Friday, January 20, 2006

Ethics and Statistical Analysis

This post is partially inspired by Jonathan Miles' comments regarding the last Moose and Dude comic. However, I felt that it was too off-topic to continue in that thread and I hope it will warrant discussion on its own here.

Jonathan mentioned, and Arthur continued to discuss, the fact that certain aspects of Jonathan's argument hinge on empirical data. In particular, in order to measure the force of Jonathan's argument, one would, minimally, need to determine what, if any, connection exists between beastiality and the "slippery slope" of harm. This brought to mind a particular concern of mine in ethics. Presumably, one probable avenue for attempting to determine whether any such connection exists would be the use of statistical analysis. For example, one might attempt to discover a significant correlation between people who engage in beastiality and people who engage in particular harmful behaviours.* This, it seems, is a fairly popular way of doing things. For example, I'm sure many of us have seen the commercials that tell us that if we eat dinner with our children, they are less likely to drink or do drugs. Presumably, this statment is based on some statistics about families eating dinner together and underaged drinking and drug-use.

I have always found such statements extremely troubling, mostly because it seems to me that saying "There is a statistically significant correlation between X and Y" is NOT justification for saying "X makes Y more likely." Largely, this is because there are other factors that might be involved. In the case above, for instance, it has always seemed likely to me that a correlation in that case indicates, not that eating dinner with one's children will keep them out of trouble, but that many parents who are able to keep their children out of trouble through good parenting also eat dinner with them. Thus, if you are a terrible parent, you should not conclude (as the commercial seems to indicate) that if you start eating dinner with your child every night they will begin behaving well.

It seems to me that such cases are particularly problematic in the sort of psychological/ethical cases we are discussing here. If we were to discover a correlation between beastiality and harmful acts, would this really indicate that beastiality leads to such things? It seems to me just ask likely that the stigmatization of beastiality in our culture might make it such that only those who are otherwise troubled psychologically would engage in the activity. (Please note that these are just possibilities, I am not asserting any of this to be the case.) If this were, indeed, the case, then one might find that were beastiality not stigmatized, such a correlation would no longer exist.

Anyway, I think I've said enough here, and I'm quite interested to hear what others think about this issue. In particular, aside from comments about my reasoning, I am curious what alternatives others might offer for discovering the answers to empirical questions such as those raised by Jonathan's argument, assuming they share my worries about statistics.

PLEASE understand that I know little to nothing about statistical analysis, and it may very well be the case that I missed something, or (perhaps the more likely case) that this is an issue that has been discussed and dealt with many times and it just happens to be that anti-drug groups don't bother to worry too much about truth in advertising.

*I say "particular" because showing that someone has sex with sheep and is mean to their brother--or engages in any number of other harmful activities--probably wouldn't tell us much, as many people who are mean to their brothers don't have sex with sheep. I have in mind here specific types of harms, most likely serious crimes, that could be shown to have a particular connection with beastiality.


  1. I'm sure that this bears some very direct resemblance to the violence-in-films/video games debate. Do violent kids tend to watch and play violent media, or does violent media make kids violent?

    I would've thought (although, like you, I'm no social science whizz) that to resolve the above debate you'd need to measure violence levels of kids at an age before they begin to observe the violent media (assuming its possible: is it easy to tell if a 3-year old is violent?). Likewise, in the bestiality example, I'd assume that measuring psychological abnormaties before people get to the appropriate age would allow you to measure the effect of bestiality on mentality.

    (you'd obviously need a large sample in both cases to ensure that some of your sample do indeed end up committing harms/bestiality)

  2. I too have seen that "dinner with kids" advertisement and rolled my eyes. I noticed a similar problem in Plato's Republic, actually, when I wrote my paper last semester. Socrates asserts that one should act justly because (mostly) a just man is happier than an unjust man. But it turns out that to be a just man you need to know the form of the good and have a well-ordered soul (which some people naturally don't have). So only just people should act justly. Everyone else would receive no extra happiness from acting justly.


    I agree with Alex Gregory that to draw solid conclusions from the video games example, you need not just data that says 70% of gamers become violent, but data that says something about how the gamers acted before being violent, and whether it looks like gaming was the cause of the violence. In other words, for it to be a good science experiment, you need to isolate the variable you are trying to test.

    I think it matters somewhat (in terms of ethics, and how we should act) when the statistics, though non-scientific, become overwhelming. For a while we didn't know categorically that HIV was the cause of AIDS (in fact, I'm not positive we know that now). We observed that with a very few exceptions, everyone who had HIV got AIDS. Since we hadn't (or haven't, I'm not sure) proved the causal connection, it was plausible that "the kind of people who got HIV were the kind of people that gots AIDS" but who in their right mind would play chicken with this? No one, because the data was so overwhelming, though causally it was no stronger than the "dinner with the kids" example. If anyone knows more about HIV than I do please correct me, but I think what I described is accurate. (a note: clearly it makes a difference how many people get AIDS without having HIV first. I'm not sure how distinct it is from other auto-immune disorders so I don't know if that is plausible)

    So if 99% of those who practiced beastiality also showed harmful behavior, I might be moved to prohibit beastiality and see what happens. If "certain types of people" still showed harmful behavior without any beastiality, let them have their sheep back.

  3. Funny this popped up. I was just having a conversation with Herman explaining how it would be cool to put together a Centre (sic) for Correlations. We could issue press releases occasionally. Like this:

    January 20, 2006, Bowling Green, Ohio - The analysts at the Centre for Correlations (CC) have found a statistically significant correlation between the regular purchase of small boxes and lung cancer. According to Jenson, founder of the Centre, purchasing one or two boxes per day is strongly correlated with cancer of the lungs, but has also been found to cause emphysema, shortness of breath, and other ailments.

    "These boxes are clearly a major health hazard," remarked Jensen, "and it is a wonder that anyone would sell them."

    The boxes, plainly on display and for sale at gas stations, supermarkets, and kiosks for approximately $4.00 in the Buckeye State and CDN$7.00 in the province of Ontario, come in different colours and are sealed in cellophane.

  4. Another statistical interpretation error (the first being presuming causation on the basis of correlation) is failing to use Bayesian reasoning (either the theory explicitly or some analogue). Studies have found disturbing tendencies amongst doctors to make this error:

    “Here's a story problem about a situation that doctors often encounter: 1% of women at age forty who participate in routine screening have breast cancer. 80% of women with breast cancer will get positive mammographies. 9.6% of women without breast cancer will also get positive mammographies. A woman in this age group had a positive mammography in a routine screening. What is the probability that she actually has breast cancer?

    Only around 15% of doctors get it right (Casscells, Schoenberger, and Grayboys 1978; Eddy 1982; Gigerenzer and Hoffrage 1995; and many other studies.)…On the story problem above, most doctors estimate the probability to be between 70% and 80%, which is wildly incorrect…The correct answer is 7.8%, obtained as follows: Out of 10,000 women, 100 have breast cancer; 80 of those 100 have positive mammographies. From the same 10,000 women, 9,900 will not have breast cancer and of those 9,900 women, 950 will also get positive mammographies. This makes the total number of women with positive mammographies 950+80 or 1,030. Of those 1,030 women with positive mammographies, 80 will have cancer. Expressed as a proportion, this is 80/1,030 or 0.07767 or 7.8%” (E. Yudkowsky (2004) An Intuitive Explanation of Bayesian Reasoning. Retrieved from: