Cognitive Biases and Free Will Part I: The Dual-Process Challenge

I’ve presented challenges to free willism in other posts. For example, I still find the prospect of a “Medicalized Society”, according to which wrongdoers are treated almost like sickly or diseased people, to pose a fascinating challenge.

More recently, I’ve become interested in cognitive biases that seem to affect humans in general. These biases suggest another possibility: the dual process hypothesis. The dual process hypothesis paints a picture of the mind like the following:

1. the mind has both (i) slower, smarter, more generalized (SlowSMG) circuits and (ii) faster, dumber, more domain specific (FastDDS) circuits
2. amongst the FastDDS circuits are circuits relating to identifying other agents,      understanding the causal relationship between events, assigning blame and praise, and the emotions such as anger, thirst-for-revenge, and so on
3. the FastDDS circuits compensate in speed for what they lack in accuracy, and so represent a “knee jerk” response that, upon cooler and more thoughtful reflection, is sometimes mistaken
4. so our prephilosophical or instinctive attitudes and beliefs about other agents,      responsibility and blame, and anger and retribution, etc. are sometimes mistaken or imprudent.

Now, we are all fairly familiar with the dual process challenge because Nichols & Knobe did fascinating studies showing that affect, or the moral salience of an event, seems to increase compatibilist responses at the expense of incompatibilist responses. For example, Nichols & Knobe suggest that “[p]erhaps the most obvious way of explaining the data reported here would be to suggest that strong affective reactions can bias and distort people’s judgments.” This would fit with the dual process challenge: the affective or morally salient situations *trigger* the faster, dumber, more domain specific circuits in our brain (such as those dealing with anger and retribution), which produce less accurate or more imprudent responses.

NOTE: I know some, such as Eddy Nahmias, take issue with this research. I would love to hear any possible explanations for the differences between their data and the Nichols & Knobe data.

Joshua Greene and Jonathan Haidt have done similar research on issues such as consequentialism/deontology, moral realism, and the relationship between the emotions and moral reasoning. For example, I think Greene has found that people who give consequentialist answers give them *slower*, consistent with the hypothesis that the SlowSMG circuitry is trumping the FastDDS circuitry. I think he has also put subjects in MRI scanners and actually mapped how different parts of the brain “light up” with respect to consequentialists and deontologists. And Haidt has done fascinating research on “moral dumbfounding” which suggests that people have FastDDS circuits for moral reasoning and, furthermore, that the more cerebral or cognitive parts of their brain *rationalize* the response, which has already been determined by the FastDDS circuits.

These lines of research suggest future directions for studies on the free will problem. Do those who give incompatibilist or non-realist/eliminativist answers give them slower (like the consequentialists)? When subjects give compatibilist or incompatibilists answers while in an MRI scanner, which parts of their brains “light up”, and can one argue that these parts of the brain are more like SlowSMG circuits or FastDDS circuits?

NOTE: This dual process picture of the mind is a hypothesis. It is an empirical question and cannot be settled by thought experiment (or whatever else philosophers do). But, as a live hypothesis, I think it is worth contemplating the ramifications, should the hypothesis turn out to be true or largely true.

Here are some questions for other Gardeners:

1. Do you tend to think that this dual process theory will turn out to be true or largely true? If not, where do you think it goes wrong and why?
2. Regardless of whether you that this dual process hypothesis is right, what ramifications do you think it would have for the free will debate (libertarianism, compatibilism, non-realism/eliminativism)? Is the existence and relevance of such FastDDS circuits likely to undermine one group more than others?

Comments

[tnadelhoffer]

Kip,

I am familiar with the literature you discuss, although I have always had my doubts concerning how to use the data to support or object to a particular philosophical position. Consider, for instance, Greene's claim that consequentialist judgments are to be preferred to deontological judgments since the latter are driven primarily--if not exclusively--by fast and loose affective cognitive mechanisms whereas the former are driven by more sophisticated "higher" level frontol lobe activity (e.g., ACC and DLPFC). Even if future studies provide further evidence that Greene's model of the cognitive undergirding of consequentialist vs. deontological judgments is correct, what justifies the assumption that because deontological judgments are driven mainly by spontaneous affective mechanisms they are less valuable (or more problematic) than their consequentialist counterparts? After all, from an evolutionary point of view, these affect-driven spontaneous judgments have served us very well in terms of our ability to navigate our complex social structures--something we see quite clearly when we examine proto-moral behavior in non-human primates.
Indeed, if the research by people like Malcolm Gladwell is roughly correct, we may have good reason to prefer the fast and loose spontaneous intuitions more than the slow and deliberative ones. Minimally, I think anyone who wishes to use the fact that a given intuition is produced by a particular kind of underlying cognitive mechanism as a reason for discrediting the intuition needs to explain why some mechanisms are to be priviledged as far as our moral intuitions are concerned. After all, if the recent work on sentimentalism is correct--see, e.g., Nichols' Sentimental Rules--affect plays a critical role in our moral psychology. Why, then, would we think that non-affect driven intuitions about free will and moral responsibility are to be preferred to those intuitions that have been "corrupted" or "biased" by affect?

