Sommers on Responsibility and Evolution

Tamler Sommers, in “The Illusion of Freedom Evolves” (posted here on April 11th), aims to describe the evolutionary origins of the belief in moral responsibility. Sommers argues that understanding the origins of our beliefs, moreover, will support the idea that moral responsibility is an illusion; Sommers believes that evolution “explains away” responsibility.

I am not convinced, however, that evolutionary considerations provide any independent support for scepticism about moral responsibility. I will try and say why.

Sommers’ argument seems to go something as follows.

(1) Ceteris paribus, it is irrational to cooperate in prisoner’s dilemmas.
(2) Cooperation enhances survival, so some organisms evolved certain (reactive) emotions that “force” them, irrationally, to cooperate in prisoner’s dilemmas.
(3) Rational organisms, however, are liable to notice the irrationality of the reactive emotions, and repress them. As a result, such organisms are likely to evolve certain entrenched irrational beliefs (such as the belief in free will) that rationalise the reactive emotions.

I will comment on each step of the argument in turn:

(1) It isn’t obvious that cooperating in one-off prisoner’s dilemmas is irrational. (See the work of people such as David Gauthier here, for instance). Moreover, most people seem to acknowledge that adopting certain cooperative strategies (such as the tit-for-tat strategy) in iterated dilemmas is rational.

(2) If it is rational to adopt the tit-for-tat strategy in iterated games, then the reactive emotions (which incline us to use the strategy) shouldn’t be deemed irrational.

(3) In claiming that the reactive emotions may require the support of certain (false) beliefs, Sommers presupposes that these emotions have cognitive content – that they purport to represent certain features of the world. However, Sommers doesn’t consider the possibility that the reactive emotions are fundamentally non-cognitive – that the question of rationality or irrationality doesn’t apply to them at all. (On this view, it is rational to cooperate, given the reactive emotions, but it is neither rational nor irrational to have them in the first place.)
     Even if the reactive attitudes do have cognitive content, Sommers doesn’t say why the beliefs that we evolved to rationalise them are false. There appears to be an equivocation between the question of the rationality of the reactive attitudes, and the rationality of the beliefs that putatively justify them.

Comments

[Neil]

For what it's worth, I think the reactive attitudes are partly cognitive (and of course cognitivism about emotions seems to be the dominant position these days). But I just don't see how step 3 is supposed to work. Suppose it is rational for me to defect, but I feel negative emotions about defecting. I also believe that we have free will. Surely I then just think that it's up to me (categorically) whether I defect or not? Belief in free will doesn't make me more likely to cooperate. If anything, the opposite might be true: believing that my emotions have causal effects that I am powerless to resist, I do not attempt to resist (there is some evidence that this kind of belief really has world effects; addicts who think that they are powerless in face of their desires are more liable to relapse).

Wouldn't (3*) a belief in the objectivity of morality be a more a likely mechanism for entrenching the dispositon to cooperate?

[Kip Werking]

I'm not familiar with Gauthier's work but I suspect that under certain specifications (including what "rational" means, and that the PD is not iterated), cooperating *is* irrational. Let's call these PDIs.

Of course, the claim that situations in the EEA were like a PDI (assuming, as seems plausible, that such things can exist) is an empirical claim, and neither Sommer's suggestion that it is true, nor Neil's suggestion that it is false, is final. And so Sommer's idea remains, for me, an intriguing possibility. I have recently grown more confident that the popularity of free will, despite (what seem to me) valid proofs against its existence (according to at least some defintions, which strike me as the most relevant and interesting), suggests that this popular illusion represents a real lacuna in our psyche, and so invites an evolutionary explanation. I realize that evolutionary psychology has its critics, but I'm fairly familiar and enthusiastic about its claims. And I don't think there are enough evolutionary approaches to philosophy (Dennett's Freedom Evolves being a notable exception).

Personally, I would explore the evolutionary angle connecting the free will "illusion" with the Fundamental Attribution Error (also called the Correspondence Bias, the two are distinct, but it's not clear to me how), instead of with people's tendency to defect in PD's despite the alleged irrationality of doing so. Unlike PD-defection, the Fundamental Attribution Error is uncontroversially irrational. I have an evolutionary "just so" story that might explain why the FAE evolved.

The more interesting question, though, is how does the FAE result in the free will "illusion" (if it does, and if free will *is* an illusion)? My suggestion is that the free illusion is a result of the FAE as applied to one's own character. The FAE says that people overestimate environment and underemphasize character when making attributions (or, crudely, we emphasize nurture at the expense of nature). A belief in free will, even in being-causa-sui, might evolve if we irrationally emphasize the role of character... in the formation of character itself. Hence our natural and mistaken intution that our characters builds our character, that we cause ourselves, that we are causa-sui.

[Neil]

Kip,

I seem to recall that people asymmetrically apply FAEs to themselves. Their successes are explained in terms of their character, but their failures are explained in terms of the environment. All in all, they don't overattribute environmental causes.

Second, the question for evolutionary explanations is not "did our ancestors encounter PDIs", but what strategies were available in the EEA, and what were the payoffs? If the payoff from cooperating was high, even though iterated games were rare, and the costs of being a sucker low, then cooperation might evolve. Are those conditions met? Lots of evidence that they are - (vampire bats, cleaner fish... the examples are familiar).

[Mark Smeltzer]

Kip,

"Hence our natural and mistaken intution that our characters builds our character, that we cause ourselves, that we are causa-sui."

I'll agree with you there.

[Mark Smeltzer]

Damage control: In other words, I agree that the notion of causa-sui is mistaken. That's about as far as we go ;)

[Tamler Sommers]

Thanks for the comments, Daniel. Here’s a quick response. As Kip says, you’re employing a much broader interpretation of the concept ‘rationality’ than I am. I give a technical definition that only serves for the purposes of my argument. Most relevant to your first objection is the second sense—b-rationality, I call it. To believe that something (a behavior or feeling) is b-rational is to believe that it serves your immediate self-interests. So while it may be in your long term self-interest to cooperate in a prisoner’s dilemma type situation—and thus rational in that sense—it is still not rational in my sense (b-rational).

The idea is that at some point in our cognitive development, we developed the capacity to assess our short-term self interests, and act accordingly, but lacked the capacity to calculate our interests in the long term. This gave rise to certain commitment problems. Nature’s short-cut solution to this problem was to select for emotions that committed us to act against our immediate self-interests in order to further long term reproductive and material gains. (This is, of course, Robert Frank’s idea, not mine.)

Other brief responses: “Sommers presupposes that these emotions have cognitive content – that they purport to represent certain features of the world. However, Sommers doesn’t consider the possibility that the reactive emotions are fundamentally non-cognitive – that the question of rationality or irrationality doesn’t apply to them at all.”

I don’t know if I “presuppose” it, although you’re right that I don’t provide a detailed argument for the claim. Remember though that for non CS creatures, the emotions do not have cognitive content. It’s only as we become more cognitively sophisticated that certain beliefs about RMR are necessary to fortify the link between emotion and action.

“There appears to be an equivocation between the question of the rationality of the reactive attitudes, and the rationality of the beliefs that putatively justify them.”

Not an equivocation, just two different senses of rationality, both explicitly defined.

[Tamler Sommers]

Neil,

I don't know if you've read the paper, but Daniel's summary of the argument wasn't complete. First of all, I don't talk all that much about prisoners' dilemmas. I focus more on commitment problems. Second, it's the belief in robust moral responsibility, not free will, that's necessary to get you to act on an emotion. In a commitment problem that requires you to act retributively (or be predisposed to do so), you won't act this way unless you believe that the offender is DESERVING of punishment. If you didn't believe this, why would you risk your own safety to punish them. The belief in free will comes in only now. You don't believe someone is DESERVING of blame or punishment unless you think they acted with (libertarian) free will. (All of these are of course defeasible claims.)

On your last point, I agree. In fact, I have a paper that runs a parallel argument in support of a Mackie-style error theory of objective moral values.

[Kip Werking]

Neil, I just arrived in Sacramento for vacation. But I want to say: I think your reminder about the FAE's asymmetry is probably fatal to my theory. I strongly suspect that people *do* overemphasize the role of character in the formation of character itself, but this phenomena may be distinct from the FAE.

I also strongly suspect that this effect you describe, people's overemphasizing the environment's role when making attributions about the self, is sensitive to whether the person wants to own that action. For example, wouldn't a person be more likely to emphasize the role of character in explaining why he saved the world? As opposed to why he went on a crime spree?

(By the way, I think the statement "All in all, they don't overattribute environmental causes." is misleading, to the extent it implies that two irrationalities can cancel each other out.)

"Second, the question for evolutionary explanations is not "did our ancestors encounter PDIs", but what strategies were available in the EEA, and what were the payoffs? If the payoff from cooperating was high, even though iterated games were rare, and the costs of being a sucker low, then cooperation might evolve. Are those conditions met? Lots of evidence that they are - (vampire bats, cleaner fish... the examples are familiar)."

I agree that cooperation can definitely evolve in this way, and I think it's a fascinating phenomonon. But I want to keep separate, as I think Tamler does, the question of whether a belief is adaptive, and the question of whether a belief is accurate. The quote above emphasizes the former question, but that alone shouldn't be dispositive (if I'm misunderstanding what you wrote, I'm sorry). Indeed, if I under Tamler correctly then, on his view, the illusion of free will evolved to prevent us from recognizing that an adaptive belief is irrational. (I'm not sure how a genuinely irrational belief could be adaptive.)

[Kip Werking]

The excellent blog "Mind Hacks" has two new posts about free will:

NewSci on autism, free will and homo florensis
http://www.mindhacks.com/blog/2005/06/newsci_on_autism_fr.html

Discusses the latest issue of New Scientist, which has a cover story about free will.

