As I mentioned, I think there are at least three categories of objections to paternalistic intervention, and I wanted to identify empirical research relevant to those objections. The first centers around the traditional claim that people know their own preferences and, as a result, are best able to identify those preferences and make their choices--certainly better than a third party. Oft-repeated, the objection is, of course, associated with John Stuart Mill: "[W]ith respect to his own feelings and circumstances, the most ordinary man or woman has means of knowledge immeasurably surpassing those that can be possessed by anyone else."
A facile response is one used to justify intervention in the first place--"look at the bad decisions people make!" In the health context, the prevalence of obesity has doubled over the last 30 years. Smokers underestimate their risk of dying from lung cancer, despite overestimating the link between smoking and cancer. In the market context, cognitive biases may allow sellers to manipulate buyers; erroneous estimates of the likelihood of one's own breach, or that of the other contracting party, may lead to parties bearing risks that they did not fully consider, or to failure to consider appropriate remedies for breach. And, of course, many people make poor financial decisions in the context of pensions or investing.
But an anti-paternalist responds--among other things--that perhaps people are not being irrational; they in fact are identifying their preferences, and deriving pleasure even from objectively detrimental activities.
One reply points to the burgeoning literature on affective forecasting. Substantial empirical research (recounted lucidly and entertainingly by Dan Gilbert in Stumbling on Happiness) shows that people are surprisingly inaccurate in identifying their emotional reactions to future events, both positive and negative, and as a result have difficulty knowing what makes them happy or unhappy. This research is becoming more familiar in the law, and has been imported into the paternalism discussion (as well as a number of other contexts). Briefly, though, such findings "challenge a fundamental assumption that [supports various] forms of autonomy in decision-making . . . , namely, the assumption that people are able to make choices in their own best interests."
Simply saying that people don't always know what's best for them isn't dispositive, of course, nor does it address the Millian notion that "I know my preferences better than someone else." But there are empirical reasons to think that in some cases, experts--or simply third parties--do better. This may particularly be so in the context of risk perception. For instance, ordinarily, third parties over-estimate the risk preferences of others. However, particularly in the context of negative events, individuals' over-optimism tends to reduce their perceptions of self-relevant risks. In a sense, the over-estimation by third-parties could in fact be relevant to bring estimates closer to the truth. More directly, Hsee and Weber's research shows that when the target of a risk judgment by a third-party is made salient--made visible, for instance, even without actual interaction--the over-estimates become more accurate. Thus, it may be that simply identifying the target individual to the third-party may ameliorate over-estimation by that third-party, and may result in more accurate risk estimation than when the judgment is left to the individual. Finally, of course, is the likelihood that a third-party may be less susceptible to immediately experienced factors that bias decisions (such as mood). More objective, "calculation"-based decision-making tends to ameliorate emotion-based "probability neglect," and might warrant placing decisions in the hands of third parties less susceptible to such immediate emotional biases.
Of course, none of these findings mandates paternalism, nor do they conclusively refute objections to paternalistic intervention. But they are helpful in moving the ball forward and incorporating more ELS information into the discussion. To the extent affective forecasting research suggests that we in fact do not know what makes us happy, and to the extent the risk perception research suggests that perhaps third parties can do better, the Millian criticism of intervention must say more.