Automaticity and Moral Judgments

According to rationalist models of moral psychology (e.g., Kohlberg 1969; Piaget 1932/1965), people’s moral judgments are reached via a deliberate, reflective, and conscious process of moral reasoning. Advocates of these models typically reject the emotive or affective theories that one finds in works of Freud, Skinner, and others. Rather than focusing on emotions or intuitions, rationalist models give moral reasoning pride of place in explanations of human moral psychology. And while these models held sway among psychologists for several decades, there is growing evidence that they paint an incomplete picture of how moral judgments are actually made. Work in cognitive science, social psychology, and evolutionary biology, for instance, has shown not only that much of human cognition occurs automatically and outside the scope of consciousness (Bargh and Chartrand 1999), but also that people are often not very adept at describing how they actually reached a particular judgment (Nisbett and Wilson 1977). One of the key notions that underlies much of the recent research on moral psychology is “automaticity” – i.e., “the mind’s ability to solve many problems, including high-level social ones, unconsciously and automatically” (Greene and Haidt 2002: 517). The seeming ubiquity of these automatic mental processes has led some researchers to reject rationalist models of moral psychology in favor of non-rationalist affective models. According to these affective models, “moral judgment is more a matter of emotion and affective intuition than deliberative reasoning” (Greene and Haidt 2002: 517). And while these new explanatory models make room for the type of higher cognition that Kohlberg and Piaget championed, they nevertheless suggest that moral judgments are produced primarily by emotional and non-rational processes rather than deliberative and rational ones. Moreover, moral reasoning is relegated to the role of making post hoc justifications for antecedent moral judgments.

What should philosophers interested in moral philosophy do with this sort of data? How, for instance, does this sort of work hook up with normative ethics? What if, as Jonathan Haidt suggests, Hume beats Plato in the end-at least descriptively?

Comments

[Neil]

A lot of the fuss made about these models of reasoning is based upon a confused idea of what cognition could be like. Of course a great deal of mental processing takes place below the level of conscious awareness. How could it all be conscious? How, for instance, could we assign weight to various considerations explicitly and consciously? In the light of what would such assignments be made? The same kind of confusion underlies a lot of responses to the Libet experiments: what, exactly, are they supposed to threaten? These results are compatible with all the major positions on morality and moral responsibility: our sub-personal mechanisms can be reasons-responsive and reasons-reactive, or can be expressions of our character, or whatever.

Tim Bayne and I have a paper on the role of automaticity in moral accountability forthcoming in the International Review of Psychiatry. Draft here:

http://www.phil.mq.edu.au/staff/tbayne/doing_without_deliberation.pdf

[tnadelhoffer]

Neil seems to have misunderstood the part of automaticity that I was worried about (perhaps my own failure to be clear is to blame). So, let me give it another shot. There are rationalist models of moral psychology just as there are rationalist models of normative ethics. Leaving aside the latter for a moment, I want to consider the likelihood that the former models are false--especially if they entail that people first engage in moral reasoning and then make moral judgments. According to the affective models of Haidt, Alicke, and others, moral reasoning only comes into play when we have to defend our moral judgments to others. The problem with this is that when we engage in moral reasoning in this post hoc way, our reasoning is often biased. So, for instance, we end up with confirmation bias, hindsight bias, outcome bias, blame validation modelling, etc. In all of these cases, people are reluctant to give up their antecedent moral judgments even in the face of what appear to be good reasons for doing so. Consider, for instance, cases of what Haidt calls "moral dumbfoundedness" whereby people are very quick to make moral judgments, but then have extreme difficulties justifying them. Indeed, in many cases, the subjects are somewhat surprised, embarrassed, and frustrated by their own inability to explain why they judged that an action was immoral (e.g. incest between two consenting adults that have taken every necessary precaution against accidental impregnation and who experience no negative emotional side effects from the consensual sex). So, the problem I was trying to bring into focus earlier was not automaticity as such (although I admittedly do not share Neil's views about the threats it poses to free will and moral responsibility)--but rather the way that automaticity affects (and in many cases taints) our ability to engage in unbiased and deliberate moral reasoning even after we have been told of the ubiquity of the cognitive biases that afflict us.

On an unrelated note, Neil's article looks interesting--everyone should check it out. Perhaps its subject matter would make an interesting post in its own right!

[Neil]

Thanks for the clarification, Thomas (Tom?) I'm still not sure that there is a problem for any view of moral psychology that we have antecedent reason to take seriously. For one thing, subject's failure to justify moral judgments could be construed as support for a rationalist view. Suppose justified moral judgments are analogous to justified (say) physical judgments. Most people have physical judgments, but these judgments are sometimes false folk physics. Even when they are true, people have a hard time justifying them. I don't think this cases doubt on the role of reason in physics. Admittedly, I think it is more plausible to give reason a more limited role in moral assessment: perhaps moral judgments are defeasible by reason. In that case, a justified judgment might be one that survives all attacks, by comparison with analogous cases and so on. In that case, subjects' failure to be able to justify their judgments should be just what one would expect. Given that biases are universal, BTW, I think it is most plausible to think of justifying judgments (in morality and in physics) as a social process: we all give it a good go, and we get justified judgments that we then internalise as a result. I judge that X, and my justification for X is the testimony of the experts who examine moral judgments.

On a related note, I have an unpopular response to Haidt's findings that moral judgments vary with socioeconomic status. Suppose (wildly implausibly) that education is good for something. Then we should expect people's judgments to vary with their educational level, and that those with higher SES to have better judgments (perhaps they have reflected on the extent of cross-cultural moral diversity).

