Wisdom grantees Michael Sargent and Shabnam Mousavi examine the question.
Anchoring Judgment in Wise Principles
Michael J. Sargent, Bates College, United States
I’ve been asked to write a blog entry about a related pair of questions: “Can the unwise recognize wisdom?” and “Can one act wisely without being conscious of it?” In thinking about this pair of issues, I suspect that the answers to these questions are, respectively, “Yes, at least in principle” and “Absolutely.” In other words, even the unwise may recognize what is wise, even though they fail to make wise choices. And there may be ways of helping them make wiser choices, even without their awareness. In other words, there may be ways to help them help themselves.
Whatever one’s definition of wisdom, even the unwise probably have an in-principle understanding of wisdom in many domains. They understand that saving for retirement is probably essential if one wants to avoid poverty in one’s old age. They understand that reducing caloric intake and exercising regularly are wise if one wants to be physically healthy. They understand these things in principle, and so at least in that sense, even the unwise can recognize wisdom.
Still, as evidenced by low savings rates and high rates of obesity (at least in the U.S.), applying these principles is often difficult. What can be done to facilitate individuals’ acting upon the principles whose wisdom they explicitly endorse, even if they would be unlikely to apply those principles under normal conditions?
One answer that Richard Thaler and Cass Sunstein propose is to “nudge” individuals’ behavior in directions that benefit them and others, which could be thought of as wise. By attending carefully to choice architecture, policymakers can influence individuals’ behavior, even as those individuals retain freedom of action. A well known example that Thaler and Sunstein describe is implementing opt-out policies to promote participation in retirement plans. In contrast to conventional approaches whereby the default is not to participate in a plan unless one chooses to, opt-out policies set the default so that employees are automatically enrolled in a plan unless they choose not to be. As Thaler and Sunstein note, such opt-out policies raise the level of participation in retirement plans. On the basis of such findings, they argue for an approach termed libertarian paternalism: Individuals retain their freedom to act as they choose, but policymakers structure the choice context in a way that is intended to promote individual and social welfare.
In my research, I’ve recently been exploring applications of such ideas to the study of attitudes toward punishing criminals. Obviously, criminal wrongdoing is a ubiquitous problem, one that societies often attempt to address, at least in part, through punishment. A reasonable question to ask is whether societies’ penal systems are constructed in a manner that is wise. What would a wise penal system look like? And will citizens support it?
Arguably, a wise penal system will optimize the deterrent effect of punishment. In the case of general deterrence, the goal of punishment is to use the punishment of a particular criminal as an example to deter other would-be criminals from the same sort of crime. Punishment is used to ensure that the expected value of punishment (i.e., the probability of punishment × the magnitude of punishment) is kept at a high enough level to deter potential criminals from the same crime as those who have already been punished. A key factor that ought to therefore influence sentences is the likelihood of detection. For crimes that are hard to detect, punishments should be set relatively high. Because the low likelihood of detection of the crime entails a lower likelihood of punishment for the criminal, a greater magnitude of punishment is needed to maintain the same expected value of punishment.
As I mentioned last June, research by social psychologist Kevin Carlsmith and others suggests that people often endorse deterrence in principle, but rarely apply it in practice. For example, Carlsmith and others have found that manipulations of detection-likelihood generally make little difference in recommended sentences of specific, hypothetical criminals. Crimes that are hard to detect are punished no more than harshly than crimes that are easy to detect. Similar results occur when individuals are asked to evaluate general policies intended to optimize deterrence. One group of researchers found that law students at the University of Chicago (who were well versed in deterrence) were generally opposed to policies that delivered lower penalties when crimes were easy to detect than when they were hard to detect, whether the agent meting out the different penalties was bureaucratic (the IRS) or judicial (a judge). Thus, even individuals with expertise seemed reluctant to embrace policies based on deterrence. Data such as these have led researchers to conclude that people’s intuitions are rooted in retribution and not deterrence.
These findings raise many questions that have motivated much of my work under the grant. One question is when individuals are willing and able to apply deterrence theory. Specifically, when will they recommend more severe penalties for crimes that are hard to detect? Like others (including Carlsmith), I have found it difficult to identify circumstances when this occurs.
