Aggregating for lifetime character (and well-being): is there an end-of-life bias?

Newman, Lockhart, and Keil recently published their finding that when judging a person’s overall moral goodness or badness across a lifetime, we seem biased toward the end of life (at Cognition here).  According to this theory, we do not judge the moral qualities of a person’s lifetime character by merely adding up the ‘moral points’ of her individual actions over that life.  Instead, if Scrooge or Andrew Carnegie turns things around at the end of their lives, we will attribute much greater goodness to them across their whole lives because we give greater consideration to what people do at the end of their lives than we give to the rest of their lives, when determining the moral character of a life.  I think that the data presented by Newman et al. are open to at least two other explanations, however.


Fans of theories of well-being will notice a parallel in the debate just described, between theories of how to aggregate for lifetime moral goodness and theories of how to aggregate for lifetime well-being.  Newman et al. are essentially comparing moral analogues of additive theories of lifetime well-being (favored by hedonists, such as Feldman and Kahneman) and a Michael Slote-style time-of-life theory.  Additive theories of lifetime well-being say that lifetime well-being is just the sum of (non-relational) well-being contained in each moment of that life; let’s stipulate that additive theories of lifetime character similarly hold that a person’s lifetime goodness is just the sum of the goodness of each action in that life.  Slote’s time-of-life theory holds, by contrast, that the contribution of a moment to lifetime well-being depends on when it happens (for him, greater contribution attaches to mid-life moments of weal and woe), so it cannot be counted without first weighting it according to temporal location; Newman et al., similarly, are theorizing that, in folk moral assessment, the contribution of an act to lifetime character depends on when in that life the moment happens (greater contribution attaches to end-of-life).

It seems to me that character-analogues of two other theories of aggregating for lifetime well-being can be put to use in the domain of lifetime goodness judgments.  Not only can these two theories explain the data presented by Newman et al.; they also might render those judgments rational.  This is important, because time-of-life theories seem to render our evaluations not just biased, but downright irrational.  Why should the mere location of a point in a life matter to that point’s contribution to lifetime well-being or character?

First, the data: Newman et al. initially presented participants with vignettes about “Jim,” who in one version spends most of his adult life engaging in bad behavior, then later in adulthood turns things around to act morally, and then dies unexpectedly a year later; in another version, the order of Jim’s change is reversed, to go from good to bad; in two other versions, Jim does not undergo any change—he just stays good or bad.  Participants rated Jim’s life nearly as good when he briefly changed to good at the end of his life as when he was good throughout, and similarly for being bad.  This seems to tell against the view that we judge lifetime character additively.  Moreover, it is the end of life that seems to matter in our judgments because if the aberrant good or bad year is placed thirty years before death, Jim doesn’t receive the same ratings.  Finally, if Jim continues living after his change of character, the effect weakens—Jim is considered more moral when he dies than when he just continues living after his change.  Thus Newman et al. settle on a time-of-life theory of folk judgments of lifetime goodness.

But another theory that can explain these data is a character analogue of the narrative theory of well-being favored by David Velleman and Doug Portmore.  According to this theory (roughly, and blurring together differences between Velleman and Portmore), later events can change the meaning of earlier events, and thereby change their contribution to lifetime well-being.  For example, having a costly investment pay off handsomely is better than paying the same costs and losing the investment but simultaneously winning a monetarily equal lottery, because only the former makes your earlier sacrifice a good investment—the value to your lifetime well-being of the earlier event has now increased, even though its momentary value didn’t retroactively change.  Some participants in the Newman et al. studies may have had something similar in mind in the case of Jim: perhaps the reason that Jim is rated so well after a change toward the end of life is that it also changed the significance of his earlier curmudgeonly behavior.  That is, we can read that early behavior as part of a narrative of moral improvement, where the earlier moments were moral costs paid in the service of learning how to be a good person, thereby counting for more than they do in the case of the Jim who is good only in early adulthood and bad for decades after that—earlier good behavior cannot render later behavior sacrifices in the service of learning how to be good.  (In fact, this theory also explains another finding of Newman et al.: Jim’s favorable rating after a positive change is secured when his change is genuine rather than simply motivated by, say, his knowing that he will die soon.)