[Shai]

"e.g. incest between two consenting adults that have taken every necessary precaution against accidental impregnation and who experience no negative emotional side effects from the consensual sex"

this is a weak example for me. many people will be unable to imagine the conditions realized as they are presented in the paragraph, and so it's hard to know whether one is making a verdict about the residue of the set where those conditions are realized or resistance to the story as it is told. in the latter i'm thinking of reasoning from stereotypes of kinds of actions, or types of people that we often do (i believe the feminist bank teller example used to illustrate errors in probability is the most famous of this type)

[tnadelhoffer]

Shai says that "this is a weak example for me. many people will be unable to imagine the conditions realized as they are presented in the paragraph, and so it's hard to know whether one is making a verdict about the residue of the set where those conditions are realized or resistance to the story as it is told." But in the original studies, it was made very clear that neither of these two things would be a problem. More importantly, once subjects appealed to these sorts of reasons for judging that the incest was immoral and were reminded that these reasons were not applicable in this particular case, they nevertheless maintained that it was immoral even though they had no reason why. This is just the sort of dumbfoundedness I was talking about earlier.

[Alexander Batthyany]

It is good to see philosophers paying attention to social psychology and cognitive science research to probe the validity of philosophical notions of agency, choice and will. It seems to me that presently, many philosophers interested in such topics mainly go for neuroscience, yet the automaticity research may be much more challenging than anything the neurosciences have to offer, philosophically speaking. So thanks for bringing this topic up.
However, my impression is that much of automaticity research, for example most of Bargh and his colleagues’ experiments, tends to overstate the role of automaticity. True, some of the findings in priming research can be truly puzzling, especially those which make use of subliminal priming procedures; here, there seems to be absolutely no role for consciousness for anything: if you can manipulate choice and will. And the effects are sometimes huge (affecting and altering the choices and behaviors of up to 50% of subjects and more) and go contrary to everything we think and believe about ourselves, our choices, actions and their causes. (Of course they do, for they are not consciously accessible.) Enters confabulation.
However, as to the role of moral reasoning, and the underlying chronic attitudes and values of agents vs. the influence of situational cognitive effects, I would like to offer the idea that since primes, moods, cognitive load, etc. are mostly modifications of a situation, a nice way to grasp the true scope of the effects of such manipulations is to have a look back at the person-situation debate (see David Funder’s papers for some excellent discussions on these). Being somewhat attached to virtue ethics, I value the idea that to be a virtous person means to automatically do what one has formerly recognized as worthwhile and good and acted out often enough for it to become a moral habit. This is another kind of automaticity, and one that I find hardly challenging for agency.
Really challenging are only those mechanisms which cause an agent to act against what he normally would think, feel, choose and do without the agent (in this case, for example the subject of a Bargh experiment) consciously being aware of doing so (or at least not knowing why). For the sake of clearity, I call such effects situational automaticity effects as opposed to chronic automaticity (i.e. habits). As to the former – however thrilling the findings of situational automaticity research may be, there are two issues which I think are too rarely addressed in this context. First of all, most of the situations in such experiments are fairly simple and are usually acted out in a fairly mindless way – and precisely because the situations or stimuli in such experiments are so ambigious and undecidable, they allow for a wide within-subject variability. My guess is that such situations never showed much within-subjects stability, and thus the observation as such is no big news; only the cause is now recognized as a cause, and not as mere coincidence, mood, etc. Maybe that's counter-intuitive, but from the viewpoint of conscious agency, there is not much a difference.
Secondly, only recently has priming research begun to look for individual differences in attitudes, values, and dispositions, and found out that the latter do modify, sometimes even eliminate, priming effects (eg. social value orientations and cooperation primes and their relative effects in social dilemma experiments). Almost in each and every study addressing this topic I know so far, individual differences made a considerable difference despite strong manipulations.
While I don’t want to claim that situational automaticity plays no significant role (it does to a huge extent), I also think that due to the fact that individual differences variables are only rarely measured in automaticity research, one easily gets the wrong impression that for example conscious attitudes and the like have no real impact on choice and behavior. Sometimes, that may be the case; but at other times, given the right situation, they do.
In a way, I’ve come to think of situational priming effects as a special case of moral luck.
But I agree that it is time for philosophers and psychologists to cooperate on this important topic; only I cannot exactly pin down my feeling that there might be a bias in automaticity and bias research itself. In any case, the outlook of empirically researching such an age-old philosophical topic is interesting and challenging.

[Hunt]

Research on goals and motivations might be relevant here, in addition to the "implicit attitudes" research. In particular, it's quite clear that people are not, in most cases, aware of the effects of their goals and motivations on their evaluations (and aren't really even directly aware of their current goals and motivations). The classic example is the role that current goals play in decisions about future actions (if I have a goal now, I will often irrationally make decisisions to carry out acts consistent with that goal in the relatively different future, even though it's unlikely that I will still have that goal), including the way that time scales effect decisions (I'm more likely to choose a behavior that's bad for me, but pleasurable, if I'm choosing something to do tonight, but will choose less pleasurable, but better for me behaviors when deciding what I will do 2 weeks from now). Also, the devaluation effect, in which individuals tend to see things not associated with their current goals as less valuable, is something that no one seems to be able to observe in his or her own behavior without first having hard of the devaluation effect.

I find it difficult to believe that people are able to rationally deliberate about the values of particular actions when they simply aren't aware of all of the factors, or even most of the factors, affecting (or determining) their evaluations. Unless, of course, "rational deliberation" just consists of saying, "Well, I think X is a good idea, even though I know that the explanation for why I think so will just be a story I've constructed to explain my positive evaluation of X, so X is morally good."