In response to my comments on his recent opinion article on the the human tendency to overvalue information presented as stories, Reid Hastie writes:
Andrew (and Commenters) … I’d like to try to clarify some of the statements and implications in my Bloomberg article on “Our Gift for Good Stories …” The essay is what it is, but some of the implications that I intended to convey do not seem to have been communicated effectively. So, let me take a shot at clarification here. (Of course I am not assuming that, once clarified, my statements are necessarily correct or that you will agree with them.)
1. What I meant in the sections of the paper that claimed the brain is naturally good at visual and causal (narrative) thinking, is that that the brain was probably selected, through evolutionary processes to be adaptively successful at those capacities. I don’t have good evidence for this claim … but, we do a lot of those kinds of thinking, we’re distinctive as a species in the ways we do them, just about every typical human in any culture does them much the same way, there seems to be some evidence for systematic localization in brain structures, and we’re pretty good at them.
In contrast, I don’t believe that mathematical thinking has those properties. (Here I mean stylized mathematical thinking that goes beyond counting – such as algebra or geometry and more elaborate forms of this capacity.) I admit I know even less about evidence for cultural (not) universality, developmental trends, and brain localization (my reading of recent neuroscience, gives me the impression that mathematics beyond counting “hijacks” circuits in the brain that were not selected for these abilities … but, I admit I’m not even sure how to define some of these terms.) I also believe, from behavioural experiments, that we are not endowed (through natural selection) with the logical reasoning skills required to solve stylized logic problems like those that are presented in classroom logic courses. Though we are pretty good at some elaborate forms of “natural deduction” that are best described as rule-based. Here I think our abilities are part of our (selected) language generation and comprehension skills.
2. I agree that my assertion that Bayesian Causal Networks (really the principles behind that notation) provide a normative, rational model for causal reasoning is highly speculative. But, it seems to me that the Bayesian Causal Networks (e.g., the principles for calculation on those networks proposed by Pearl, and many others) are more popular than any previous proposal for a normative model, and that approach seems to distill the truth from many of the most plausible prior accounts. I appreciate that the Bayesian Networks approach is still under construction and it may fail gain the status of a generally accepted normative framework. But, every so often you have to bet on a new theoretical development, and I’m placing a wager on Bayesian Causal Networks … I certainly have not won that bet yet. (I also think Andrew is right that most realistic causal systems studied in the behavioural and biological sciences are defined by many small-influence causal relations and the current versions of Bayesian Causal Networks may not be a practically useful framework for modeling those phenomena.)
3. I am not apologetic about my choice of the Tversky & Kahneman causal conjunction brainteaser: Which is more likely “A flood drowns 1,000 Californians” versus “An earthquake followed by a flood drowns 1,000 Californians”? (And, hey, nothing against Californians: some of my best friends are Californians.) I agree that the standard criticisms of the heuristics & biases brainteasers apply (to the specific example I cited): Are there linguistic or conversational biases in the wording; how can untutored respondents use the probability response scale in a consistent and reliable manner; if there is no referent, how can we talk meaningfully about probabilities or errors; does this “brainteaser” really illustrate the bias introduced by “narrative causal coherence,” or is it just simple availability. (I did check on the numbers for that problem before I cited it in the article and I now regret overstating the magnitude of the effect size.) I used the flood-earthquake problem because it was the most relevant to readership of a Bloomberg editorial. And, I used this problem because I believe that it “works” rhetorically to illustrate a form of the “narrative fallacy”; our bias or tendency to believe “good narratives” are likely to occur and to think the “good story” confers other advantages, such as the ability to predict the next similar event.
Andrew says he does not see how this example demonstrates a problem with logical thinking skills. My answer is not subtle; I just buy the Tversky-Kahneman conclusion that there is an implied violation of logical set-superset membership relationships when we rate the probability of a conjunction higher than the probability of a component category. If you don’t accept that interpretation, you can certainly argue the alternative (see my list of methodological weaknesses above, plus the fact that different samples of respondents made the ratings that I claim “violate the logical principle”). But, I still prefer my interpretation.
4. I’m not sure how to respond to the criticism that my claim that many interpretations of the recent financial crises are examples of narrative fallacies is actually motivated by my desire to excuse specific individuals and institutions from responsibility for the crises. I can say it was certainly not my intention to excuse bankers and others from responsibility when I included that example. Again, I thought I was providing an illustration that typical Bloomberg readers could understand; and that I was warning them not to commit the fallacy in their professional roles – and especially to be more modest about their own abilities to forecast future events. Although I’m (happily) located in the University of Chicago Business School, I am not knowledgeable about financial events. If I were to characterize my own views on responsibility for the recent financial events, I would say that many bankers, analysts, raters, regulators, and politicians should be punished, much more harshly than they have been (or will be). (To cite an analogy, I reject most philosophical conceptions of “free will,” but I also firmly endorse incentives and sanctions to control socially consequential behaviour. Okay, argument from analogy is always dubious.) In any case, the reaction to my citation of (some) discussions of the financial crisis as examples of the “narrative fallacy” was an unintended consequence and I’m grateful to the blog for pointing that out.
5. I want to thank Andrew for pointing out that my tone seemed smug. That too was unintended (though maybe that’s not very remarkable, who sets out to intentionally write an article that sounds smug – maybe the occasional humorist). I’m not used to writing “thought pieces,” and that smugness is a style problem I need to work on. Many of the assertions in the article were tentative in my mind, but I wanted to say something definite and engaging. That’s probably a reason that many popular science writers get in trouble.
6. If you were interested in the topic of my essay, I would recommend a couple of other commentaries on the same topic: My beloved co-author, Robyn Dawes published a much more thoughtful essay on the same theme, “Prediction of the future versus an understanding of the past” (American Journal of Psychology, 106(1), 1-24); and I admire Duncan Watts recent (2011) popular book, “Everything is Obvious … Once You Know the Answer.”
Finally, I’d like to thank Andrew, and the rest of you bloggers for taking my article seriously enough to comment. As a further testimonial, I’ve enjoyed Andrew’s blog for many years and it is the most-visited non-news site on my browser toolbar. And, finally, finally … thanks for the Repo Man quotes (My favourite? p. 139, Hastie & Dawes, 2010).
When blogging I spend so much time reacting to journalists who seem to either ignore criticism or incorporate it without acknowledgment. I doubt, for example, I’ll be getting any responses from Michael Barone, Doug Schoen, Gregg Easterbrook, or Campbell Brown. Academic researchers, though, have a tradition of thoughtful dialogue (with some exceptions). I appreciate Hastie’s thoughtful reply.