In one of my recent studies, though, I’ve found that a familiar phenomenon known as anchoring can be used to produce differences between those individuals exposed to high detection-likelihood crimes and those exposed to low detection-likelihood crimes. Anchoring (famously described by Tversky and Kahneman) describes a way of making judgments under uncertainty where one begins with a specific starting point and adjusts from it. A classic demonstration is the Mississippi River problem. Half of a sample is asked whether the Mississippi River is longer or shorter than 500 miles (most say longer) and then to estimate its length. The remaining individuals are asked if it’s longer or shorter than 5,000 miles (most say shorter) and then asked to estimate its length. Individuals in the second group generate larger estimates (typically differing by over 1,000 miles). The assumption is that the initial value that’s mentioned acts as an anchor from which participants adjust to get to a final judgment. But, critically, different starting points lead to different outcomes.
I’ve begun applying this framework to studies of deterrence and detection-likelihood. For example, in one study participants in the low detection-likelihood condition read about a case of embezzlement in which the crime was described as hard to detect. Critically, they were also told that in such cases state governments recommended a sentence of 15 years (a high value on the sentencing scale they would ultimately use to report their recommended sentence). By contrast, other participants read about the same case of embezzlement but were told that it was easy to detect and that, as a consequence, state governments only recommended a sentence of 2 weeks (a low value). In short, each group was given an anchor, in addition to information about detection-likelihood. In this study, participants in the first group recommended higher sentences than those in the second group. This was true, even after dropping from analysis those participants whose recommended sentence was equivalent to the anchor, arguing against the possibility that the effect was driven by participants merely repeating the number that had just been presented.
Of course, anchoring effects are highly robust, as anyone who has used the Mississippi River example as a teaching demonstration can attest. Consequently, it would have been surprising not to obtain an anchoring effect. What we are more interested in is our next step, which is to recruit a new set of participants and describe to them the sentencing disparity obtained in our first anchoring study. We also plan to experimentally vary our description of the process that produced such a disparity. A key condition will be one in which we attribute the discrepancy to jurors whose starting point for deliberations on a sentence was recommendations that penalized hard-to-detect crimes more severely. (Their freedom to ignore the recommendations will be emphasized.) Will being told that a jury of their peers decided to punish hard-to-detect crimes more severely lead to more acceptance of such a discrepancy than if it is the work of a bureaucracy or an individual judge? Might it even lead to majority support for such discrepancies in penalties?
We don’t yet know what we will find in the second stage of this work, but if we did find this result, it could have implications for the issue of popular support for public policies that advance deterrence. Previous scholars have suggested that, because citizens are intuitive retributivists, they will not accept policies that optimize deterrence. In the words of the research team who studied the University of Chicago law students, “the fact that optimal deterrence policies are rejected in both the administrative and judicial domains among a group likely to be predisposed in their favor strongly suggests that any effort to move in the direction of optimal deterrence would encounter significant popular resistance” (Sunstein et al., 2000, p. 248). Perhaps this is true under many conditions. But if we find that attribution of a sentencing disparity to jury behavior legitimizes that disparity, even if the juries’ decisions were clearly rooted in numerical anchors provided to them, then it would suggest one condition under which citizens may accept optimal deterrence policies. Despite the fact that such policies are at odds with citizens’ intuitions about how punishment should be applied, if the disparity is said to be due to the free choice of their fellow citizens perhaps they will be accepted.
Again, it is yet to be determined what we will find. But the data we already have shows that, whatever individuals’ willingness or capacity to apply deterrence theory in an explicit fashion, they can be induced through anchoring effects to behave “as if” they apply it. Insofar as optimizing deterrence through penalties is wise sentencing, then this would be a case of individuals adhering to wise principles despite a lack of awareness of what they are doing. As with Sunstein and Thaler’s work, it would suggest that the choice context can be structured to nudge individuals toward wise choices, even without their awareness. Additionally, where juries play a role in setting penalties for offenders, it would suggest that optimal deterrence may be legitimized in the public’s eye by their influence, even if anchoring strategies are used to influence their decisions.
My students are just finishing the online survey that we will use to collect the data for the second part of this project, so time will soon tell what our findings are. In the interim, though, I welcome comments and questions.
