selection between groups tends to create selfless behavior, expressed in greater generosity and altruism, which in turn promote stronger cohesion and strength of the group as a whole.

E.O. Wilson’s varying use of these terms points to one of the problems group selection has in popular discourse. The term group selection has been used so inconsistently and used to refer to so many different dynamics, it is often hard to know what someone means when they refer to it. This article (ungated pdf) points to four different uses of the term “group selection”, although I have seen some suggestions that there are six different uses in the literature. When people like Wilson present their arguments in such a confusing manner, it is no surprise that others with less expertise are similarly confused. Look at Jonathan Haidt’s confusion of old group and multilevel selection as a prime example.

The other bits

Beyond Wilson’s take on group selection, there are some interesting parts to the book.

One is Wilson’s argument that many examples of kin selection can be explained as pure self interest. For example, he describes how some bird and mammal offspring remain at their parents’ nest. This has been interpreted as an example of kin selection – it helps the bird or mammal’s parents and siblings. However, Wilson suggests direct self interest is at play. In cases of resource or territory scarcity, they remain with the parents to inherit the parents’ nest when the parents fall off the perch. Wilson provides several examples of this type, suggesting that the focus on kin selection clouds the assessment of what is actually occurring.

Funnily enough, these arguments mirror an argument I often make about apparently altruistic acts sought to be explained by multilevel or group selection. Many apparently altruistic acts are self interested, such as the trade that characterises our economies. If you classed two people trading with each other as a group, as you might in a multilevel selection framework, you could class the person who gained the least surplus from the trade as an “altruist”. But the simplest explanation is that they seek to gain from trade.

The final sections of the book seek to explain “who we are”. I can only say that there are better places to read about the evolutionary origins of religion, art or language. While the last chapter of Sociobiology was revolutionary in its application of evolutionary theory to humans, the short snapshots Wilson provides in The Social Conquest of Earth do not do justice to the work that has occurred in the last 30 years. But that large body of work is, of course, one of Wilson’s great legacies. As Dawkins noted, despite The Social Conquest of Earth, Wilson’s place in history is assured.

A week of links

Links this week:

  1. W. Brian Arthur on economic complexity.
  2. A great article on humans as imitators.
  3. Higher latitudes have colder weather which leads to larger people which causes lower population and higher investment in children which triggers economic growth.
  4. An epidemic of over-diagnosis.
  5. Financial price data are converted into music, the music is played to a rat, then the rat guesses whether the price will fall or rise.
  6. Is being good at science a matter of nature?
  7. Women earn less even when they set the pay.
  8. Social and cognitive skills are complements.

Ignorance feels so much like expertise

In the Pacific Standard, David Dunning of the Dunning-Kruger effect writes:

A whole battery of studies conducted by myself and others have confirmed that people who don’t know much about a given set of cognitive, technical, or social skills tend to grossly overestimate their prowess and performance, whether it’s grammar, emotional intelligence, logical reasoning, firearm care and safety, debating, or financial knowledge. College students who hand in exams that will earn them Ds and Fs tend to think their efforts will be worthy of far higher grades; low-performing chess players, bridge players, and medical students, and elderly people applying for a renewed driver’s license, similarly overestimate their competence by a long shot.

But education is not always the answer:

While educating people about evolution can indeed lead them from being uninformed to being well informed, in some stubborn instances it also moves them into the confidently misinformed category. In 2014, Tony Yates and Edmund Marek published a study that tracked the effect of high school biology classes on 536 Oklahoma high school students’ understanding of evolutionary theory. The students were rigorously quizzed on their knowledge of evolution before taking introductory biology, and then again just afterward. Not surprisingly, the students’ confidence in their knowledge of evolutionary theory shot up after instruction, and they endorsed a greater number of accurate statements. So far, so good.

The trouble is that the number of misconceptions the group endorsed also shot up. For example, instruction caused the percentage of students strongly agreeing with the true statement “Evolution cannot cause an organism’s traits to change during its lifetime” to rise from 17 to 20 percent—but it also caused those strongly disagreeing to rise from 16 to 19 percent. In response to the likewise true statement “Variation among individuals is important for evolution to occur,” exposure to instruction produced an increase in strong agreement from 11 to 22 percent, but strong disagreement also rose from nine to 12 percent. Tellingly, the only response that uniformly went down after instruction was “I don’t know.”

