Mo(o)re on TNR (Transfer of Non-Responsibility) Principles

Here’s an argument I find compelling:

1. If TNR is unjustified, then incompatibilism is unjustified.
2. If TNR is justified, then no one is morally responsible for anything.
3. So, either incompatibilism is unjustified or no one is morally responsible for anything.

As I suggested in response to Tamler’s post about Transfer of Non-responsibility (TNR) principles, I think the conclusion forces a Moorean dilemma of the sort that was leading Tamler’s Jack, even after reflection, to impale himself on the first horn of the dilemma, despite finding TNR and incompatibilism (and my premise 2) plausible.  I’ll offer some of my thoughts on these issues (sorry for the length) and then I’d love to hear what others think.

 

I find premise 1 compelling because I think TNR principles serve as the crux of every incompatibilist argument (including hard incompatibilist/skeptic arguments), often explicitly, sometimes implicitly.  Peter van Inwagen’s Principle Beta (and the modified versions of it like “Beta-box”) is a TNR principle, and he says the Consequence Argument (and incompatibilism) relies on its validity.  TNR appears in Robert Kane’s UR principle: “if an agent is personally responsible for X and if Y is an arche (sufficient condition, cause or motive) for X, then the agent must also be personally responsible for Y.”  Derk Pereboom’s Principle O is a TNR principle:  “If an agent is morally responsible for her deciding to perform an action, then the production of this decision must be something over which the agent has control,” and he seems to suggest that the best explanation for our intuitions in his four-case argument are based on TNR.  And Galen Strawson’s Basic Argument relies on a TNR principle in every one of its versions.  In one long version, it is premise 3:  “If, therefore [sic], one is to be truly responsible for how one acts, one must be truly responsible for how one is, mentally speaking—in certain respects” (1986, 28); I use “sic” because I do not see that this premise follows from the prior two premises rather than just being asserted as the crux of the rest of the argument.  I simply don’t see how an incompatibilist argument will succeed without a TNR principle (I’d like to add “any more than I can see how an argument for skepticism about knowledge can succeed without a closure principle” but I don’t know that literature well enough to be confident about that—Joe C., isn’t that your view?).

 

Before I move onto premise 2, you may be wondering what I (and Tamler) mean by TNR, since no one has explained what it means in these posts.  The basic idea, as captured in the above examples, is that, in order to be morally responsible for an action (or decision), one must be morally responsible for some part of what brought about that action (or decision) (I’ll stick with talk of moral responsibility, though I think we could just as well plug in one sort of free will people in the debate are talking about).   This basic idea sounds eminently plausible, since in real life the people we consider morally responsible are highly involved in the conditions that bring about their actions and decisions (they made lots of decisions that formed their character and got them into the situations they are in), and this prior involvement in their actions is clearly one of the reasons we hold them responsible.  But, of course, the basic idea then gets you on an infinite regress going back to a time when the (infant) agent is not involved in the right sort of way in the conditions that bring about her actions and decisions.  The basic idea makes less and less sense the farther back in an agent’s life you go. 

 

(In a separate post I might argue that the way to get out of the silly infinite regress is to realize that moral responsibility (and desert and free will) is a graded concept and getting from none to just a little bit and then more and more doesn’t require some magical injection of agent-causal powers but the slow development of relevant psychological capacities.  The infinite regress dissolves into a Sorites paradox, and we all know that there are, in fact, heaps.)

 

But let me try to be more precise about TNR.  Here is a version I find at least as plausible as the ones found in the literature (it is similar to Kane’s U):

 

TNR:  If X is a set of conditions that (ultimately) brings about Y, and an agent A is not morally responsible in any way for any part of X, then A is not morally responsible in any way for Y.

 

(Or perhaps we could try a version replacing “brings about Y” with “in principle could explain Y as much as it is possible for Y to be explained.”)

 

Notice that this TNR principle is neutral between deterministic and indeterministic universes or causal explanations.  After all, why should it matter if the prior conditions X that precede any possible contribution from the agent are logically sufficient for Y or simply as complete a cause/explanation as is possible for conditions that ultimately bring about Y. 

 

Consider the informal version of van Inwagen’s Consequence argument, which basically expresses the idea behind TNR:  “If determinism is true, then our acts are the consequences of the laws of nature and events in the remote past. But it is not up to us what went on before we were born, and neither is it up to us what the laws of nature are. Therefore, the consequences of these things (including our present acts) are not up to us.”

 

Has anyone noticed that the opening antecedent does no work as written?  With the antecedent, we need to add “then our acts are logically necessary consequences of…”  But we could just start the sentence without the antecedent like this:  “Our acts are ultimately the consequences of…”  Is anything really lost by taking out determinism here and replacing the logical necessity with the idea of ultimately bringing about?  This allows us to see the similarity between the Consequence argument and Strawson’s Basic argument; to see what many on all sides of the debate have accepted—that the issue is really about sourcehood rather than alternative possibilities; to see why van Inwagen has moved towards combining the Consequence argument and the Mind argument into a general argument for mysterianism (skepticism by another name?); and to see that incompatibilism is really—ultimately!—about TNR (and not determinism specifically).

 

Now, why would a compatibilist like me accept premise 2 above?  First, because “moral responsibility” is built right into the TNR principle (it doesn’t work through some tenuous claim about ultimate sourcehood being required for MR), and second, because it seems obvious that for any agent, there must be conditions for which the agent is not MR that bring about or explain (as much as is possible) what that agent does (except perhaps, as Joe C. points out, for weird agents that spring into existence fully formed or infinite agents, such as God, though the “as much as is possible” clause might come into play here, and in any case, shouldn’t we really be interested in whether any humans can be MR for anything?).

 

So, this has all been set up for what I really want to ask people about (of course, feel free to crush me on any of the above):

1.      Are there any arguments for TNR principles?  If so, what are they?  If not, does its support come from intuition and lack of clear counterexamples (as van Inwagen suggested about Beta, though of course, there were counterexamples to the original Beta and there is always the non-question-begging stipulated counterexample of the adult human who is (clearly!) responsible for doing A even though she is clearly not responsible for the set of conditions in the distant past and laws that ultimately bring about A, either deterministically or indeterministically)? 

2.      If TNR is just supposed to be a plausible idea supported by our pre- and/or post-theoretical thinking, why should we think it trumps our very strong, perhaps un-give-up-able (cf. Peter Strawson and van Inwagen) pre- and post-theoretical intuitions that people are morally responsible for much of what they do and are apt targets for reactive attitudes and genuine desert?

Comments

[Neil]

Great post, Eddy! I recently began and then set aside (for lack of time) a paper with the provisional title "Free will without physics". The basic idea is that all the major issues in the debate arise independently of the worry about the causal structure of the universe. I wasn't thinking in terms of TNR at all, but I now suspect that TNR was doing some work in the idea. I have been accused of being a closet incompatibilist. I think it is more plausible (and worrying) that my views rest on TNR without resting on incompatibilism.

My only thought at this stage is this: you're asking what grounds we have for plumping for TNR over the idea that agents are morally responsible. But I'm not sure that we can choose MR rather than TNR, because I suspect that some kind of transfer principle is built into the notion of MR. Think about Derk's Principle O again: yes, you can detect a whiff of transfer about it, but that degree of transfer seems built into the notion of MR. How can agent be responsible for a decision over the production of which that agent had no control (granted Scanlon might bite this bullet, but I think he does so by changing the subject).

[Justin Coates]

Hi Eddy,

What about the following argument for incompatibilism (similar to van Inwagen's first version of CA, 1983*, and Fischer's Basic Argument, 1994)?

(1) Suppose determinism obtains.(2) Jones X's.(3) If Jones were to refrain from X-ing the past would have to be different or the laws would have to be different(4) Jones cannot do anything such that the past is different (i.e. it's fixed)(5) Jones cannot do anything such that the laws are different (they're fixed)(6) So, Jones cannot refrain from X-ing(7) One is responsible for X-ing only if one could refrain from X-ing (PAP)(8) So, Jones is not responsible for X-ing.

This is rough, but I think it can be refined into a pretty powerful (although unsound) argument for incompatibilism that doesn't obviously rely on TNR.

*If I'm remembering correctly, van Inwagen thinks each of the three versions of CA that he gives relies on Beta, but it's just not obvious to me how the version that runs pretty much identically to the one I've offered relies on any TNR (Beta or otherwise).

[Justin A. Capes]

I suspect something like Beta-box slips into Justin's argument around premise (3), though I'm not sure about this. I too have sometimes wondered where, exactly, Beta is supposed to come in to play in PvI's 1st formulation of the CA, which, for what it's worth, I think is the most clear and compelling version of the argument. In any event, I agree that the argument Justin presents is unsound--premise 7 is false, for familiar reasons. However, I'm inclined to endorse an alternative possibilities principle along the lines of PvI's PPP2. If such a principle is defensible, then the argument Justin gives (or something like it) would be sound, in my view. If it is sound, then it's not clear to me where TNR slips into the picture.

In any event, it's at least as obvious to me that some people at least sometimes have at least some control over the causes of their actions (and thus over their actions) as it is that people are sometimes morally responsible. Hence, I feel no pressure to abandon TNR or incompatibilism (if it rests on TNR), and I would be content to rest the case for TNR on *mere* intuition, though of course there might be further arguments for TNR that people smarter than I can find.

[Neil]

One addition: is TNR necessarily at issue in any historical view (like Fischer's)?

[Kip]

I just want to add that I agree with Eddy 100% that the structure of the universe is irrelevant to the compatibility question. The compatibility question is really nothing more than semantics: whose definition of free will is right?

I also agree with Eddy 100% that TNR (interpreted very broadly) is the heart of the matter.

Why do I lean towards denying free will instead of TNR?

In a nutshell: I don't know of any error theory for TNR, but I have a strong error theory for free will.

Why might people believe in free will even though it doesn't exist?

1. Wishful thinking---not to be underestimated
2. The fundamental attribution error
3. The illusion of control
4. The illusion of spontaneous evil (the inability to understand the motives of harmers)
5. The empathy gap
6. Reactance: we want worthless choices (and, because of wishful thinking, we believe in what we want)
7. The just world phenomenon: people will blame the victim when bad things happen, even if Hitler is one of the victims
8. Religious baggage: most language users believe in an almight God, who ensures that the world is just (fitting with the Just World Phenomenon) and who gave us a soul and, perhaps, libertarian and self-creating powers. That way we can't blame our genes, parents, or anything else, we can just blame ourselves.

That's just off the top of my head. All of these reasons put 'free will' at a disadvantage. But I can't think of a single reason why someone would have mistakenly believed in TNR. What motivation would they have had? What cognitive bias, as demonstrated in the social sciences, does it invoke? None, as far as I can tell.

TNR is a neutral premise that leads to a conclusion in which we have a highly charged, emotional investment. If I have to doubt the premise, or the conclusion, I'm going to lean towards the conclusion.

[Cihan]

Eddy,

I don't know about (1). I don't know any arguments for TNR - but it seems to me to be so intuitively true as to require no argumentation.

But for (2), isn't TNR in some sense a more *basic* intuition then our "un-giv-up-able" intuition that "people are morally responsible for much of what they do and are apt targets for reactive attitudes and genuine desert".

After all, do we not suppress morally reactive attitudes when we discover some exculpating/excusing factor about the agent? ('She was really a puppet for the neuroscientist!' 'JoJo was brainwashed as a child!'.) And I think these suppressions of morally reactive attitudes are dictated by even more basic principles such as TNR. ('Jojo wasn't responsible for his parent'.)

I think this is a point Nagel does in "Fredom and Autonomy" when he criticizes Strawson. Basically, what I'm saying is that TNR is already embedded in what you call un-give-up-able attitudes. So doesn't TNR have precedence?

[Cihan]

Eddy, you may be right - these attitudes may really be unsheddable.

Basically, after writing a post defending, I felt oddly guilty about procrastinating in my work.

Nonetheless, just because I feel guilty doesn't mean I'm morally responsible.

[Tamler Sommers]

"After all, do we not suppress morally reactive attitudes when we discover some exculpating/excusing factor about the agent? ('She was really a puppet for the neuroscientist!' 'JoJo was brainwashed as a child!'.)"

Yes, but for TNR to be "more basic," the suppression would have to go exclusively in that direction. But it doesn't. We also suppress TNR in cases where blame just seems really appropriate. Take the attitude almost all liberal academics take towards Dick Cheney. Do they care that he's not ultimately responsible for his character? Do people care that Hitler wasn't ultimately responsible for character? Surely some people are (a) aware that Hilter is not causa sui, but (b) think he's blameworthy anyway. Or take Jack the skeptic's attitude towards the person who deliberately harmed his daughter.

This is why I think the 'TNR is more basic strategy' won't work. And I certainly don't think one (e.g. Neil) can argue that TNR is built into the moral responsibility concept. People use the concept all the time without committing themselves to even the most basic control conditions never mind something as hard or impossible to satisfy as a ultimacy condition.

[Tamler Sommers]

One thing to add: Kip's strategy in fact strikes me as the most promising. But it requires spelling out what counts as a debunking explanation and what doesn't. And remember, it's moral responsibility that's at issue here, not free will.

[Eddy Nahmias]

Lots of interesting responses. Thanks!
Briefly, I'd say:
1. Justin, you are right that there may be ways to frame incompatibilist arguments that make TNR fade into the background or even avoid implicit use of it. I think those arguments are the weakest. For instance, in your version I think premises 4 and 7 are very questionable. I think to make those premises stronger, one would have to appeal to a TNR principle (that's likely why PvI thinks all his versions of CA rely on Beta). As I said, there seems to be significant movement on all sides of the debate away from arguments that rely on PAPistry and towards arguments that appeal to sourcehood, and I think sourcehood arguments really need TNR.

2. I don't think our concept of moral responsibility has TNR built into it. It may appear to when TNR is phrased more like Pereboom's O (that we can't be MR if we had no *control* over what caused our action), but, even though I brought it up, I don't think O gets you the sort of TNR that leads to an infinite regress.

Notice the middle of my post where I explain why TNR seems so plausible (because in just about every adults' action or decision for which we want to hold them MR, the agent is partially MR--in compatibilist/control sense--for part of the conditions that brought about the action or decision). TNR is implausible, I think, only when it is used in an infinite regress to get us to look for a *first* MR action.

3. Hence, Kip, I think we could develop an error-theory for our intuitions about TNR. It'd look something like the common philosophical move of arguing that too much philosophizing is what leads you to find certain principles to be universally valid, when in fact there is an explanation for why the claim looks universally valid and also for why it fails in certain instances. But working out this move would take a lot more (consider how hard it is to come up with counterexamples to Beta and PAP... but it has been done!).

[villa te huur]

I happy to see this brilliant brief related to transfer of non-responsibility.Thanks

[Neil]

Eddy, as my second post attempted (somewhat elliptically) to suggest, transfer principles need have nothing to do with infinite regresses. For instance, many compatibilists (eg Fischer, Haji, Mele) believe that, with the possible exception of instant agents, in order to be MR for A-ing, one must be MR for the springs of A-ing. That's apparently a transfer principle, and its not obvious that something like this isn't built into MR.

[Paul Torek]

Eddy,

You are right about everything of course (what did you expect to hear from your fan club). But in answer to (1), yes, there are abductive arguments for TNR. The Four Case argument is an example.

Maybe we should speak of transfer principles rather than TNR principles? For example, suppose one supplemented Pereboom's O with a Transfer of Non-Control principle. I suppose you could then call the combo a TNR package, in a broad sense.

[Mark Smeltzer]

Neil,

Can you offer a quote from Fischer to back up that claim? I know of plenty where he discusses transfer principles for responsibility, and I also know plenty where he attacks transfer principles of non-responsibility. I'm unaware of anything I've read of Fischer's where he endorses the need for being MR for the "springs" of MR (this sounds like TNR in reverse to me).

[Neil]

Mark:

"we contend that the agent must have taken responsibility in order to be morally responsible for his behavior. In taking responsibility for the springs of one's behavior, one makes the mechanism that issues in it 'one's own'."

Responsibility and Control, p. 230.

[Neal Tognazzini]

Neil,

I don't think that taking responsibility for a mechanism results in one's being morally responsible for having that mechanism. Think of the conditions that are necessary and sufficient for taking responsibility for a mechanism, and note that exercising guidance control over which mechanism one acts on isn't among them (and it's not even clear what that would mean). Given that guidance control is necessary for moral responsibility, I would think this means that Fischer is not committed to the view that morally responsible agents are also morally responsible for having the mechanisms that issue in their actions. (Perhaps this makes the phrase 'taking responsibility for' a bit unfortunate in this context, since you can take responsibility for something without thereby becoming responsible for it.)

