Overall, you could say that people analyze arguments on (at least) two dimensions – the rhetorical dimension and the logical dimension, and how likely you are to convince someone that you have a valid point depends on some weighted sum of the two. As far as I can tell, this can cause serious problems for the use of logical discourse to aggregate information. In the most simplified case, you have on piece of information – whether or not your opponent agrees with you – but two variables – how effective your rhetoric was and how good your logic and information are. Without knowing how these factors are weighted, you can’t reliably use discourse to aggregate information.

I’m not really sure how everything plays out, though. The nature of the effect of rhetoric on someone’s analysis of a given topic is not well-defined in this model and as such it is hard to determine what the possible aggregate effects are. This does, however, lead me to a much softer stance on failing to concede than I once took. A few years ago, I was frequently extremely frustrated by the fact that many people would concede the fact that they could not offer a counter-argument to points that I made in debate, but they also refused to change their belief. If their belief was rational, they should be able to defend it. If not, they should give it up. However, this ignores the fact that it is possible that they found my arguments convincing for some rhetorical reason. Perhaps I used some subtle wordplay to distract them from the weak points in my arguments (maybe even without knowing about it). Without knowing how much of my argument was strong because I am good at debate and how much was strong because I am good at logic, the fact that they could not win an argument with me did not necessarily mean that I was right. Hopefully the fact that I bested you weighs in favor of re-analyzing the beliefs that you found yourself unable to defend, but it need not mean that you should immediately switch sides.

One thing to note is that even viewing all logical discourse as a collaborative social activity, the adversarial mechanism for debate (wherein each participant has a “side”) may very well be the best way to undertake the task. Since it is cheap to communicate a conclusion but costly to undertake a full accounting of all reasoning, it is reasonable to only check all your answers if you know that at least one (or both) parties have made a mistake. If the parties come to contradictory conclusions, necessarily at least one has made a mistake (or they do not share common premises).

In the view of discourse-as-debate, wherein one party’s only goal is to attempt to convince another party that their own view is correct, discourse is a zero-sum game (a variation on this is the political debating style, wherein parties aren’t trying to convince one another, but they are trying to score points with the audience). In this view, you score points against your opponent. However, discourse-as-information-aggregation is a non-zero sum game. All parties come to the table with their own information (some of it possibly redundant), but they don’t have to spend information to transmit it to the other participants. In this model, it is possible for everyone to win. Because of this fact, there should specifically be an incentive to make concessions as soon as you believe they are warranted.

Consider that the best strategy (once you are already at the table with your information in hand) will involve strongly considering efficiency in communicating and assimilating information from other participants – if you don’t give in to a better idea as soon as it comes along, you will waste effort arguing that could be spent getting more information for everyone. This is not to say that, in practice, you should be looking for places where you can concede your point because it means you’ve assimilated information quickly, of course – efficiency is the ratio of value to energy expended. I’m sure most people have had debates where their opponent had an extremely compelling case that fell apart upon greater reflection or the introduction of new information. The point is that ideally, a concession of a point tells you about who had the better information or reasoning at the outset of the debate, but afterwards both parties are better off because they now know (of the two views) which is more likely to be correct.

This is, by the way, not a positive statement about what discourse, as a whole, is, nor is it a normative statement about what discourse should become. My goal with this is to illustrate the way that I view discourse in terms of information. Additionally, I would not be surprised to find that no debates can be perfectly modeled from a pure logical-discourse standpoint. However, I think that logical discourse is an important subgame in many debates that is not to be overlooked.

The book is a case for how and why standard rational voter models in public choice fail and what the implications are for democracy. I don’t think I’m particularly well-qualified to criticize the empirical work (in the sense that I haven’t exactly pored over it), but even if the book were framed as a counter-factual, I would still find it fascinating. He lays out in excellent detail what factors are at play in democratic systems with and without various assumptions of voter rationality and eviscerates a number of common misperceptions about the mechanisms of politics. For example, Caplan argues that interest groups do not and cannot change policies that voters care about, but rather they are most effective by pushing policy on the margins of indifference. A huge lobbying campaign could never get murder legalized, for example, because any politician who implemented such an unpopular policy would be immediately ousted from office. If, however, voters want protectionist policies in general, but don’t care about the specifics, a lobbying group has some slack to choose which industry is protected. This sort of, “I never thought of it that way, but now that you mention it…”-style insight is a common and extremely enjoyable occurrence in the book.