Another theory that can explain the data is a character analogue of my own loss/gain account of lifetime well-being.  On this account, losing momentary well-being is bad for a life and gaining momentary well-being is good for a life.  Similarly, it may be that having one’s actions go from good to bad (or bad to worse) itself contributes to having a bad lifetime character, while making moral progress itself contributes to good lifetime character; making moral progress is itself morally good and moral backsliding is itself morally bad.  The employment of such a theory would explain the evaluations of Jim’s lifetime character: when he gets better, that itself is good, and to some degree makes up for his earlier failings, which would bring those evaluations closer to evaluations of a Jim who was always good.

We can see why, then, comparing the improved-but-died Jim to the improved-and-continues-living Jim doesn’t tell us enough.  Because “it is…ambiguous whether or not the [improved-and-continues-living] person would change back,” that pair of cases doesn’t discriminate between the effect of Jim’s death, the effect his reform has on the narrative significance of his earlier bad acts (because we don’t know if still-living Jim’s reform is true), or the place of his reform in an overall trend of moral improvement or decline (because we don’t know whether still-living Jim will continue the trend upward).  We need more research, then, to discriminate between whether we ordinarily follow an irrational time-of-life theory of lifetime character, a narrative theory of lifetime character, or an improve/decline theory of lifetime character.  It would also be interesting to see whether the latter two theories could be characterized in a way that makes them seem as justifiable as their well-being counterparts.

20 Replies to “Aggregating for lifetime character (and well-being): is there an end-of-life bias?

  1. What is the point of these overall assessments of a person’s moral goodness or badness? Assessments of overall well-being are dubious too, but at least they are relevant to what we ought to do on some utilitarian theories of that, but assessments of overall goodness or badness seem not just abstract but disconnected from anything that matters or might matter. Maybe making such assessments is like comparing how “influential” or “great” different historical figures were: not really meaningful unless you stipulate some standards for judging.

  2. Hi Josh,
    You talk about how the narrative theory of well-being and the loss/gain theory of lifetime well-being can explain the data that Newman et al. collected regarding reactions to various vignettes. But it seems to me that we must do psychology, not philosophy, to explain people’s reactions. I think that a philosophical theory, such as the narrative theory of well-being, could explain why an uphill life is better than a downhill life even when each contain equal sums of non-relational well-being, but I don’t see how it can explain why people tended to react one way rather than another unless we can just assume that people’s assessments of the levels of lifetime well-being of people that are describe in various vignettes reliably tracts the actual levels of lifetime well-being of people that are describe in those vignettes. Are you making this assumption? Or, if not, what do you mean when you say that these philosophical theories can explain the data?

  3. Doug, right, I was just being careless at a couple of points. To be more precise (as in, say, the second, fourth, and last paragraphs): the psychological claim that the participants in Newman et al’s study might have adopted a narrative theory of character or a decline/improvement theory of character would explain the data. The philosophical claim that such theories are justifiable would make it reasonable for people to adopt those theories. Is it safe to assume that something in those neighborhoods sounds okay to you?

  4. Josh, I feel pretty confident that I can see the difference between the “mere point in time” view and the other views. But I’m less confident that I see the difference between, e.g., your view and Doug’s.
    On one view, all the good in a person’s life is located at one time or another, but the temporally intrinsic properties don’t themselves determine the amount of good. On the other view, the temporally intrinsic properties do determine the amount of good, but there can be additional ‘pattern’ goodness, not located at any particular time.
    But does this different really amount to a real difference in the theories, or might it be just two ways of putting the same thing?

  5. Hi Josh,
    I’m not sure that it is safe to assume that the participants in Newman et al’s study have adopted any theory on such matters. I suspect that most of them have only various intuitions about the vignettes, which they have not reflected upon much, which may or may not be consistent with other intuitions that they have, and which have probably not been systematized into any sort of theory. (I’m assuming that the participants were not philosophers specializing in well-being.)
    Now you also say: “The philosophical claim that such theories are justifiable would make it reasonable for people to adopt those theories.” Isn’t this just tautological? Rationally justified theories are theories that people can be rationally justified in adopting. But, in that case, what does such a tautology have to do with Newman et al’s study? It seems that Newman et al’s study is irrelevant with respect to this tautological claim.
    Moreover, it seems that you need something like the following claim to make Newman et al’s study at relevant to the other claim that you made: The people in Newman et al’s study would tend to adopt only theories that are rationally justifiable. But why should I accept that? Is there reason to think that these participants have research and reflected on these matters and have thereby come to adopt an informed, consistent, and rationally justifiable theory?
    So I’m not seeing what work you take the data from Newman et al’s study to be doing.

  6. Mark,
    I don’t have a full-blown answer as to why overall assessments of lifetime well-being or character matter, but we clearly make them, and we seem to think that it is better to lead a good life than a bad life, so I’m not sure why we should think that they are outright “dubious.” In any case, I think that if we want to examine our practices of judging people, it makes sense to ask, with Newman et al., what are standards for judging whole lives are and whether those standards are rational.

  7. Josh: Shouldn’t you just say that part of what rationally justifies someone in adopting a certain philosophical theory is that it does a better job of accounting for and systematizing that someone’s intuitions than any of the alternative theories do. The point, then, of the data from Newman et al’s study is to establish that many people share the intuitions that your theory is trying to account for. Thus, it shows that the sort of intuitions that your theory is trying to explain are not just idiosyncratic to you. If that’s it, then I fine with your use of the study.

  8. Jamie,
    I’m not sure I know what a ‘temporally intrinsic property’ is. Can you gloss that for me?
    In any case, the difference between Doug’s view and mine comes down to his rejecting and my accepting the value of meaningless gains or losses. So on the additive view that he and I both reject, a life that goes simply from (non-relational) momentary value n to (non-relational) momentary value n+1 has the additive value of n + n + 1. For Doug’s view, that life *may* also have the additional value that comes from the later moment changing the earlier moment’s significance, e.g., making it a worthwhile sacrifice. On my view (which is consistent with accepting Doug’s view but adds a component that Doug doesn’t accept), that life has more than merely additive value just by virtue of getting better, regardless of whether that getting better is meaningful–tied to some earlier sacrifice, say–or meaningless, as when one wins the lottery. Thus on my view meaningless gains themselves have value (as do meaningful gains), on Doug’s view only meaningful gains themselves have value, and on the additive theory no gains themselves have value–there’s just the value of the moments themselves added together. (And on Slote’s view, we weight each non-relational momentary value according to what point in the life it occurs.)

  9. Oh, that difference I do understand.
    Yeah, I’m not sure whether I can say anything helpful about temporally intrinsic properties! I meant, properties the thing has at a time merely in virtue of that time slice, shared by everything that is a duplicate of it at that time. For instance, your weight is temporally intrinsic, but your age is not.
    Here’s what I was worried about. Take the view you’re ascribing to Doug. Now suppose someone says, no, that’s not possible, because all the value in a life has to occur at some time or other. But, he says, Doug’s view is almost right. At the time, it appears that the moment in question has a certain value, but later we discover its value is greater because of the meaning it has in retrospect. So, all the value is indeed located at a moment, but how much each has might be determined only in retrospect.
    So, my worry is that these are not really different views.

  10. Doug,
    If we don’t like talk about the folk “adopting a theory T” (and of course I agree that the folk don’t literally sit around doing a lot of theorizing about this stuff), how about “follow the principle P”? And yes, what you say is a tautology strikes me as one, too (which is why I was hoping you’d agree to it!): the point is just that, as you suggested, to say that people believe P and that it is rational to believe P is to make two different claims. (That is, I was agreeing with you and just trying to expose a more careful way of talking than I used in the original post.)
    I want, in the end, to say two things about the study from Newman et al. First, there are other possible explanations for the data than the two (additive, time-of-life) that they consider, and at least two of those are consistent with their data. Second, those two additional explanations promise greater likelihood of rendering the folk rational than the time-of-life folk theory/principle/pattern-of-response/whatever that they consider. (Thus, I don’t think this, as you suggested: “The people in Newman et al’s study would tend to adopt only theories that are rationally justifiable.” That would be nice, but I’m not confident it’s true.) I also want to make the third claim, not about the Newman et al. study but inspired by it, that just as philosophers have had these conversations about well-being, it might be good for us to have them about character, too, in order to see if the various theories can be shown to be justifiable. Maybe that’s been done, and I just missed it; but in any case the fact that Newman et al. got these interesting results might breathe new life into philosophical reflection on the subject.
    Regarding your final claim, that we want the theory that best accommodates our intuitions, and that psychological studies can evoke relevant intuitions, yes, I agree with that (my stuff on race has leaned heavily on just that kind of experimental philosophy methodology). I’m not sure it vindicates any one theory of well-being here, since this study is about character rather than well-being, and given that their data are consistent with three theories of character, those data don’t vindicate any one theory of character, again. But they do nicely seem to be a strike against additive theories of folk character–to the extent that we want to do the kind of x-phi that you seem to endorse (and that I endorse).

  11. Jamie,
    Okay, I think I’m getting closer to understanding your concern. It may be that your envisioned reply is just going to require a repackaging of the difference between (say) Feldman, Slote, Doug, and myself, rather than a flattening of that difference. If one is tempted to say “At the time, it appears that the moment in question has a certain value, but later we discover its value is greater because of the meaning it has in retrospect,” this is because one is packing relational value into the value of a moment, whereas in order to get some traction in the discussion, Doug and I (and others) have been putting it in terms of non-relational value–value the moment would have if it existed alone. That’s just a handy way of isolating whether the relationship between moments can impact lifetime/episodic well-being in addition to the value the moments have in isolation, contrary to the theorist who wants to merely add the non-relational value (say, in hedons) of all moments into a sum. If we want to use this alternative way of talking, though, we could instead talk about whether moments have only non-relational value, as additive theorists say, or also have relational value, as Doug and I say, and then we’re off trying to sort out which kinds of relational value count.

  12. Josh: I wasn’t sure that the results of the experiment demonstrate that subjects of the experiment weren’t simply adding virtues. The vignettes, as described, seem vague and subjects might just be attributing virtues and vices to characters that are not described in the vignettes.
    They seem to endorse:
    C: A person, P, who is bad at time t for duration d and good at t2 for d is typically morally better than another person, P2, who is good at t for d and bad at t2 for d.
    One explanation why they may endorse this claim is that a natural explanation why P became good at t2 is that P had some virtue at t that we have no reason to attribute to P2 at t2. Another is that a natural explanation why P2 became bad at t2 is that P2 had some vice at t that we have no reason to attribute to P at t2.
    We need an explanation why P became good. That he had some virtue at t that was exercised in the process of becoming good is a plausible explanation. Similarly, we need an explanation why P2 became bad. That he had some vice at t provides a plausible explanation. Perhaps those responding to the experiment tended to attribute these virtues and vices to people. If so, their judgements may exemplify an additive view of moral goodness.

  13. Victor,
    A few notes on that interesting hypothesis. Your suggestion — let’s call it the ‘projection-of-virtues’ explanation — actually sounds kind of close to the one put forth by Newman and colleagues, in their elaboration on the time-of-life theory they attribute to the participants. The principle they attribute to their participants is that Jim’s “true self” is revealed in his end-of-life conversion, and so, yes, the participants may think that he was bad (vicious) all along or good (virtuous) all along. One thing fans of this theory would have to explain, though, is why the favorable rating only attaches to *end* of life good behavior, and not behavior thirty years prior: if the virtue can be covered up before becoming good at the end of life, couldn’t it be covered up after it was exposed earlier, too? The way that Newman et al account for the asymmetry in their results, of course, is by saying we simply have an end-of-life bias, and one advantage of the narrative and decline/improvement theories is that they can explain the asymmetry as well. I’m not sure whether/how a projection-of-virtues theory can handle that result as elegantly.
    Also, this is just a terminological clarification, but note that the projection-of-virtues theory isn’t an additive theory as defined above. Additive theories in that sense determine the goodness of the whole life by simply adding together the goodness of the actions in that life (without weighting for time-of-life). The projection-of-virtues theory adds to additive goodness a quotient of character goodness for virtue that was there but latent during the years of bad acting (and similarly for vice).
    One other clarification: I don’t think that the participants endorsed your principle C, because the length of time for the good period and the bad period — “d” in your principle C — is not equal. Instead, it’s 29 good/bad years followed by 1 bad/good year. (By the way, for the curious, the vignettes can be read in their paper, linked to above, in Appendix A.)

  14. Thanks Josh
    I’ll have to read the vignettes more carefully when I have a moment. Thanks for rightly correcting inaccuracies in my description.
    In the meantime, consider a virtue like ‘learning from experience’
    A person with this virtue reflects on her own actions, correcting her tendency to act on bad inclinations and reinforcing her tendency to act on good inclinations. This seems to be a virtue, but it seems exhibited only by agents who continue to improve in their other actions. Improving oneself in these ways also seems to be a kind of action. P could have had and exercised this virtue. P2 could not. The case where the person acts well thirty years prior and then fails to act well subsequently seems to lack the virtue.

  15. Victor,
    Yes, I think that would probably explain the asymmetry. So if I’m following you now, it’s not just having any allegedly latent virtue, but having the specific virtue of moral improvement, that gets ‘extra points.’ In fact, that’s sounding kind of like a virtue-theoretical variation on the possibility that the raters might (in a character decline/improvement analogue of the loss/gain approach to well-being) give Improving Jim such a high rating because there is goodness in moral improvement itself, beyond what goodness lies in each moral action.

  16. Right. My thought was that the theory could be additive whereas you contrasted the additive theory (which, in the post, you suggested might have trouble explaining the data) with the narrative theory and the loss gain theory (which might explain it).
    You described additive theories as:
    a person’s lifetime goodness is just the sum of the goodness of each action in that life
    acts of moral improvement can be added to the other acts in the life. I take it that omissions must count too. letting oneself (morally) go to the dogs, or be corrupted, might count as an especially bad omission. it seems that an additive approach might explain the data.
    Perhaps we need to know more about the non-relational character of the additive view. I don’t know as much about the well-being literature as I ought to, but I take it that feeling happy that I was happy earlier in my life advances my well-being on the Feldman view. Similarly, acting to correct my tendency to act badly should also count as advancing my goodness on the additive view of goodness. I take it that the non-relational character of the additive view is simply that the mere fact that the good act is later should not make a difference to its importance to the goodness of a life. this leaves open the possibility that certain good deeds that are important to one’s overall goodness (like rectifying one’s bad character, or paying compensation for harm wrongfully caused) can only occur later in life than the bad deeds or character to which they relate. these features might explain the judgements made in the cases without the fancy feature of the narrative view, which (as described) discounts the importance earlier wrongful acts to some extent because they are in the service of improvement.

  17. Victor,
    That’s an interesting suggestion. So now if I’m following, the good rating isn’t due to the fact that Jim has a (latent) virtue at t, or due to the fact of improvement itself, but instead due to the fact of performing acts of moral improvement, which are just weighted more heavily than other good acts in an additive formula of lifetime goodness (to skyrocket Improved Jim up towards the level that at which Always Good Jim is rated). If I’ve got that right, let me just float a couple of follow-ups.
    First, I wonder if the ‘boost’ an act of moral improvement gets can really be so heavily weighted: why should an act of improvement, say, rectifying past misdeeds or seeking moral therapy (what exactly is an act of moral improvement? I’m not sure…okay, back to the point at hand), be so much more significant than Always Good Jim’s regular good acts, like helping needy people? To be sure (and anticipate Doug), there are two possible questions here: (1) why should we think that the two classes of good act are so different in importance, and (2) is it plausible to project that theory onto the participants?
    (2) gets me to my second question: it’s not clear that Improved Jim’s good actions would be likely read as acts of moral improvement per se, given how the vignettes were worded: he raised employee salaries, donated to charity, etc. These are straightforwardly read as just ‘regular’ good acts, rather than acts of improvement themselves.
    Conveniently, this second hesitation could be tested, by running the same study and adding another vignette that specifies that Jim’s acts in a way that are clearly acts of moral improvement, to see if that kind of Improved Jim gets a different rating that the regular Improved Jim.

  18. Thanks Josh.
    I agree that it’s puzzling why participants seemed to think that many years of wrongdoing have little weight in rendering a person’s life morally bad when she becomes good right at the end. I wasn’t sure that any of the theories you proposed do better than the additive theory in explaining this. For example, the loss/gain account was described as:
    “having one’s actions go from good to bad (or bad to worse) itself contributes to having a bad lifetime character, while making moral progress itself contributes to good lifetime character; making moral progress is itself morally good and moral backsliding is itself morally bad.”
    One version of the additive account might be similar to this view except that moral progress/backsliding are not things that happen to a person, they are actions.
    It seems no more plausible that non-action-guided progress can be sufficiently important almost to completely negate 30 years of wrongdoing than it is that action-guided progress can be sufficiently important almost to completely negate the significance 30 years of wrongdoing. What explanation could be given on the loss/gain account for this judgement?
    I should say that I don’t find the judgements of the participants very intuitive with respect to the lack of weight they give to 30 years of wrongdoing where the person improves at the end, at least if the question is ‘how good a life did the person lead’.

  19. Victor,
    We’re agreed on the puzzle bit! As I said at the outset, it would be good to see philosophers spend more time on the question of whether it is rational to think that the shape of moral development can impact lifetime goodness.
    When it comes to explaining the data from Newman and colleagues, though, I do suspect that (psychological versions of) the time-of-life, narrative, and decline/improve explanations can do better than the theory that the improvement is in the acts themselves–that is, that Improved Jim gets a high rating because he is performing a bunch of (heavily weighted) acts of moral improvement. For, again, he is not described in the vignettes as performing such acts; his good acts are simply ‘regular’ good acts. Of course, it’s not an impossible theory: perhaps the participants assumed that he must have performed a bunch of acts of moral improvement if he were to perform regular good acts, too. So okay, if you’re just saying saying that an additive explanation could make some move to accommodate the data, right. But, and maybe I’m just being closed-minded here, that seems to me like more of a stretch than a narrative, decline/improve, or time-of-life bias read on their judgments. (And again, someone could test at least some of these explanations experimentally.)
    To switch back to the philosophical question for a second, note also that there is one difference between the Improvement Acts theory of what might make the data reasonable and the decline/improve theory (just to be clear, of *reasonableness* now, not psychological frequency). The Improvement Acts theory weights improvement actions really heavily compared to other good actions, comparing apples-to-apples in a somewhat counterintuitive way. The decline/improve theory is factoring in the trend of a person’s actions independently of the actions themselves, in an apples-to-oranges comparison that, because it compares different kinds of things, strikes me as less counterintuitive.

Comments are closed.