[Al Mele]

Neil wrote: "many compatibilists (eg Fischer, Haji, Mele) believe that, with the possible exception of instant agents, in order to be MR for A-ing, one must be MR for the springs of A-ing." Actually, I explicitly deny this and argue against it. See, for example, my *Autonomous Agents*, pp. 223-230. Neil may be confusing my claim that in order to be morally responsible for actions agents must lack histories of certain kinds with the claim that "(with the possible exception of instant agents) in order to be MR for A-ing, one must be MR for the springs of A-ing."

[Neil]

I am aware of the claim, Al. But so far as I can see, instant agents seem on your view to be co-extensive with the class of agents who lack histories of the relevant kind but do not satisfy the positive conditions on MR. If that is right, then the quotation from me you cite is correct.

[Neil]

Neal, I can't see where we are disagreeing. Fischer says that to be MR for an action, one must satisfy an ownership condition wrt to the mechanism upon which one acts. It is, I agree, a further question whether satisfying the ownership condition makes one MR for having that mechanism (though I suspect Fischer holds that too).

[Mark Smeltzer]

Neil,

"It is, I agree, a further question whether satisfying the ownership condition makes one MR for having that mechanism (though I suspect Fischer holds that too)."

On Fischer's view, the key ingredients for MR are (1) a moderately RR mechanism and (2) taking responsibility for that mechanism. Given that, how could we make sense out of the claim that "satisfying the ownership condition makes one MR for having that mechanism"? This seems like a unsupported leap.

Taking responsibility for the mechanism is Fischer's answer to criticisms regarding agent-mechanism identification.

I think Fischer's rationale here has its roots in the general concept of ownership: if something is unowned, then the mere assertion of ownership is a sufficient criterion for ownership.

[Al Mele]

Neil wrote: "I am aware of the claim, Al. But so far as I can see, instant agents seem on your view to be co-extensive with the class of agents who lack histories of the relevant kind but do not satisfy the positive conditions on MR. If that is right, then the quotation from me you cite is correct." In the chapter of *Autonomous Agents* in which I argue against the claim attributed to me -- namely, that "with the possible exception of instant agents, in order to be MR for A-ing, one must be MR for the springs of A-ing" -- I try to explain how children who have not done anything freely yet may act freely now (pp. 227-230). In *Free Will and Luck* I return to this issue, this time emphasizing moral responsibility too. I try to explain how children who have not been morally responsible for anything yet may be morally responsible for something they do now (pp. 129-133). Given that this is the first thing for which they're morally responsible, they're not responsible for its springs. About my "positive conditions on MR": If someone were to confuse my proposal of sufficient conditions for moral responsibility with a proposal of necessary conditions (or individually necessary and jointly sufficient conditions) for MR, I could see why he or she would say what Neil said. But, of course, Neil wouldn't confuse an explicit statement of sufficient conditions with a statement of necessary conditions. So I find the remark puzzling.

[Neal Tognazzini]

Neil,

The answer to the further question -- whether taking responsibility for the springs of one's action means that one is morally responsible for them -- will tell us whether Fischer's view has to appeal to some transfer principle, which is what I was taking issue with. I think the view doesn't require any transfer principle precisely because Fischer wouldn't go on to make the claim that moral responsibility for actions requires moral responsibility for their springs. It just requires ownership of the springs.

I suppose there is SOME sort of transfer going on here -- that if you don't own the springs, and the springs lead to action, then you aren't morally responsible for the action -- but it's a far cry from TNR. And surely we wouldn't want to water down TNR so far that any historical view of moral responsibility requires TNR.

[Cihan]

Also, note that Eddy (and almost all others) assumes that there is some fact of the matter about which intuition should trump the other.

I don't think so. I am no longer so sure that there is some fact of the matter when it comes to free will. I think it's more of a linguistic+intuition dispute and that there is no way to resolve it. I'm not so sure that most of the issues can be meaningfully/substantially stated, let alone resolved.

For me, I think the more meaningful project is the normative question "when should we employ reactive attitudes if at all?".

And I never thought I'd be interested in ethics...

[Paul Torek]

Cihan,

Welcome to the view that the moral horse should go before the semantic cart. Now let's see if the horse is strong enough to pull the cart: i.e., let's use our answer to the normative questions to settle the linguistic ones.

[Neil]

Al, I am using 'instant agent' in a slightly non-standard, but I think defensible way: an instant agent is an agent who (like Diana) lacks a certain sort of normative history. Of such agents, if they are MR, it is not in virtue of lacking a normative history of a certain sort that they are MR.