Monthly Archives: January 2007

Agree With Young Duplicate

On the last two Sundays I gave examples where disagreement is justified, and where it is unjustified.  Today let me try to give a harder example where disagreement is not justified:

Again we have Star Trek style transporter modified to preserve the original while creating an atom-by-atom copy with identical personality, memories, and cognitive styles.  But this time the information to create this copy was stored for twenty years, and you find yourself standing outside the transporter as your duplicate emerges. 

You explain to your younger self that over the last twenty years you have reconsidered and rejected some of your most precious beliefs.   Following common aging trends, you are now more politically conservative and believe less in true love, your future fame, and the nobility of your profession.  Your younger duplicate suspects you are biased by self-interest and a need to explain your career failure, while you suspect he suffers youthful overconfidence.   How hard should you try to agree with him?

I say you should try hard to satisfy the rationality constraint "we can’t forsee to disagree."  You started out with exactly his biases, and you both know you may have overcome some, and some may be worse.  He should on average defer to you more than you to he, as he accepts that your belief changes come partly from more information and analysis.  But when he hesitates to fully adopt your views, you should accept that he may have good reasons to do so.   You should random walk your way to common estimates about your relative biases. 

GD Star Rating
loading...
Tagged as:

Sick of Textbook Errors

One of the most well-worn examples in introductions to Bayesian reasoning is testing for rare diseases: if the prior probability that a patient has a disease is sufficiently low, the probability that the patient has the disease conditional on a positive diagnostic test result may also be low, even for very accurate tests. One might hope that every epidemiologist would be familiar with this textbook problem, but this New York Times story suggests otherwise:

For months, nearly everyone involved thought the medical center had had a huge whooping cough outbreak, with extensive ramifications. […]

Then, about eight months later, health care workers were dumbfounded to receive an e-mail message from the hospital administration informing them that the whole thing was a false alarm.

Now, as they look back on the episode, epidemiologists and infectious disease specialists say the problem was that they placed too much faith in a quick and highly sensitive molecular test that led them astray.

While medical professionals can modestly improve their performance on inventories of cognitive bias when coached, we should not overestimate the extent to which formal instruction such as statistics or epidemiology classes will improve actual behavior in the field.

GD Star Rating
loading...
Tagged as: , ,

Manipulating Jury Biases

The top google link for the phrase "Overcoming Bias", other than links associated with this forum, is to a book length pdf called "Overcoming Jury Bias."  This is a well written guide to lawyers on how to get a jury to agree with them.   It does show how to overcome jury bias when that would help you, but it also shows how to enhance jury bias when that would help you:

A woman who claims to be the victim of a sexual assault must appear in court as a "victim". If the woman appears before the jury wearing tight clothing, ostentatious jewelry, overdone cosmetics and a wild hair style, this will create an incongruence in the minds of the juror of this woman as a "victim". …

Continue reading "Manipulating Jury Biases" »

GD Star Rating
loading...
Tagged as: ,

What Insight in Innocence?

A recent New Yorker review of "The Bloodless Revolution: A Cultural History of Vegetarianism from 1600 to Modern Times" shows how bias theories are central to the vegetarian debate: 

Commentators argued that the habit of killing, like that of meat eating itself, hardened the heart and the nerves, both figuratively and literally. The squeamish human response to animal suffering was the authentic one; the callous reaction induced by familiarity was accounted artificial or false.  … 

Cartesians had a response: any such human reaction was itself just a mechanical reflex. …  it’s true that many of those who have little experience of what goes on in an abattoir are repulsed by any kind of firsthand knowledge, or even by reading vivid accounts. But things are different on the other side of the slaughterhouse wall. Those who kill animals in the course of their working day may quickly become habituated to it, and to dismiss this effect as mere desensitization effectively discounts great knowledge of animal death in favor of slight knowledge.  Similarly, those who like to romanticize country people are frequently discomfited by their uncuddly ways with livestock. … Why is it "natural" not to know very much about "nature"? …

Continue reading "What Insight in Innocence?" »

GD Star Rating
loading...
Tagged as:

On Policy, Fact Experts Ignore Facts

I teach Health Economics starting today, and every year I start with data assuring students that learning data will not change their health policy opinions.  Regarding health economics:

In a survey of 50 leading health economists, Fuchs (1996) found considerable disagreement regarding major issues of health policy.  The extent of disagreement was particularly striking when compared with the high level of agreement among the same economists about the determinants of health and the determinants of health expenditures.   Furthermore, the small disagreement that did exist regarding the positive questions seemed to play no role in explaining policy differences.  ….