The way we traditionally conceive of ignorance—as an absence of knowledge—leads us to think of education as its natural antidote. But education, even when done skillfully, can produce illusory confidence. Here’s a particularly frightful example: Driver’s education courses, particularly those aimed at handling emergency maneuvers, tend to increase, rather than decrease, accident rates. They do so because training people to handle, say, snow and ice leaves them with the lasting impression that they’re permanent experts on the subject. In fact, their skills usually erode rapidly after they leave the course. And so, months or even decades later, they have confidence but little leftover competence when their wheels begin to spin.

In cases like this, the most enlightened approach, as proposed by Swedish researcher Nils Petter Gregersen, may be to avoid teaching such skills at all. Instead of training drivers how to negotiate icy conditions, Gregersen suggests, perhaps classes should just convey their inherent danger—they should scare inexperienced students away from driving in winter conditions in the first place, and leave it at that.

The full article is worth reading.

A week of links

Links this week:

  1. The freedom to pursue informed self-harm has a long and noble tradition.
  2. What happens when behavioural economics is used to explain rational behaviour.
  3. A great summary of some of Gordon Tullock’s work. HT: Garett Jones
  4. Another study on the limited effect of parenting on IQ. HT: Billare via Stuart Ritchie
  5. What Hayek might say to Republicans.
  6. The long shadow of history on the distribution of human capital in Europe. HT: Ben Southwood
  7. Opposition to urban development by “environmentalists” is among my bigger gripes. Left-leaning cities are less affordable.
  8. I have only just come across Dominic Cummings. Some interesting thoughts. Check out his blog.
  9. Affirmative action to overcome liberal bias.
  10. How your brain decides without you.

Genome Wide Association Studies and socioeconomic outcomes

A few months back, I posted about a Conference on Genetics and Behaviour held by the Human Capital and Economic Opportunity Global Working Group at the University of Chicago. In that post, I linked to a series of videos from the first session on the effect of genes on socioeconomic aggregates.

Over the last couple of days, I watched the videos from the session on Genome Wide Association Studies (GWAS). As for the first set of videos, they are technical (as you might expect for a bunch of academics) – particularly the questions – but cover some important points.

In early studies linking genetic factors to behaviour and socioeconomic outcomes, candidate gene studies were the dominant method. In a candidate gene study, a gene is hypothesised to have an effect, and that hypothesis is tested directly. However, there are some major problems with candidate gene studies, with the literature littered with claims of the “gene for X” that simply can’t be replicated.

David Cesarini opened the session by pointing to this low level of replication of candidate gene studies. He suggests three problems might be causing this failure to replicate. These are multiple hypothesis testing coupled with publication bias, population stratification, and the low power of the small samples typically used.

Multiple hypothesis testing in candidate gene studies arises because more than one gene tends to be tested. In that case, the significance level of the tests should be adjusted to account for the multiple tests. But the reality is that the many negative tests never see the light of day, with the successful ones presented as successfully meeting a threshold appropriate for a single test. Publication bias exacerbates that problem as negative results tend not the be published and you don’t know how many tests have been conducted.

In contrast, GWAS is a hypothesis free approach. All SNPs in a sample (single nucleotide polymorphisms – DNA sequence variations in which a single nucleotide varies in the population) are tested for association with a trait. As there are as many hypotheses being tested as there are SNPs, very high significance thresholds are applied to avoid false positives. But as the number of SNPs in an array is known from the start, there is no doubt about the appropriate threshold.

Cesarini’s talk focused on the second problem, population stratification. This occurs where allele (variants of a gene) frequencies correlate with confounding variables. A classic example is analysing a mixed population of Asians and Caucasians and discovering the chopsticks gene. This can be overcome in GWAS by a technique called principal components analysis, which can be used to model the ancestry of the population and correct for stratification before conducting the analysis.

The next speaker, Daniel Benjamin, spoke on the third problem – the low power of candidate gene studies. Power is the ability to statistically demonstrate an association when that association exists. A test with low power will miss the associations most of the time.

The low power of candidate gene studies is partly due to their typically low sample size, usually between 50 and 3,000 people. Benjamin points out that there may not be any genes in social science with effects large enough to be detected in samples of this size.

The low power of a study has an important implication beyond the inability to find any effects that exist. If real results are rare, they will be swamped by the false positives, which would occur for 1 in 20 tests using the typical significance level. Benjamin runs through some numerical examples and shows that given the expected effect sizes of genes on social science outcomes, you simply shouldn’t trust most candidate gene study results. False positives will drown the real findings. This contrasts with GWAS. Once you get to decent sample sizes in the order of 100,000, you can be relatively confident that what you do find (even though you miss a lot) will be true.

Benjamin also talks about the Social Science Genetic Association Consortium (SSGAC), which is an attempt to build datasets large enough to apply GWAS to social outcomes such as IQ and risk aversion. The proof of concept was on educational attainment, which the next speaker covers in more detail.

Philipp Koellinger opens by asking why there are so many null results in the search for genetic influences. Is it because the effects are small? Because they are non-linear? Or there are gene-environment interactions? Maybe the results of twin studies showing most social outcomes are heritable are wrong?

Part of the answer was given by a study of educational attainment in which Koellinger and the previous two speakers were involved. They used a GWAS to search for SNPs that affected educational attainment in an initial sample of 100,000 people. They then replicated the result in another sample of 25,000 people. All three SNPs found in the discovery stage were replicated.

Importantly, the effect sizes were smaller than expected, with those three SNPs explaining 0.02% of the variation in educational attainment. If you added up the effects of all the SNPs in their sample, you could explain around 2 to 2.5% of the variation.

While this sounds low, it provides a basis for hope. Based on projections for larger sample size, it should be possible to explain 20% of the variation in education attainment through genetic factors.

Jason Fletcher was next, and he asked two main questions. First, how much should we believe GWAS results given how differently GWAS is done compared to normal science procedure. Second, what use are GWAS results? He spends more time on the second question and points out the usual possibilities, such as providing measures for latent variables. For example, if you don’t know the IQ of your sample but have their genomes and know how this affects intelligence, the genetic information could be used to attempt to determine the effect of IQ on a certain outcome.

Fletcher also points to the potential for exploration of gene-environment effects. He gives the example of people responding differently to tobacco taxation based on having different alleles. His paper on this topic is here.

Within his talk, Fletcher asks an interesting question about whether the SSGAC will become a natural monopoly in GWAS. Do we need a second SSGAC to enable people to check the results, and is it feasible for one to emerge? Others may be more viable as genetic testing becomes cheaper, but the tendency for one to dominate may still remain.

In the questions to Fletcher’s presentation, Benjamin makes the important point that the use of GWAS results as control variables could give much more precision to the estimates of the effect that a social science experiment is designed to measure. He gives the example of the Perry pre-school project – expensive educational interventions with a small sample, in which any added precision as to their effects would be of great value.

The last speaker, Dalton Conley, returned to the population stratification problem. His argument is that it may not be as easy to solve as it seems. Conley refers mainly to a technique called Genomic-relatedness-matrix restricted maximum likelihood (GREML) or Genome-wide complex trait analysis (GCTA) (which I have posted about before). This technique seeks to determine the contribution of all the sampled SNPs combined to variation in a trait. The output is a lower bound estimate of heritability. This technique relies, however, on an assumption that among those who are less related than second cousins (higher degrees of relatedness are removed), they share alleles in a way that is uncorrelated with any similarity in environment.

Conley argues that this assumption is false, and shows that using GREML, he can obtain a finding that birth in an urban or rural environment is heritable, in direct violation of the assumption. This result does not disappear after controlling for population stratification.

To deal with this problem, consideration should be given to testing for variation within families – any differences in genes between siblings will truly be random. The problem with this is that most massive datasets for which GWAS is performed don’t have pedigree data of that nature. The good news, however, is that the violation of the assumption does not seem to puncture the GWAS results. It is violated but the consequences are trivial. A paper by Conley and friends on this paper can be found here.

A week of links

Links this week:

  1. A good Jared Diamond interview.
  2. The 10,000 hours rule – the best you can do is find the peak of your own ability.
  3. Tinder works because a picture is “worth that fabled thousand words, but your actual words are worth… almost nothing”. (HT: Razib)
  4. Dumb incentives, although economists would be the first to point out a lot of the unintended consequences.
  5. No evidence for the benefits of expertise for fund managers.
  6. Drunks are more utilitarian. And maybe you should do that drinking on an empty stomach.
  7. Are social psychologists biased against Republicans?