I was also happy to see that Caplan addressed and elaborated a few of the concerns I’ve been thinking about in his section on the four common biases – anti-market bias, anti-foreign bias, pessimistic bias and make-work bias and in his chapter on “rational irrationality.” Specifically, I’ve blogged here about my nascent ideas on what Caplan calls make-work bias and rational irrationality, and while the section on biases is far too short for my tastes, it has already given me something to think about. There’s also a brief section towards the end about discourse and the role of rhetoric has to play in an irrational world that leads me to believe that Caplan wrote this book specifically for me – who else would like a book about the details of economics, rationality, bias, politics and discourse? (Oh, right.)

The one negative thing I have to say about the book is that it was a bit too practical for my tastes. By this I mean that Caplan places more weight on forming coherent counter-arguments to pro-democracy advocates than he places on just forming a coherent, generalized argument. For example, there is an entire chapter on “Market Fundamentalism” vs. “Democractic Fundamentalism”, wherein he makes a fairly strong case that while the popular perception is that economists are “market fundamentalists”, the opposite is in fact true and, to take it one step further, those leveling the accusation are often guilty of actually being “fundamentalists” when it comes to democracy. While I understand why he would want to discredit the idea that economists are “market fundamentalists”, it seems like a minor point to examine the hypocrisy of those making the accusation – and one that I doubt Caplan would have included if it were not for practical concerns.

I hate to have ended this review with a paragraph about this one minor quibble, because I don’t want to leave the impression that I was anything but satisfied with my experience. Overall, the book is fantastic and worth reading for anyone interested in politics, discourse or rationality. Hell, even if you aren’t interested but you do like to vote I suggest reading it (or at least listening to the EconTalk podcast).
Rating: 10 of 10

And that is why I find it so grating to hear the standard poorly-articulated disparagements of Twitter (and before that Facebook and mySpace and LiveJournal, etc) – the question is not “why should I care what you are doing?” (or at least, that is not necessarily an interesting question). In fact, the more interesting question is, “What, if any benefit does Twitter provide to its users, and if it does not provide any benefit, why do they use it?” If you don’t think that there is any reason to want to know about the minutiae of other peoples’ lives, then you should be puzzled by Twitter and you should almost certainly not be dismissive. Think about it – if millions of people derive some benefit from using Twitter but you can’t even fathom why, by determining why they find it entertaining, you could be discovering a previously unknown form of pleasure. Or maybe you’ll have a better understanding of how the millions of people around you think and interact. That’s not to say that the search costs won’t be higher than the potential benefits (that’s up to you), but I do think that it’s worthwhile to recognize that these sorts of puzzle are more interesting than you might think a priori.

If X then Y. Y is an undesirable consequence, therefore not X.

An example of a non-fallacious argument from final consequences would be:

If X then Y, not Y, therefore not X.

The way that I’ve always found reductios useful is in demonstrating that someone has more premises than they actually take account of. For example, in a recent conversation I was told that a certain law would be enforced “regardless of cost”. My response was to ask whether the law should be enforced if the cost were 100% of GDP or 200% of GDP. If the person actually only cared about enforcing this law, the reductio would still serve to illustrate that, but if they also cared about being able to buy cars and food and shelter and healthcare, they would have to take into account the fact that either their premise of “enforcement irrespective of cost” must fail or their premise of “we should be able to buy food and shelter” must fail. The form of the argument is then:

If X then Y. If Z then not Y. Therefore, not X and Z.

Note that the effect is not to say that X is not true, just that both X and Z cannot be true at the same time. It also doesn’t say anything about the truth of Z. It can be true that neither X or Z is true.

Note that this is also not just a way to dash someone’s arguments. It’s also a very useful tool for your own thinking. Take any of your cherished positions and try and bring them to the extreme. “I think free speech should be upheld universally” becomes, “I would kill everyone I loved so to uphold a Nazi’s non-precedential (that is to say that this is a complete one-off) right to call Jews filthy” – maybe you think you would do that, or maybe you would come to the conclusion that free speech doesn’t have the intrinsic value that your premise assumes. And once you know that your premises aren’t consistent in that absurd context, you can walk it back and try and find out where the premises are consistent to try and formulate a new rule that includes only the essential elements of the original premise.

In my view, what a good model does is reveal the symmetry in a class or between classes of problems. What this does is allow you to solve many problems at the same time – if you know how to solve a fifteen puzzle with numbers on it, you know how to solve a fifteen puzzle with a picture on it, because even though the goal is different (in one case you are trying to order the numbers and in another case you are trying to make an image), the “hard part” is identical. In the same way, you can form a model for any sort of problem by separating out the “hard parts” from the “easy parts”, then solving the hard parts. Any time you come across something with the same hard parts, you can apply the same solution.

Russell Roberts at EconTalk has frequently said that he is coming to the view that economics is less valuable as an empirical science than as a framework for how to look at the world. I tend to sympathize with this view, but I am not sure that this is limited to economics. The same can be said for physics and chemistry as well – the empirical side is important there, but the bigger focus tends to be on developing ways to think about the world that are easier to parse. We find some problems and solve them, then run around trying to transform other problems into the ones we’ve already solved by modeling those new problems as the old ones plus something else. In fact, in computer science, the entire field of complexity theory is more or less devoted to finding isomorphisms between problems that we’d like to solve (and they do a damn good job of it, too).

This is why I think that how agents are modeled is the relevant part of the question of behavioral economics. If the models are the same either way, the puzzles are all still the same and still have the same solutions. If the models have to be adjusted, then you still have those earlier solutions, they just happen to be solutions to a different problem.

While I’m not saying this is the reason for “just-below” prices, it is an example that does a good job of illustrating the concept of rational ignorance — where the cost of additional information is greater than the potential benefit. The reason I find the concept, and many others in the book, so intriguing is that it starts to delve into the strategies that develop in a world where agents don’t have perfect access to information and where attaining information has an associated cost. The fact of the matter is that in this universe information isn’t always easy to come by, which can be a huge limiting factor in many fields. As I see it, logic is useful precisely because it allows us to economize on information-gathering and transmission. We won’t ever have perfect information and as such we have to live in an uncertain world — how we learn to quantify that uncertainty and make decisions in the face of imperfect information is, I think, an incredibly interesting puzzle.

The book looks at a large number of pricing puzzles and tries to provide rational explanations for why they might be the case. Largely, these focus on the degree to which different pricing structures arise in different situations. For example, why should manufacturer’s rebates or coupons exist? If everyone redeems the rebates or has the coupon, there’s no reason to bear the transaction cost of the extra hurdle, and so firms should just set the discounted price as the regular price. On the other hand, if no one uses coupons or rebates, there’s no reason to even offer the discount at all. So the only time when coupons or rebates could exist is when only a portion of consumers will take advantage of them – and then you get into the puzzle of why some consumers might want to take advantage of them and others would not. McKenzie takes great pains to illustrate the possible ways to resolve these puzzles.

There is also a strongly Hayekian component to the book. The presumption underlying most of his cases is that the people with the incentives to get prices right are more likely to have done so than the casual observer. The fact that a puzzle exists is taken as evidence that the person who is puzzled by the behavior doesn’t fully understand the situation. However, just because you aren’t puzzled by the situation doesn’t mean you fully understand it – the best thing about the book is how he makes the case that most simple answers to pricing puzzles actually disguise how complicated the puzzle is. He talks about how most people, when asked why there are occasional water shortages in California, will answer that this is because of the near-desert conditions that lead to droughts. While this seems to explain the phenomenon, he destroys this explanation by pointing out that while there isn’t much rain, there are literally no candy bars or automobiles falling from the sky and yet we never have shortages in these goods.

Overall, I recommend this book to anyone and everyone who is even a little interested in understanding economics as the science of making decisions. Most of his pricing puzzles go a long way to explaining many of the basic concepts in economics in a fun and fascinating way that is easy to connect to.

Rating: 10 of 10

When people take actions in any kind of market, they are providing some information about their preferences to everyone else. If I won’t buy a can of peas for more than $1, I’m signaling to everyone that I value peas at $1 or less, and that if you can provide them to me at that price or less, you are creating value. If I am rational, my behavior is presumably a perfect signal of my preferences – even if you think that those preferences are weird. If, however, I am irrational, I am sending false signals with my systematically biased behavior – if I truly value peas at $0.50, but I irrationally consistently choose to pay $1 for them every time, you will think that you are creating value by making peas for $0.75 when in fact you are destroying value.

This problem exists whenever you are trying to harness the information-aggregating power of any kind of market (including democracy, which is a specific kind of market). Imagine a market in, say, health care. If people honestly and truly would prefer more health, but rather than optimizing their utility function for health, they optimize some rough heuristic based on correlates of health (equating more health care with more health, equating more antibiotics with more health, etc), they will be less healthy in the end because of their irrational behavior – which sucks for them.

In the end, however, this only matters if you want to measure peoples’ actual utility functions (which probably only matters from a normative, utilitarian point of view). If you want to actually maximize social utility, then irrationality can throw a monkey wrench in the works. This comes up in The Myth of the Rational Voter in the form of preferences over beliefs and how they effect democracy. If democracies are constructed to aggregate information about the actual social utility created by a given program, if voters are rational with strange preferences democracies work perfectly (even if they seem to fail because the outcome doesn’t match what people say they want). If voters are actually irrational, however, they are maximizing some function that is not necessarily correlated with their utility functions, and as such democracies will fail at their task of maximizing societal utility.

One last thing I’d like to note in this post is that the fact that a difference does arise here doesn’t necessarily imply that there is a simple experiment that can be done to differentiate between whether people are rational or not. The only way to do that would again be to have an independent measure of utility that could be compared to the actual outcomes. Since we do not have a direct measure of utility, I don’t think that this difference leads to a way to distinguish between the two cases.

One way to look at this is as a democracy where each dollar is one “citizen” getting a single vote, with the actual players controlling the population dynamics of their own citizenry. If everyone starts out with the same amount of money, in the first round of “voting”, you will get an answer with some noise based on the bias in each of the bettors, after which the money in the pot is redistributed according to how people bet. Assume now that no bettor changes their strategy – if they bet 25% of their money on X in the first round, they will still bet 25% of their money on X in the second round. Even though the strategies of the people spending the money didn’t change, the people who did the best are now betting more in absolute dollars and the weighted average gets closer to the best answer. Each additional round of betting continues to give more weight to the people who are consistently right.

This is the real discipline of the market, in my understanding. It’s not necessarily that once you get burned by making bad decisions, you stop making those decisions – it’s that if you don’t stop making those decisions, you don’t get to make any more decisions. That’s not to say that no one does stop making bad decisions – in fact, the fact that people respond to incentives re-enforces this mechanism by not only changing weight given to the choices of good decision-makers, but by inducing a larger share of the population to become good decision-makers. These sorts of effects are why I find many of Dan Ariely’s points somewhat lacking – population dynamics and selection biases can significantly change the aggregate dynamics of a system and it is not sufficient even to show that there are systematic biases in human decision-making ability. It’s why I am a big fan of evolutionarily stable strategies in Game Theory, and where I’m starting to think that the real action is in economic analysis (at least in going from a micro- to a macro- picture).

One problem that I think markets are particularly well-suited to solve is the weighting problem. If I were to take a random sample of 100 people from the world and ask them each to solve a complicated physics problem, the best answer might not necessarily be the answer that comes up most frequently or the straight average of all the answers because of the possibility of inhomogeneity in the distribution of knowledge. Imagine that 10 of those people are physicists specializing in solving that type of problem and the other 90 are high-school drop-outs. If there is no systematic bias in the way the high-school drop-outs answer the question, then the straight average will be the same as the average of the 10 physicists. If, however, the high school drop-outs are systematically wrong in their answers, then the average answer is unlikely to be the best answer – even though the information needed to solve the problem is present in the sample. The correct answer would probably give a heavier weight to the answers provided by people who know what they are doing – and that is precisely what markets allow you to do. If you now change the question such that each of the 100 people are asked to place a bet on what the answer to the problem is (assuming that it is empirically verifiable), people will have a greater incentive to gather the information necessary to make an informed bet. The high school drop-outs will convey that they don’t have much information about the answer by not making particularly large bets, while the physicists will convey their confidence by making large bets if they know the answer.

