Jan 29 2019

The Law of Unintended Consequences

Psychologists have come to recognize that, because of the complexity of human emotion and behavior, we are often motivated to engage in activity which produces the exact opposite effect that we intend. If you are fearful of losing someone, you may become clingy and possessive, driving them away.

The same is true on a societal level – interventions designed to have one effect may have the opposite effect if we are not careful. A classic example is the “scared straight” approach to public service announcements – it doesn’t work. In fact, it may have the opposite of the intended effect. Warning kids about the dangers of alcohol, for example, may just romanticize alcohol use and suggest that it is more popular or common than it is, creating social pressure to use. This is the main idea behind the social norming approach – tell kids, instead, statistics about how few of their peers are getting drunk regularly, reducing the social pressure to use.

This overall pattern is fairly consistent in the literature (although, of course, researching such questions is complex and the details matter to the outcome). Another recent example is a study which finds that fat shaming obese people does not motivate them to lose weight, which is sometimes the motivation (or at least the justification) of the person doing the fat shaming. Rather, fat shaming leads to more weight gain.

Another study from November 2018 is making the social media rounds which also reflects this basic principle of unintended consequences –

Federal abstinence-only funding had no effect on adolescent birthrates overall but displayed a perverse effect, increasing adolescent birthrates in conservative states. Adolescent pregnancy–prevention and sexuality education funding eclipsed this effect, reducing adolescent birthrates in those states.

So in conservative states, abstinence-only education actually increased adolescent birthrates, meanwhile sex education reduced adolescent birthrates. The conventional interpretation of these and similar results are that by teaching children about sex you demystify it, while encouraging abstinence just increases the allure of “forbidden fruit.” This interpretation is also probably simplistic, but it is consistent with the data. At the very least the notion that teaching kids about sex will encourage them to engage in it (rather than empower them to make better decisions) is naive, only surpassed by the notion that encouraging children to be abstinent will actually work.

While it is tempting to also interpret these results as having something to do with the ideology of the states studies (and this was the hypothesis being tested) there are many possible confounders. The authors did try to control for “state-level confounders” but it is difficult to anticipate them all (as the history of psychological research shows over and over). For example, the teen birthrate is significantly higher in rural areas vs urban areas, and rural areas are also more conservative.

There is another article making the rounds that relates to a similar misconception about how psychology works – the relationship between violence in movies and video games and violent behavior. The primary point in this article is that over the last few decades violence in movies has been increasing, while overall crime rates have been decreasing.

The evidence is fairly objective that movies are getting more violent – a PG-13 movie today is more equivalent to an R-rated movie in the 1980s in terms of violence (not necessarily nudity – another factor that gets the higher rating). The authors of the new study conclude:

Raw correlations suggest that PG-13 rated movie violence is inversely related to actual violence in society. However, controlling for autocorrelations suggests that the best interpretation is that PG-13 rated movie violence is unrelated to violence in society. Caution is advised for scholars to avoid implying that PG-13 rated movie violence may have a causal effect on crime in society.

So the best current conclusion is that movie violence has no effect on societal violence, but the authors are correct to caution about drawing any firm conclusions because the question of what causes societal violence is horrifically complex. We still can’t fully explain why crime has been generally decreasing since the 1990s.

The lesson is all of this is that we need to be exceedingly cautious when drawing simplistic conclusions about human psychology and behavior. Naive notions of cause and effect are likely not to be true, and in fact may be the opposite of what our gut tells us. This not only applies to our personal decisions, but to public policy. It’s easy to create perverse incentives, for example. We also cannot assume that people are generally rational actors – people are emotional and social creatures with a complex web of motivating factors. Predicting how that will all play out is like predicting the weather.

What this means practically is that care should be taken when crafting policy. Up front it should be as evidence-based as possible. But even then, predicting effects will be difficult. Therefore, new policies intended to have a specific effect should build in a review period in which objective evidence is gathered to determine the net effect.

This is basically what we do in medicine – we may think we can predict the net effect of an intervention, but we know from experience that we have to study net effects and then make decisions based on this evidence. Otherwise you are not practicing medicine – you are practicing witchcraft. Imagine if the same were generally true for government policy – evidence-based government. This doesn’t mean there is no role for ideology, which determines priorities and value judgments. But it does mean that we do our best to ensure that a new policy will have the effect we intend, whatever that is.

No responses yet