[Patricia Greenspan]

Let me just mention (underlining the previous comment) that I recently was appalled to read a psychology article a student passed on to me that referred to as the "logical" approach the justification of killing one person in order to save several. A Kantian view, on the other hand, was emotional....

[Kip Werking]

Thomas,

Thanks for commenting on my post. I'm glad you at least understand where I am coming from. I was a little worried that others (especially those who may be less familiar with this literature) would just think that the dual process theory is adsurd or ridiculous.

Let me first offer a caveat and then try to answer your question. The caveat is that I am not familiar with the work by Gladwell or Nichols, although I do want to learn about it because I am fascinated by moral psychology. I do have a passing familiarity with studies showing that, for example, retribution or altruistic punishment often leads to better outcomes for groups (Nichols mentioned this in his discussion paper for the GFP Reading Group).

Now, while I will try to answer your question, I am not the best person to do so. I'm sure someone like Josh Greene would have a more encyclopedic grasp of the relevant literature, and could cite, with more eloquence and sophistications, arguments in favor of my position. That said, here are some points:

1. One point seems almost tautological: FastDDS circuits are, by definition, dumber. Perhaps you are only considering fast, *smart*, domain-specific circuits, in which case you might be talking about a different hypothesis than the one I present. It seems to be a live possibility that such fast, domain specific circuits are smarter or dumber. Knobe and Nichols, for example, have offered the alternative explanation for the affect/compatibilism finding: that affect improves moral competence (so to speak).

One way in which these FastDDS circuits are dumber is in the sense that they are *heuristics* that trade accuracy for speed. Such circuits, on this hypothesis, are willing to get the answer wrong once in a hundred cases (or whatever number), in order to get the answer 50 milliseconds faster (or whatever number). So, it seems strange to me, to cite the *speed* of the circuits in favor of using them ("these affect-driven spontaneous judgments have served us very well"), when what we are concerned with is the *accuracy* of their judgments, and they specifically sacrifice accuracy in favor of speed.

Now, I recognize that there is conflicting evidence about whether affect helps or hurts moral competence. Certainly, there is some evidence that emotions can trigger us to do the less moral thing: a husband comes home, finds his wife sleeping with another man, and in a fit of rage, kills them both. And so this seems to be a subject ripe for empirical investigation.

2. The notion of a heuristic is also important with respect to Darwinian fitness. You wrote: "After all, from an evolutionary point of view, these affect-driven spontaneous judgments have served us very well in terms of our ability to navigate our complex social structures--something we see quite clearly when we examine proto-moral behavior in non-human primates." One problem with this comment, at least on one interpretation of it, is that Mother Nature "is a wicked old witch." There is no shortage of cruelty and viciousness in the animal kingdom, that evolved because it served the respective genes/organisms well. "If he hits you on the nose, always punch him back" and "shoot both the adultering wife and her love" may have served organisms well over deep time. But I don't think the self-serving benefit that these FastDDS circuits might provide is an argument in favor of using or listening to them.

3. It's important to note that all participants (putting aside, for a moment, those with more religious beliefs) in this debate agree that, in a sense, we have split from Mother Nature, according to which we just evolved to maximize fitness and put more genes into the next generation. I, personally, could care less how long my genes live, and I bet you feel the same way. Instead, we probably care about happiness, or flourishing, or some such thing. But, given this break from nature, we cannot expect natural or evolved mechanisms to be a good or infallible guide to happiness, or flourishing, etc.

4. Let me give you a specific example, which I happen to think has some truth to it, about how these heuristics can lead us astray: suppose we have evolved a retributive impulse, and suppose it has served us well over time. Now, further suppose, that if all of our ancestors had just had a "moral pill", which they could have given to aggressors, which would have changed nothing about these aggressors, except for minimizing their future odds of striking again (deterring them just as much as a punch in the nose, inspired by the retributive impulse, would). We can imagine here that we would have evolved to give transgressors the moral pill instead of enacting retribution (suppose it involves less danger and uncertainty than punching someone in the nose).