Regarding labor and public economics, Fuchs, Krueger, and Poterba (1997) found

Only one of the 13 [policy] proposals (a 25 cent per gallon increase in the gasoline tax) elicited a strong consensus either in favor or in opposition.  … Economists, like experts in many fields, reveal considerable "overconfidence" in their estimates of the economic parameters.  For most questions, a large portion [typically 30-40%] of the individual [95%] confidence intervals do not include the average best estimate, or even include the value that is covered by the largest number of confidence intervals. … respondents whose best estimates are farther from the median tend to give wider confidence intervals for those estimates.

Fuchs et. al. estimated economists’ policy opinions from their opinions on factual economic parameters, their values, and their political party.  Political party was never (10%) significant.  For 65 labor economists, values were always (3%) significant, but parameters were (10%) significant only on eliminating affirmative action.   For 69 public economists’ opinions on five topics, values were always (2%) significant, and parameters were never (30%) significant.  On expanding IRAs both parameters and values were (3%) significant, on state education financing only parameters were (1%) significant, and on mandatory savings accounts neither were (10%) significant.   (The other topics were increase AFDC benefits, increase gasoline tax, adopt VAT, eliminate OASI Cap, increase minimum wage, eliminate job training, and increase unionization.)

While most economists think that the facts they spend years studying influence their policy positions, most policy opinions are apparently determined almost entirely by values.  Since it is obvious that facts are very relevant, this all makes me ashamed to be … human. 

Added: Bryan Caplan suggests that learning facts may change our values, thereby making our policy opinions fact-dependent.   This seems a slim hope to me, but I hope we can check it with data on how opinions change with time. 

GD Star Rating
loading...
Tagged as: ,

The Butler Did It, of Course!

Here is a paper showing the potential practical utility of detecting and reducing biases: Confirmation bias in criminal investigations by O’Brien and Ellsworth. In an experiment subjects read a police file and were asked halfway through about their hypotheses of who the murderer was; practically everybody named the obvious suspect. On completing the entire file, where a second and stronger suspect emerges in the later half, they still tended to suspect the first guy. In a second experiment the subjects were asked to generate counter-hypotheses about why their suspect might be innocent, and this reduced the confirmation bias.

Another troubling source of bias is false confessions triggered by this confirmation bias and then strongly supporting the erroneous conclusion. The Psychology of Confessions by Kassin and Gudjonsson reviews this. During the preinterrogation review police, believing themselves to be better at detecting deception than they are, tend to confidently make false positive detections of deception in innocent people. Once they have convinced themselves they have caught a suspect a police interrogation then becomes guilt-presumptive and rather effective in generating false confessions, in particular in cognitively challenged people. And finally, juries and judges are easily convinced by the confessions.

Nothing of this may be total news to anybody on this blog, but it is still rather worrying how strong biases are accepted in police investigations and the legal system. Maybe the counter-hypothesis trick at least could be made part of police procedure: at certain points during investigations the investigators have to state possible disconfirming hypotheses for the record.

GD Star Rating
loading...
Tagged as:

No Death of a Buyerman

Do ads bias beliefs?  Many think so, and want to regulate ads.  And yes, many ads don’t seem to offer much useful info.   Here are four ad theories:

  1. Exploiting bias – the small leaks of our ancestors’ biases are now torn into gaping holes.
  2. Burning money – a company willing to waste money on ads signals confidence in its product.
  3. Identity – ads associate a product with a personal style, with which people want to identify.
  4. Hidden info – if you look closely, you will see that ads do in fact have lots of useful info.

Here is important clue:  Every transaction has both a buyer and a seller.  Yet we hear much more about salesmen, and how to sell, than we do about buyermen and how to buy.  There are few thoughtful plays about "The Death of a Buyerman" or books on "How to be won as a friend and influenced by people."  Why?

Buyers are usually more uncertain about their value than sellers are about their cost.  So both sides tend to talk about buyer value.  Thus to the extent that individual ability is relevant, whether a sale happens is more clearly a signal of seller ability than of buyer ability.  This suggests a fifth ad theory:

5. Seller signaling – buyers and others like to see and affiliate with impressive sellers.

Just as we like to see and affiliate with sports stars, musicians, actors, writers, and professors, we like to see and affiliate with people who have impressive abilities associated with sales.  Watching ads or listening to a salesman’s pitch may be like watching a sports star, and buying a salesman’s product may be like getting a star’s signature or t-shirt.   