Additionally, consider repeated interactions of the same sort. The people who bet imprudently now have less money to bet for the next round, and they will have even more incentive to be prudent, lest they lose all of their money. And even if they don’t become more prudent, it doesn’t matter because they will have less and less money with which to be imprudent. The more prudent people whose confidence in their ability most closely matches their actual ability will be more and more heavily weighted, filtering out people who don’t properly weight their information. Notice that the market need not filter out people who don’t have information. High school drop-outs could just choose to abstain from the market by making no bet and they will still do better than their comrades who bet imprudently. As long as peoples’ confidence in their information matches the validity of that information, the market can aggregate all that information up together with the proper weights.

In The Black Swan, Taleb is constantly warning people not to construct stories to match their facts, because you’ll view future facts in the light most favorable to your earlier story. Personally, I have mixed feelings in general about these carte blanche proclamations about hazardous but potentially useful devices. On the one hand, I don’t think that storytelling (or other hazardous devices) is a bad thing prima facie – if you understand the limitations of the tool, you can use it effectively (similarly I don’t think you should abandon logic because Spock, the archetype of the logical being doesn’t get it). However, he does make a very good point – there are a huge number of stories that you can tell from any one piece of evidence, and just because you already have a story in mind that fits doesn’t mean that one of the other stories is less likely to be correct. Each additional piece of evidence should not be used to strengthen your resolve in your conclusion, it should be used to eliminate alternative stories that don’t fit the new evidence. So, in the rationality example, Kahneman and Tversky’s work can’t (as far as I can tell) differentiate between strange preferences/rational behavior and normal preference/irrational behavior, but it can differentiate between either of those two and the combination of normal preferences/rational behavior.

So, in the end, I have a cautious prescription: if you must pick one, assume that whichever equivalent story is easier to work with is true, but keep in mind that that doesn’t make it more likely to actually be true. So if it’s easier and equivalent to model behavior as rational with strange preferences, then use that as your framework, always keeping in mind that any asymmetry between the two models that you develop must be taken into account. Any time that equivalency breaks down, your model also breaks down, which is why it’s a good idea to try and find out which story is actually true (the closer you are to the truth, the fewer sources of inequivalency there are that can screw up your predictions).

Addendum: The more I’m thinking about this, the more I worry that I am attacking a straw man here. I don’t want to say that people are arguing some semantics, just that that’s how it seems to be framed. In my posts on rationality, I’m assuming that “rational” means that people always attempt to make the decisions that with their level of knowledge they believe will optimize their utility. I get the impression that the behavioral economists recognize the fact that if you can’t measure someone’s utility function directly then you can’t distinguish between a whacky utility function and irrationality, and so they just call any evidence of either thing “irrationality” and take that to mean “not a combination of normal preferences and rational behavior.” My point here isn’t that these guys are dumb, I just wanted to explain my thinking about what the real problem is.

So far, the only way I can see that could throw a major monkey-wrench into the task and make everything very complicated is if preferences don’t commute – that is to say that if you prefer A to B and B to C you necessarily prefer A to C. If preferences commute, then you are working entirely with an Abelian group, but if they don’t commute, you are working with a non-Abelian group and their properties are very different. If preferences are commutative, you can easily convert preferences directly into money and back into preferences (would you rather have A or $x, B or $y, C or $z), which means that you can define a simple measure of value. However, if they do not commute, you have to define a much more complicated function that defines the relative value associated with preferences. Note that you still can define a measure of value and map this type of irrationality onto a map of preferences, it’s just not as simple.

I think one thing that is often overlooked is that most of these economic models that “assume rationality” are formulated in an extremely general way, which makes it much easier to define a map from irrationality to preferences. For example, in game theory a game is defined by its payoff matrix, not its outcome matrix. For example, in the prisoner’s dilemma, the outcome matrix is as set forth in Fig. 1, but the “game” of the prisoner’s dilemma is anything with a payoff matrix of the form Fig 2, where a_1/2 < b_1/2 and a_1/2 < d_1/2. Just because we expect rational people only considering the length of their sentences to have a payoff matrix of that form when the outcome matrix looks like Fig 1. doesn’t mean that any time you are presented with that outcome matrix you can assume you are in a prisoner’s dilemma. A rational person might simply prefer not to rat out their compatriot – the stronger the preference for keeping tight-lipped, the lower that person’s payoff for defecting. Even if you found that no people, when put in a literal prisoner’s dilemma ended up in the Nash equilibrium, that doesn’t necessarily mean that the Nash equilibrium analysis doesn’t work, it could very well mean that the players are playing a different game (with a different payout matrix).