Here's the point: once we actually do invent moral pills, we won't have the "proper" intuition to use them, if we only (slavishly?) follow intuitions, as produced by these FastDDS mechanisms, that evolved over millions of years, in which moral pills never existed. We would be biased, and the bias exists because of the difference between the Stone Age and the Nuclear Age.

5. One final problem, I will raise is the notion of moral dumbfounding. Something we tend to want (at least in West, but perhaps all over the globe, or at least I would like to think) is for our moral theories to have rational justification. But Haidt's work (putting aside the even more disturbing findings about rationalization) suggests that, at bottom, our moral intuitions are rationally bankrupt. Why is secret, non-hurtful incest between two consenting adults wrong? Well, according to Haidt, people just think it's wrong. Period. There is something disturbing and anti-intellectual (at least to my ear) about our moral theories having such fragile or illusory rational foundations. And if FastDDS circuits are producing these intuitions, then the simplistic and non-rational nature of these intuitions would seem to argue against using or listening to these circuits.

Note: I tend to agree with Josh Greene's moral anti-realism.

[tnadelhoffer]

I suspect that the psychology paper Professor Greenspan alludes to is one of the co-authored papers by Joshua Greene. For what it's worth, I think it is somewhat misleading to characterize Greene's view as one according to which killing one in order to save several is "logical" while the Kantian view is merely "emotional." The main point of the studies is that different kinds of cognitive mechanisms get implicated when people are presented with different kinds of thought experiments (e.g., Thomson's Bystander at the Switch case vs. her Large Man on the Footbridge case). Moreover, even when two people are given the same thought experiment--e.g., Sophie's choice--their answers differ according to which mechanisms happen to get implicated as the result of each participant's unique psychological constitution.

So, for instance, it appears that all non-psycopaths initially have prepotent affective responses to vignettes involving harm to others. In some people, these affect-driven intuitions issue straightaway in moral judgments that tend to be deontological in nature (e.g., never intentionally kill the baby). However, in other people, these prepotent affective responses give rise to cognitive conflict--which implicates both the ACC and the DLPFC. In these cases, the people overcome their initial aversion to harming one in order to save several. As a result, they make moral judgments that are consequentialist rather than deontological.

But it is not quite right to suggest that the consequentialist is making the "logical" decision any more than it is right to categorize the deontological judgments as merely 'emotional'--the picture is markedly more complicated than these labels suggest. The point is just that different kinds of mechanisms are required for different kinds of moral judgments. More specifically, it appears that you cannot make consequentialist judgments without the implication of "higher-level" cognitive mechanisms in the frontal lobe. Deontological intuitions/judgments, on the other hand, appear to require far less sophisticated cognitive processes. Indeed, it appears that the sort of mechanisms required for what Greene et al call deontological intuitions/judgments may be simple enough to be shared by other non-human primates that appear to engage in proto-moral behavior with respect to harm, justice, etc. The capacity to make consequentialist judgments, on the other hand, may be something humans alone have.

Of course, having said all this, it is still unclear to me what principles or premises get us from the data from neuro-ethics to conclusions in moral philosophy with respect to which kinds of judgments or intuitions are normatively the best. Minimally, the is-ought specter rears its ugly head in this context. And I, for one, am not sure Greene or anyone else has adequately addressed the problem.

[RICHARD DOUBLE]

Wouldn’t one expect that deontological reasoning would be faster due to its not needing the calculations inherent in consequentialist thinking? Fewer steps -> faster >< [necessarily] dumber?

[James Gibson]

To add to what Richard Double noted, it was pointed out to me a couple years ago that there is neurological research that suggests lying involves more work by the brain than truth-telling. I wasn't interested enough in the claim to confirm if this is correct; but if that is correct, it also shouldn't be very surprising.

[Kip Werking]

Thomas,

I apologize if I have oversimplified Greene's, or anyone else's, view. That is quite possible, considering that this area of psychology is new to me.

I think Greene agrees with you about is/ought gap. This is from an article he wrote:

"I agree with traditional ethicists that
there is a sharp and crucial distinction
between the ‘is’ of science and the ‘ought’
of ethics, but maintain nonetheless that
science, and neuroscience in particular, can
have profound ethical implications by
providing us with information that will prompt
us to re-evaluate our moral values and our
conceptions of morality."

Greene can respond: we cannot breach the is/ought gap, on the fullest and most natural reading of "ought", and so there is no ought. He's a moral non-realist.

Yet Greene also maintains that, as a practically matter, humans want to go about satisfying various desires, many of which are shared with the community (or even species), and that humans can go about doing this, by recommending either consequentialist or deontological systems of ethics. And he argues that consequentialist systems are preferrable, given these practical considerations.

To help you understand how Greene tried to get from "is" to this weaker-and-more-practical "ought", let me restate the key points of his argument in his paper on free will.