References
Carlsmith, K. M., Darley, J. M., & Robinson, P. H. (2002). Why do we punish? Deterrence and just deserts as motives for punishment. Journal of Personality and Social Psychology, 83, 284-298.
Sunstein, C. R., Schkade, D., & Kahneman, D. (2000). Do people want optimal deterrence? Journal of Legal Studies, 29, 237-253.
Thaler, R. H., & Sunstein, C. R. (2009). Nudge: Improving decisions about health, wealth, and happiness. New York: Penguin Books.
Tversky, A., & Kahneman, D. (1974). Judgment under uncertainty: Heuristics and biases. Science, 185, 1124-1131.
Is Wisdom Intuitive?
Shabnam Mousavi, Max Planck Institute for Human Development, Germany
IS WISDOM INTUITIVE? From what I have read and heard so far, it is often at least assumed that this is the case. What is wisdom? We do not know exactly, and we do not aim to provide a universal definition. But we all agree that wisdom is good, useful, and worth exploring. Can we measure wisdom? Yes, and we do so by eliciting people’s judgments of their own wisdom (self-assessment), or their judgments of others’ wisdom (asking for nominations). Why do we want to measure wisdom? To specify this admirable human capability, unravel its secrets, and obtain more of it; to make it accessible to all; to learn from wise decisions and to develop wise strategies. We want to learn about wisdom to spread it and to expand it. In sum, we all seem to agree that everybody would be better off if we could produce more of this indefinable wisdom; if we were to increase our wisdom scores. Once again, a rational principle prevails: “more is (of course) better.”
SO, PEOPLE’S JUDGMENT is the basis for discovering and specifying wise people, whose behavior is then studied to unravel wisdom. This practice must assume that people are reliable in judging, observing, and identifying wisdom, even though they are not necessarily wise. However, some researchers would not, for instance, accept a self-nomination when seeking nominations for wise people. This takes me back to my first question in one step: How are people able to identify wisdom? Is it intuitive? If yes, why can’t they judge their own wisdom? Is it because this recognition is based on an intuition limited to outside oneself? If so, how can we rely on any kind of self-assessment? If one requires some primary level of wisdom to identify wisdom, do we then face circularity? As if these questions aren’t challenging enough, allow me to throw in another twist: We should not forget that people have been shown to suffer from persistent judgmental biases, and irrationalities. After all, many scientists have dedicated themselves to the task of finding treatments for this widespread captivity of human kind to irrationality.
Wrestling with these questions, I couldn’t help but wonder: Can I imagine a way of specifying and understanding wise actions that can avoid these difficulties? I have put my bets on one idea: A study of wisdom as heuristic processes, which rely on intuitive inference. Allow me to sketch this idea here. (And of course, I need a reference, but I’m doing my best to avoid its anchoring effect.) The RATIONAL AGENT has been a popular scientific reference for judging, modeling, and theorizing human behavior and choice. This agent is axiomatized as omniscient in overcoming epistemological uncertainty, as inferring statistically and as complying to logical truth (-tables). I define a WISE AGENT as one that is comfortable with ontological uncertainty through flexibility of intuition and robustness of simple heuristic strategies. This agent makes good decisions based on intuitive inferences and ecological reasoning. I define ACTION as a mapping between representation of information and matching/triggered heuristic strategy. I replace, (1) The information set by representation of information; and avoid the necessity of pre-specifying goals (that cause reduction of actual situation to a solvable form) by specifying a relevant/triggered heuristic strategy; (2) Statistical inference by intuitive inference and gain flexibility; (3) Logical truth with ecological reasoning to rethink habitual norms of behavior and to develop requirements (and methods) for construction of content-sensitive norms. So far I have one suspected corollary: The axiomatization of a wise agent does not necessitate the imposition of invariance. That is, replacements (1), (2), and (3) above remove invariance from the list of requirements for the study of human choice behavior framed with reference to wise actions.
Am I too ambitious in expecting this pursuit to produce a primary framework for the study of actual human behavior (including wise action)? Well, of course, time will tell…. But maybe in this case “faster is better”? In the interest of wise academic practice: Can anyone help me get wiser? Can you give me a good (or ‘good enough,’ because we are all boundedly rational) reason for stopping this imaginative inquiry?
Photo from Flickr Creative Commons.
[]
[]