Overall, I think the debate seems like it’s being framed very poorly. It’s very easy to model the behavior of individuals if you assume that you know their utility functions at all points in the decision-space and that they are rational always. However, if either of those assumptions breaks down, you don’t have much to go on. Additionally, deviations from either of those assumptions can be mapped onto a model of deviations from the other, so it is sufficient to choose whichever one you like more and assign all prediction errors to a breakdown of the other assumption. Effectively, what this means to me is that there’s no reason to debate whether or not people are rational – you can assume that they are rational and measure peoples’ utility functions (you are going to have to measure them anyway, the only difference here is that you have to account for the function having a more complicated form).

Updated: Oops, accidentally switched the “cooperate” and “defect” labels in the figures. Thanks to Iron Man for pointing out the error. The situation has now been remedied.

First off, the arguments that I see put forth by people like Ariely seem to take the form of “people aren’t rational, therefore we throw standard economic theory out the window, because standard economic theory assumes that people are rational.” But consider the fact that if people are irrational, but their irrationality is random and doesn’t introduce any systematic bias into their decisions, the average agent acts as if he were rational (because any mistake that one person makes in one direction is canceled out by mistakes that another person makes in the other direction). Ariely himself makes the case that people are systematically irrational, which does mean that the average agent acts irrationally, but that irrationality could be modeled as a a rational preference for what outsiders view as irrational. For example, in one chapter Ariely explains how people are happier to do an action for free than to be underpaid for it. You can say that this is irrational because getting paid is everything you got doing it for free, plus more. You could also say that this is a rational preference for giving things away for free. Getting paid takes away that benefit, and so you need to estimate the relative magnitudes. Effectively, predictable “irrationality” is isomorphic to having rational preferences for seemingly “irrational” things. Since your utility function is effectively the premise upon which you base your decisions, it is impossible for your true preferences to be irrational.

This is not to say that people aren’t irrational, but the fact of the matter is that since predictable irrationality is necessarily isomorphic to rational preference for irrationality and unpredictable irrationality is irrelevant, this means that as far as I can tell, there is no meaningful distinction between the two cases. The main difference is how useful you will find various stories. The more I read, though, the more I’m finding that people in the field do realize that this is the case, and it’s just populist rhetoric to spout out that “everything economists have learned is wrong!”

To me, this book seems like a very good introduction to the important results in behavioral economics. Ariely does a very good job both making the case for the validity of his experimental results and in making the book entertaining. He takes specific case studies out of his samples to build an interesting narrative, then goes into details about how controls were implemented and the statistical significance of the results. As far as popular descriptions of scientific experiments go, he did a good job of making the case that his experiments were well-designed.

His descriptions of the nature of human rationality could also be very useful in your own day-to-day decision-making. The book goes into explicit detail about the most common irrational behaviors that we humans are likely to engage in and explains why they are irrational. While I certainly wouldn’t recommend this as some perfect guide to improving your life, he does bring up some ideas that might be worth considering.

Where the book goes bad is whenever Ariely tries to generalize his results. The conclusions he draws are far too broad and he does not make a compelling case for them. For example, one chapter in his book is entitled “The Fallacy of Supply And Demand: Why The Price of Pearls – And Everything Else – Is Up In The Air”. In this chapter, he very effectively lays out evidence that people will anchor to essentially random prices, and then concludes that this means that prices have nothing to do with supply and demand. Clearly it’s evidence that people aren’t necessarily rational about how they evaluate prices, but a huge part of how prices are set is due to competition. Without addressing the aggregate dynamics of predictably irrational actors, he can’t draw conclusions about aggregate phenomena like pricing.

Overall, I’d say the book is worth reading just for the descriptions of Ariely’s experimental work – the book contains a large number of interesting results in the field of behavioral economics. However, I would take with a grain of salt any of the conclusions he draws from that work, as he frequently seems to skip steps in the logic of his arguments.

Rating: 7 of 10