The cognitive basis of good and evil
http://www.mindhacks.com/blog/2005/06/the_cognitive_basis_.html

Discusses the exact same phenomena which we explore in the Garden thread: the Fundamental Attribution Error. The author here argues the FAE helps explain our notions of good and evil.

[Daniel Cohen]

Tamler,

I admit I was a bit gung-ho in my reconstruction of your argument. I can see now that your stipulations about a and b rationality are central, but I’m no longer confident I’m getting your argument. (By the way, I found it a little confusing, when reading your paper, that “a-rational” didn’t mean arational (as in ‘not rational’). Perhaps it would be better to talk about x-rationality and y-rationality instead.)

Let me say a little more about why I chose to use the concept of rationality unanalysed in my earlier reconstruction. I figured that the argument went essentially as follows (yes, I’m afraid it’s going to be another gung-ho reconstruction):

(a) It’s irrational to cooperate (b) therefore the reactive attitudes (which make us cooperate) are irrational (c) therefore the belief in free will (which underwrites the reactive attitudes) is irrational.

In this argument, it is crucial that ‘rational’ holds its meaning in (a) (b) and (c), on pain of equivocation. My response, essentially, was that any questions regarding the rationality (a) of cooperation (b) of the reactive attitudes, and (c) of whatever beliefs we may have evolved to justify the reactive attitudes, are in fact separate questions, perhaps involving distinct notions of rationality. Given this, the argument, as I just put it, would seem to rest on at least 2 equivocations.

I’m sure my ‘reconstruction’ doesn’t really capture your argument at all, however. Perhaps you can take this all as an invitation to provide a more accurate reconstruction.

[Tamler Sommers]

Daniel,

There's a fundamental confusion that needs to cleared up right off the bat. My paper does not purport to be a direct argument against free will and RMR. It is an attempt to explain why we might experience or believe ourselves to have free will and RMR even if in fact we do not. To do this, I ASSUME that theoretical arguments against free will and RMR (of the kind presented by G. Strawson and Pereboom etc.) are sound. I then ask: if they are sound, then why is the belief to the contrary so widespread? Why would everyone have this false set of beliefs? That's the question I set out to answer in this paper. But I don't argue directly that the belief IS false.

(A relevant aside: I gave this paper at the recent SPP conference at Wake Forest. Bob Kane saw it and told me afterwards that he agreed with everything except the 2nd claim of the error theory, a claim I simply assumed was true. How is that possible, I wondered. That's when I realized that my theory is fully compatible with a naturalistic libertarian view like Kane's. But whereas I think my theory explains, in part anyway, why we MISTAKENLY believe in RMR, Kane thinks it helps explain why we ACCURATELY believe in RMR, and why the belief in RMR is so valuable.)

That said, I do think my paper contains an indirect argument against free will and RMR, since a plausible explanation for why we have a false belief is further evidence that the belief is IN FACT false (assuming of course that the explanation does not require that the belief be true). But the explanation is not a complete argument on its own. It needs to be combined with independent arguments against free will and RMR.

Now you can see that your recent formulation is not an accurate representation of the argument (since your conclusion is that a belief in free will is irrational, which is not my conclusion.)

Here is the reconstruction I give in the paper itself (I take your point about 'a-rational' being a confusing term, but I'll leave it that way for now):

1. Retaining the link between the reactive attitudes, or commitment devices, and the accompanying behavior was important for biological fitness in hominids.
2. Greater cognitive sophistication, which evolved for other reasons, undermined this link. The ability to assess the a-rationality and b-rationality of our attitudes made it less probable that the attitude would motivate behavior the adaptive behavior in question.
3. The belief in and experience of robust moral responsibility would offset this undermining effect, and would therefore have been adaptive.
4. However, CS creatures do not consider it a-rational to attribute robust moral responsibility to agents who are not radically free.
5. Therefore, unless we believed ourselves and others to be radically free agents, we would not consider it rational to attribute RMR.
6. Experiencing ourselves as radically free agents allows us to believe that we and others are radically free.
7. Therefore, the experience of radical free agency, the phenomenology of free choice, would have been adaptive for CS creatures.

As you can see, at no point do I claim that a belief in free will is irrational in your sense of the term. (Although, of course, I believe that it is.)

For the record, I think 1-3 are more solid than 4-7. I'm intrigued by 4-7, but I think there are a lot of good competing explanations for why we believe ourselves to have radical freedom. Spinoza's, for example. And many other more recent theories.)

Sorry for the long reply, but does that help?

[Metta Spencer]

The insight that comes from playing aany PD game (iterated or not) is that there are two kinds of rationality, which are not necessarily congruent. At the individual level, it is rational for me to defect. As a member of a collectivity, it is even better for us both to cooperate. It is only because of the incompatibility between the two forms of rationality that we may succumb to the temptation to defect, even knowing what we know.

http://metta-spencer.blogspot.com