Under this theory, we need no bias to want to watch ads that tell us little about a product, and then buy that product.  There might still be reasons to regulate such ads, but they would not be reasons of bias. 

GD Star Rating
loading...
Tagged as:

Is There Such a Thing as Bigotry?

Statements of the form “members of Group X are, on average, less good along dimension A than are members of Group Y” are commonly regarded as bigoted. The first instinct of someone committed to Overcoming Bias is to understand that statements like that are empirical statements that could be true or false, and if the evidence was that they were true, then saying so should not carry any moral taint.

There is much to be said for this view, but to apply it in the broadest possible way would seem to require one to believe that there there is no such thing as bigotry at all; statements are either true or false, true ones are good and false ones are bad, and the intent with which they are offered is irrelevant. I got to thinking about this in response to some comments that were left in response to this post about Israeli Jews and Arabs each thinking that members of the other group are “unclean.” Some of the commenters seemed to think that the factual question of which group took more showers was the issue. But this is a case where the facts are just about completely irrelevant. It is perfectly possible that one group is objectively less clean than the other. It is clear, however, that charges of “uncleanliness” regarding a rival group, whether the uncleanliness is supposed to be physical or spiritual, whether it is real or imagined, is a way of arousing the emotion of disgust, which is often a necessary emotional precondition to hatred. The actual facts about uncleanliness are beside the point, and the statements are a reflection of extremely worrisome bigotry.

A working definition of a bigoted factual claim should be any claim, whether true or not, that is intended to have the effect of inducing disgust or hatred in members of another group. As a practical matter it may be difficult to tease out when a particular true statement is offered with this intent (or whether members of the group in question might reasonbly suspect that it is offered with this intent, which is not necessarily the same thing) but I don’t see how one can avoid the conclusion that there is such a thing as a statement that is true but still bigoted.

GD Star Rating
loading...
Tagged as:

Moby Dick Seeks Thee Not!

“Oh! Ahab,” cried Starbuck, “not too late is it, even now, the third day, to desist. See! Moby Dick seeks thee not. It is thou, thou, that madly seekest him!”

In my copy, these lines appear on page 599 of a 604 page book. Ahab is behaving the way we expect literary characters to behave at the end of a book; he is going to keep on pressing until there is some kind of dramatic resolution one way or the other. Starbuck is reminding him that just because it’s page 599 doesn’t actually mean anything; there is nothing objective about the situation that prevents Ahab from coming to his senses, turning the ship around, and taking everybody home.

It seems like a lot conflicts have this property. On top of everything else that makes up a conflict, there is the tendency for the combatants to cast themselves as playing a role in a great drama, and their range of options is restricted by the unexamined requirement that great dramas are supposed to have dramatic resolutions. This tends to crowd out plain old pragmatism, and also crowds out morality to the extent that the moral solution doesn’t correspond to one or both sides’ dramatic one (notwithstanding the fact that the holders of dramatic positions loudly proclaim morality to be unambiguously on their side).

The Israeli writer Amos Oz has made a similar point about the Palestinian-Israeli conflict.

GD Star Rating
loading...
Tagged as: ,

Morale Markets vs Decision Markets

This diagram was from from my talk last Thursday:

Morale_vs_decision_markets

The easy way to start a corporate prediction market is to grab some cheap software, fill in some easy topics, and then hope people will play in their free time for bragging rights.  You might get lucky; offering only weak incentives might still provide key valuable information, and put you in the sunny upper left box of this diagram.

But typically weak incentives and fun topics don’t give decision makers especially valuable info.  Managers allow morale markets (the lower left box) mainly to be in on a new fad, to track wider moods and which groups think what, and to let people feel their voice is heard.   I’d put the Hollywood Stock Exchange and Google prediction markets in this category. 

Morale markets might get a foot in the door, but I fear that won’t last without a move to decision markets (upper right box), offering enough incentives to elicit info directly relevant to key corporate decisions.   Decision markets tend to be politically sensitive and less fun and easy for traders, so they need non-trivial management support and trader incentives, such as cash, job credit, or decision influence.  My hope rides on the handful of decision market trials now underway.

(The lower right box is high cost and low value – why bother?)

GD Star Rating
loading...
Tagged as: