Problematic perhaps, but not irrational.

• Author: Kelman, Mark
• Position: Response to Cass R. Sunstein, Daniel Kahneman, David Schkade and Ilana Ritov in this issue, p. 1153

1. INTRODUCTION

   Predictably Incoherent Judgments (1) is an impressive work: It guides us both to rethink the acceptability of the processes by which we generate legislative punishments, punitive damages, and administrative sanctions and to reconsider our capacity as individuals to make stable, defensible comparative judgments about just how bad (or good) a particular outcome is.

   In one quite significant respect, I am certain that Sunstein, Kahneman, Schkade, and Ritov's argument is right: Individuals asked either to describe or evaluate a trait or an outcome will do so differently depending on what other events or traits they are thinking about when they are making their descriptions or evaluations. As the authors recognize, this is in part simply a function of linguistic convention: (2) If asked to communicate to some interrogator how they would describe or evaluate a trait or event, individuals may well recognize that the interrogator is (implicitly) asking a category-bound question. Thus, the stand-alone question, "Is an eagle big?" may properly be understood to imply the more complete and tailored question, "Is an eagle big for a bird?" even though the speaker would recognize perfectly well that she could have been explicitly asked, "Is an eagle big compared to most dinosaurs?" or "... to trucks?" etc. (3) Sometimes, the slipperiness or opacity of usually settled linguistic implications are apparent to us. In the weeks after the September 11 terrorist bombings, the simple question calling for an evaluation of a state, "How are you?" was (transparently) opaque and indecipherable: Did it mean how were you setting aside the big event completely, solely in relationship to the big event, overall (including the big event), or "considering" the big event (but not letting the fact that we were all upset force an undifferentiated answer)? (4) One knew that one could readily get confused: The answer "fine"--meant to be "fine, considering"--could readily be misinterpreted as a signal of a dire, perverse lack of sensitivity.

   But the authors are likely right to conclude that it is not simply the way we communicate our underlying judgments of traits and events, but the judgments themselves that are context-specific: For example, while people will typically not misassess the relative size of an eagle and a cabin once they recognize that (in conversational convention terms) they are being asked to compare them to one another (rather than to birds and homes respectively), it is also the case that when asked to judge the size of two identical circles, it is quite difficult not to judge the circle that is seen against a backdrop of smaller circles as bigger than the one that is placed amidst a backdrop of larger ones. (This is true even though there is no sense in which they believe they are being explicitly or implicitly asked to assess the size of either circle relative to the background ones.) Similarly, it appears that the authors are right that an event will seem "worse" (an evaluative, rather than purely descriptive, term) if evaluated against the backdrop of globally trivial bad events than if evaluated against more serious bad events. It is quite credible, too, that the "events" to which we typically compare the outcome we are asked to evaluate are events in (loosely conventionally defined) "similar categories," rather than the whole global range of events we might conceivably judge, at least unless we are explicitly forced to focus on extra-categorical comparison events. If that is true, of course, bad events that are conventionally members (or elements) of "minor harm" category sets will typically be "unduly devalued" and bad events that are members of "severe harm" category sets will be "not devalued enough"--at least so long as "unduly" and "not enough" are defined relative to the judgments that the evaluators would have made had they been comparing the events to a wider, more global range of events outside their "natural," local category.

   Where I depart company from the authors is that I think they are wrong to assume that the person whose judgment of the "badness" of an event changes depending on the context in which she makes her judgment has, in some sense, committed a cognitive error or "been irrational" (in precisely the same way that the person who thinks two identical circles are different in size depending on the backdrop circles in the field is mistaken?). Underlying the authors' assumption here is their unpersuasive claim about the nature of rationality. Let me illuminate this point, at least tentatively, with another example from the period succeeding the initial terrorist assault. I believe the attack (temporarily?) blunted the capacity many of us have to make just the sort of typical, intra-categorical judgments of events in our life that the authors find so troubling. Thus, if your kid cheated at school or if the auto repair shop kept your car too long and charged too much--events that might have filled you with rage and sadness in ordinary times--you were atypically prone to feel, "No biggie. It's nothing compared to the post-terror transformation of my world." My claim is that such global, non-categorical judgments, which serve largely to blunt the emotional significance of most of the events in our ordinary lives because they pale by comparison with true horrors, are best understood as (hopefully temporary) madness, not heightened rationality. People whose response to virtually all of the bad events they confront is akin to "It's not cancer" suffer from blunted affect, and that affect may negatively affect their capacity to respond reasonably.

   If that is the case, I will argue, it would be irrational to adopt mechanisms that made one more prone to respond (unduly) globally. I do not mean to downplay the importance of "seeing things in perspective." Having the insight that one's petty daily woes "aren't cancer" can be important, helpful and rational, but it is by no means obvious, as the authors seem to assume, what the "optimal" level of such perspective might be. The authors' view that the global perspective is always optimal seems to rest entirely on the empirical claim that "local" judgments tend to be unstable when one is confronted with more global comparisons, while (presumably) global reactions when one is (somehow) reminded of the advantages of localism in judgment are more stable. (5) Even if the empirical claim is true, I will argue, the relative instability of local judgments does not render them irrational. Assessing the authors' claim that "local" judgments are irrational (at least when they are unstable) is the topic of Part III of this comment; in many ways, the hesitations I express in that section are my most serious hesitations about the wonderful work the authors are doing here. (6)

   If it is not, in some general sense, irrational to evaluate events relative to "similar" events rather than relative to the whole range of perceptible events, then the live question is whether there are narrow reasons, especially applicable to either levying distinct levels of criminal punishment or punitive damages (or to valuing environmental goods through contingent valuation), that make "local" context-dependent judgments undesirable when performing these particular tasks. There may be specific policy reasons that the relative size of punitive damages should correspond precisely to the relative outrage we would feel learning of the defendant's conduct, comparing that outrage to all known or conceivable outrages (though there may be counterarguments too). But it is a mistake to assume that such reasons carry over to other "policy" realms--for example, in deciding the appropriate response to misbehavior by one's child. There may be good reasons, in terms of "parenting policy," not to mute one's anger (or disappointment) with such misbehavior even though it would seem relatively trivial in a global frame. (This is true even if "extreme," non-muted reactions would be difficult to sustain once one is reminded of more globally salient horrors.) But it seems unlikely that the reasons this is good or bad parenting policy will have much to do with the reasons for taking (or foregoing) a global perspective in the punitive damage context. In Part IV, I discuss, rather critically, the authors' relatively offhand and implicit defense of the desirability of "global" framing in the context of awarding punitive damages.

   The authors suggest, but never directly claim, that individuals will express even identical valuations in an inconsistent fashion over time when there are few agreed-upon or obvious translation metrics that we can use to express these consistent evaluations in relevant terms, so that their actual punishments will be irrationally inconsistent. The intuition behind this proposition is that in the absence of a strong social convention translating one form of judgment (e.g., how bad a pain is) into another form (how many dollars is the pain "worth") that individuals would be consistent only if they could recall the rather arbitrary translation metric they had used on prior occasions. If, on the other hand, there were a stronger social consensus on how to translate from outrage to punishment, they could rely on the more reliable ongoing knowledge of social practice that participants in a culture typically have.

   The evidence that the authors present here, however, does not involve intrapersonal inconsistency of this sort. Instead, it shows that different people, on different juries, for instance, translate (a similar sense of) outrage into penalties differently from one another. That phenomenon of interpersonal inconsistency is plainly a conceptually distinct point. (7) For reasons I will explain in a bit more detail (in Part II), this second fact does not imply incoherence or judgment reversals of the sort the authors worry about, though they consistently imply that it does. If I am...