So, I want to take you to a Wikipedia page that I first saw when someone tweeted that they had found “the best page on the internet”. The “List of cognitive biases” was up to 165 entries on the day I took this snapshot, and it contains most of your behavioural science favourites … the availability heuristic, confirmation bias, the decoy effect – a favourite of marketers, the endowment effect and so on ….
But this page, to me, points to what I see as a fundamental problem with behavioural economics.
Let me draw an analogy with the history of astronomy. In 1500, the dominant model of the universe involved the sun, planets and stars orbiting around the earth.
Since that wasn’t what was actually happening, there was a huge list of deviations from this model. We have the Venus effect, where Venus appears in the evening and morning and never crosses the night sky. We have the Jupiter bias, where it moves across the night sky, but then suddenly starts going the other way.
Putting all the biases in the orbits of the planets and sun together, we end up with a picture of the orbits that looks something like this picture – epicycles on epicycles.
But instead of this model of biases, deviations and epicycles, what about an alternative model?
The earth and the planets orbit the sun.
Of course, it’s not quite as simple as this picture – the orbits of the planets around the sun are elliptical, not circular. But, essentially, by adopting this new model of how the solar system worked, a large collection of “biases” was able to become a coherent theory.
Behavioural economics has some similarities to the state of astronomy in 1500 – it is still at the collection of deviation stage. There aren’t 165 human biases. There are 165 deviations from the wrong model.
So what is this unifying theory? I suggest the first place to look is evolutionary biology. Human minds are the product of evolution, shaped by millions of years of natural selection.
Wake me up when evolutionary biology can predict all those 165 things from first principles and a very little input the way modern astronomy can predict the motion of planets.
I would agree that a collection of biases points to a need for a theory, but I don’t think such a theory is likely to be central to the economics model simply because those deviations are irrelevant in a large number of cases. Simple rational expectations can be quite predictive of human behavior in many cases even though it is clearly completely absurd. Think of the relationship between quantum mechanics and relativity. Relativity doesn’t seem to fit at all in any reasonable way into quantum mechanics, and yet relativity is quite useful and accurate for problems at the atomic level and above.
Jason Collins’ reasoning can be used for almost any scientific theory imaginable. If you examine any scientific discipline closely enough, you will find deviations which don’t fit the standard model. But the existence of deviations does not necessarily prove the need for a new model; particularly if those deviations do not appear to be central to the model’s primary predictions. I would say a rigorously tested theory of how cognitive biases develop and are maintained may provide some useful insights into economics, but that it’s unlikely that they will disprove the basic model of supply and demand.
There are other issues in that essay. Present bias isn’t normally considered a bias. It’s referred to in economics as temporal preferences. Hyperbolic discounting has from its conception been considered an issue of preference, and only later as one of rationality. He then discusses conspicuous consumption in the context of mating signals except that isn’t a new idea either. Economics already has a theory of signalling that roughly matches with what he is referring to, and they’ve already considered social status as a type of signal, and that conspicuous consumption is used to signal social status. That also isn’t an issue of rationality, but one of preference.
Relativity doesn’t seem to fit at all in any reasonable way into quantum mechanics, and yet relativity is quite useful and accurate for problems at the atomic level and above.
General relativity doesn’t seem to fit at all in any reasonable way into quantum mechanics, and it is way overkill for problems at the atomic level (and for a sizeable fraction of problems at the planetary level, too), where special relativity (which does fit with quantum mechanics, save for a few theoretical loose ends that few non-mathematicians have reasons to care about) suffices.
I don’t think that the fact that Wikipedia has a list of 165 cognitive biases says more about Wikipedia than it says about behavioural economics.
The core idea from Kahnmann is that humans use heuristics to make decisons.
Evolution certainly affected human cognition but most designs for intelligent agents don’t produce intelligent agents. The space of set of heuristics that produce intelligent agents is small. It’s not clear that you can make an intelligent agent of something like a neural net that doesn’t engage in something like the availability heuristic. Confirmation bias isn’t something substantially different than the availability heuristic in action.
When Google’s dreaming neural nets reproduce quirks of the human brain it’s hard to argue that those quirks exit because they provide advantages in sexual competition.
So what does an evolutionary approach tell us about the human mind?
For a start, it tells us something about our objectives.
That’s basically saying Darwin was wrong and his critics who object to organism that don’t evolve according to objectives were right. Darwin wasn’t controversial because he invented evolution. Lamarks already did that decades before Darwin. Darwin was controversial because he proposed to get rid of teleology.
Economics makes errors because it assumes that humans have objectives. You don’t fix that by explaining how humans have different objectives. You fix it by looking at the heuristics of human beings and also studying heuristics of effective decision making in general.
Seems like some people replace the teleological model of “it evolved this way because the Spirit of Nature wanted it to evolve this way” by a simplistic pseudo-evolutionary model of “it evolved because it helps you to survive and get more sex”.
Nope. Some things evolve as side effects of the things that help us “survive and get more sex”; because they are cheaper solutions, or because the random algorithm found them first. There are historical coincidences and path-dependency.
For example, that fact that we have five fingers on each hand doesn’t prove that having five fingers is inherently more sexy or more useful for survival than six or four. Instead, historically, the fish that were our ancestors had five bones in their fins (I hope I remember this correctly), and there was a series of mutations that transformed them into fingers. So, “having fingers” was an advantage over “having no fingers”, but the number five got there by coincidence. Trying to prove that five is the perfect number of fingers would be trying to prove too much.
Analogically, having an imperfect brain was an advantage over having no brain. But many traits of the brain are similar historical artefacts, or design trade-offs, or even historical artefacts of the design trade-offs of our ancestors. A different history could lead to brains with different quirks. Using “neural nets” (as opposed to something else) already is a design decision that brings some artifacts. Having the brain divided into multiple components is another design decision; etc. Each path only proves that going this path was better than not going there; it doesn’t prove that this path is better than all possible alternatives. Some paths could later turn out to be dead ends.
I agree that treating humans as “rational beings with objectives” can be a nice first approximation, but later it’s just adding more epicycles on a fundamentally wrong assumption.
Please, not another bias! An evolutionary take on behavioural economics by Jason Collins
Wake me up when evolutionary biology can predict all those 165 things from first principles and a very little input the way modern astronomy can predict the motion of planets.
I would agree that a collection of biases points to a need for a theory, but I don’t think such a theory is likely to be central to the economics model simply because those deviations are irrelevant in a large number of cases. Simple rational expectations can be quite predictive of human behavior in many cases even though it is clearly completely absurd. Think of the relationship between quantum mechanics and relativity. Relativity doesn’t seem to fit at all in any reasonable way into quantum mechanics, and yet relativity is quite useful and accurate for problems at the atomic level and above.
Jason Collins’ reasoning can be used for almost any scientific theory imaginable. If you examine any scientific discipline closely enough, you will find deviations which don’t fit the standard model. But the existence of deviations does not necessarily prove the need for a new model; particularly if those deviations do not appear to be central to the model’s primary predictions. I would say a rigorously tested theory of how cognitive biases develop and are maintained may provide some useful insights into economics, but that it’s unlikely that they will disprove the basic model of supply and demand.
There are other issues in that essay. Present bias isn’t normally considered a bias. It’s referred to in economics as temporal preferences. Hyperbolic discounting has from its conception been considered an issue of preference, and only later as one of rationality. He then discusses conspicuous consumption in the context of mating signals except that isn’t a new idea either. Economics already has a theory of signalling that roughly matches with what he is referring to, and they’ve already considered social status as a type of signal, and that conspicuous consumption is used to signal social status. That also isn’t an issue of rationality, but one of preference.
I don’t think that the fact that Wikipedia has a list of 165 cognitive biases says more about Wikipedia than it says about behavioural economics.
The core idea from Kahnmann is that humans use heuristics to make decisons.
Evolution certainly affected human cognition but most designs for intelligent agents don’t produce intelligent agents. The space of set of heuristics that produce intelligent agents is small. It’s not clear that you can make an intelligent agent of something like a neural net that doesn’t engage in something like the availability heuristic. Confirmation bias isn’t something substantially different than the availability heuristic in action.
When Google’s dreaming neural nets reproduce quirks of the human brain it’s hard to argue that those quirks exit because they provide advantages in sexual competition.
That’s basically saying Darwin was wrong and his critics who object to organism that don’t evolve according to objectives were right. Darwin wasn’t controversial because he invented evolution. Lamarks already did that decades before Darwin. Darwin was controversial because he proposed to get rid of teleology.
Economics makes errors because it assumes that humans have objectives. You don’t fix that by explaining how humans have different objectives. You fix it by looking at the heuristics of human beings and also studying heuristics of effective decision making in general.
Seems like some people replace the teleological model of “it evolved this way because the Spirit of Nature wanted it to evolve this way” by a simplistic pseudo-evolutionary model of “it evolved because it helps you to survive and get more sex”.
Nope. Some things evolve as side effects of the things that help us “survive and get more sex”; because they are cheaper solutions, or because the random algorithm found them first. There are historical coincidences and path-dependency.
For example, that fact that we have five fingers on each hand doesn’t prove that having five fingers is inherently more sexy or more useful for survival than six or four. Instead, historically, the fish that were our ancestors had five bones in their fins (I hope I remember this correctly), and there was a series of mutations that transformed them into fingers. So, “having fingers” was an advantage over “having no fingers”, but the number five got there by coincidence. Trying to prove that five is the perfect number of fingers would be trying to prove too much.
Analogically, having an imperfect brain was an advantage over having no brain. But many traits of the brain are similar historical artefacts, or design trade-offs, or even historical artefacts of the design trade-offs of our ancestors. A different history could lead to brains with different quirks. Using “neural nets” (as opposed to something else) already is a design decision that brings some artifacts. Having the brain divided into multiple components is another design decision; etc. Each path only proves that going this path was better than not going there; it doesn’t prove that this path is better than all possible alternatives. Some paths could later turn out to be dead ends.
I agree that treating humans as “rational beings with objectives” can be a nice first approximation, but later it’s just adding more epicycles on a fundamentally wrong assumption.