1. Humans have different systems for thinking about animated things like animals, and non-animated things like rocks.
2. As a convenient heuristic, humans developed a system for thinking about other minds as "uncaused causes."
3. If determinism is true, this is false, and even on macro-deterministic or event-causal libertarian pictures, this is largely false.
4. Once we recognize that other minds are not uncaused causes, we can know longer hold them morally responsible.
5. If we can no longer regard other minds as responsible, we must reject deontological rules, which seem to be based upon such responsibility, and instead we can embrace consequentialism, which does not much care about who is "responsible" (in a backward looking sense), so much as it cares about what action produces the best outcome.

Now, I think Greene's account is perhaps oversimplified and overstated. I think multiple biases, evolving for multiple reasons, have made belief in free will popular. And, as Nichols notes, I don't think we always or consistently regard others as totally uncaused causes. But I suspect that in a highly morally salient event (e.g. September 11 or Columbine), we do tend to think of aggressors this way. Such events seem to trigger demonizing processes, which may aggrevate or magnify the Fundamental Attribution Error.

So Greene provides one account, which is more articulate than I anything I had suggested, for how the dual process hypothesis can pose a challenge for compatibilists. And this may come some small way towards answering your questions.

Thanks again for your feedback.

[Tom Clark]

Greene’s reductive non-realist account of morality says that deontological and consequentialist moral theories are in essence avatars for two different brain systems involved in moral decision-making – philosophers’ projections of their own and others’ psychology that emphasize one system at the expense of the other. His research suggests that the two systems (and maybe others for all we know) get differentially activated by situational variables (e.g., on a personal vs. impersonal dimension), and of course individuals differ in the baseline balance between them.

One of the take home messages in a recent talk Greene gave to a non-specialist audience at Harvard was that it’s likely that there's no soul involved in moral decision-making, rather it’s just a matter of the mechanisms involved in these systems. Further, he suggested that it's the idea of the soul that incites the retributive impulse. It's less likely that you'll want to "stick it" to someone once you really understand that they're a biological machine. Neuroscience solidifies this understanding, in which case the neuroscientific mechanistic picture of ourselves might induce us to endorse consequentialism instead of deontological retributivism.

Of course the reason the mechanistic picture might undercut retributive judgments is that it activates the slow, impersonal, cool, cause-and-consequence considering system, while deactivating the hot reactive system that for very practical purposes takes the agent as a soul-like first cause (taking time to consider what caused the offending behavior isn’t adaptive when you’re protecting kith and kin). We can always induce the opposite effect by telling stories of sadistic serial killers.

Both systems are necessary and present in a morally competent agent, but which one do we most want to encourage as a matter of general policy? Looked at naturalistically, a compatibilist deontological retributivism seems to depend on hot reactive attitudes since (by definition) for a retributivist there needn’t be a cool consequentialist justification for inflicting suffering on an offender. Compatibilist retributivists will of course reply that there’s a cool, principled justification for inflicting deontological suffering that doesn’t involve the soul, but what’s really going on, Greene says, is an elaborate apologetics for the retributive impulse. DMR/FW skeptics think that these apologetics ultimately fail, and the issue is how best to manage the retributive impulse, and that’s best done by looking at the consequences of expressing it. The governing consequentialist value here is the minimization of unnecessary suffering. Greene would say (and of course I’m way oversimplifying here) that this whole dispute is driven by two brain systems dukeing it out, with Kant and Mill as their avatars.

[Joe]

Kip,

Interesting posts! But it seems that you’re overall argument encounters at least one problem. I’m interested to see what you think about it.

In point (2) of your initial post you write: “amongst the FastDDS circuits are circuits relating to identifying other agents, UNDERSTANDING THE CAUSAL RELATIONSHIP BETWEEN EVENTS, assigning blame and praise, and the emotions such as anger, thirst-for-revenge, and so on” (my emphasis in caps). You go on, in point (4), to conclude: “so our prephilosophical or instinctive attitudes and beliefs about other agents, responsibility and blame, and anger and retribution, etc. are sometimes mistaken or imprudent.”

But wouldn’t it equally follow that our “understanding of the causal relationship between events” is equally “mistaken or imprudent”? Why not conclude, for instance, that mind-independent causation is “a convenient heuristic” for thinking about the world, given the “simplistic and non-rational nature” of our causal intuitions? But then what does that tell us about the strength of van Inwagen’s consequence argument or Strawson’s ultimacy argument, which at least appear to rely on a more robust view of causality?

I’m having a hard time putting together all of the pieces that one would need in order to take the dual-process challenge seriously.