Of Minds and Language (58 page)

C
HOMSKY
: Yes. The other question is whether you are proving something that I have always believed, namely that teenagers are a different species. [laughter]

H
AUSER
: Like right, man.

P
IATTELLI
-P
ALMARINI
: Marc, regarding the domain of decision-making, as you know Thomas Gilovitch and Daniel Kahneman and others have shown that, in the short term at least, there is more regret for something you did than for something you didn't do, even though the consequences are exactly the same.
¹⁰
This is a traditional thing, and you seem to have it here, you know, omission vs. action. But on the other hand, Connolly and Zeelemberg
¹¹
and others have shown that a crucial factor is whether you are somebody who is supposed to be doing something – for example a doctor is called to do something, and he either
does nothing and the patient dies, or he does something and the patient also dies. That is, the same result happens when the doctor fails to act, as when he does something but makes the wrong decision. The doctor who was called to do something and failed to act is considered worse morally. So I wonder whether you plan to extend it. It would be interesting because it has to do with counterfactuals, and it also has to do with a number of other tests that have been made, where there is something anomalous in the series of actions and you pinpoint the anomalous thing as being
the
cause of what happens.

H
AUSER
: Yes, we are. Take the classic case that James Rachels brought forward,
¹²
which some of you may well know, of the greedy uncle who wants to do away with his nephew who is first in line for the family inheritance. In Story 1, he's babysitting the nephew and he goes upstairs while the nephew is taking a bath, and he intends to kill the kid and he drowns him. So he intends to kill and he kills. In Story 2 he has the same exact intentions, he goes upstairs, the kid has flipped over in the bathtub and is drowning, and he just walks away and lets him drown. Now the first is an action, the second is an omission, but we don't want to see those cases as different, in fact we see them both as the same. So in some cases we don't see a difference between action and omission, and in some cases we do, and the question is, how much information do you attribute to the agent, that then makes you either lose it or pick it up? I think it is not clear in the philosophical literature at all. And it is also not clear to what extent we are vulnerable to the action–omission distinction. So that is what we are trying to do in two ways. One is to play around with when you get the information, whether you get the consequences first or the intentionality first, and I will come back to that in a second, and the second question is, to what extent is this distinction available early in life? We know almost nothing about action–omission in its ontogeny in young children; we know almost nothing about the means–foreseen distinction either. Oddly, even though there has been a rich literature on theory of mind, these distinctions don't enter into the discourse because they have largely been developed within moral psychology and the law. But they bear directly on the agent's mental states.

One more case. This is a nice case by a philosopher named Joshua Knobe.
¹³
There is a CEO of a company, and the President of the company comes to the CEO and says “Look, I've got a policy which, if implemented, will make us millions of dollars.” Now there are two versions of the story. In Story 1, the President says “If we implement the policy, there is a good chance it'll
make millions of dollars, but it will probably harm the environment.” In Story 2, the policy will probably help the environment. In the first case, the CEO says, “Look, I don't care about harming the environment, I just care about making money. Implement the policy.” The company implements the policy, and they make millions of dollars, and it harms the environment. Now you ask people: did the CEO intend to harm the environment? Here, subjects say yes. In the help case, however, the CEO says the same thing, “I care about making money, I don't care about the environment. Implement the policy.” Now you ask people: did the CEO intend to help the environment? Here, subjects say no. The idea here is that labeling an individual with some kind of moral attribute like
blame
, or
blameworthy
, actually affects the intentional attribution, at least at some level. So again, there can be all sorts of ways in which these patterns unfold, depending upon the temporal flow of them through time.

J
ANET
D
EAN
F
ODOR
: Making these moral distinctions is actually very distressing. I have to kill at least one guy if not five. And so the natural thing is to reach for a rationale, an excuse of some kind, and a very common excuse it seems to me is to blame the victim. I think we can all remember cases where we have tried to blame the victim. So this is a factor that could be introduced into the experiment so that the five guys have been told it is stupid to walk on the tracks, it is dangerous to them and everybody else; the one fat guy has been told that this is a track that is never used, it is perfectly safe to walk here, or vice-versa. Is this a universal that would make a difference to the study?

H
AUSER
: YES, the problem here is that the space at some level is unconstrained in terms of the number and kind of permutations you could run. You could ask about in-group vs. out-group, you could change the numbers – there are all sorts of things you could change, and several papers have explored this part of the problem. We have taken a different route, which is to hold these personality traits constant in order to explore the causal–intentional structure of event perception in the moral domain. As soon as you put things like responsibility (like they're workmen, they should be there; or it is the conductor's job, he's got a responsibility towards the others), you are going to change lots of the dynamics. And I find these to be very difficult problems, headed more toward social psychology, and a zone that I am less familiar with. It is not that these are the wrong kinds of questions, but rather, that I have less confidence with regard to the experimental questions. I feel more comfortable with the primitives underlying causal–intentional attributions because I am quite convinced that we can explore these issues in infants, animals, patient populations, and so forth. That is, in the same way that Lila Gleitman has been exploring the foundations of
giving and hugging in infants (see pages 207–211), looking at how infants dig beneath the surface dynamics, we have moved in a similar direction.

What is also of interest is how confident people can be about their judgments for one or two cases, but as they pile up, they lose this confidence, and the reason we think this is happening is because they don't really have access to the underlying principles, just the surface features of each case. Let me illustrate with my father, a very smart, rational physicist. He had asked me what I was working on, and so I decided to illustrate by giving him some moral dilemmas. I started with the bystander case and he answered, “Well of course you flip the switch.” I then gave him the fat man trolley case and he said, “Well of course you push the man.” This is, you will recall, a relatively rare response so I asked him why. He replied, “Well, because it is always better to save five than one.” I then give him the classic organ-donor case, where there is a surgeon in the hospital and a nurse comes in and says, “We've got five people in critical care, each needs an organ, we have no time to ship out for the organs, but you know what? This guy just walked in to visit his friend. We could take his organs and save five lives. Is that okay?” My father says, “That's ridiculous, you can't just kill a healthy person off the streets!” “But wait,” I say, “you just killed the fat man.” He says, “Okay, you can't kill the fat man.” I say, “What about the switch case where you killed the guy on the side track?” He says, “Okay, you can't do that either.” So, ultimately, the whole thing unravels, because you can only locally explain one dilemma but you can't explain the cluster, because you don't have access to the underlying principles. This is the core intuition driving our work.

F
ODOR
: I am not sure why you think the universals are likely to be about the agent and not about the patient.

H
AUSER
: Oh, I think the universals may come in at the patient level too. My guess is that it is a universal, and so there are studies by Lewis Petrinovich showing that if you put kinship groups as the patients, you are going to get evolutionary sociobiology to work.
¹⁴
“I will favor those who are more genetically related to me over those who are not, all else equal.” You can get species effects. “I will favor human over even an endangered species like the chimpanzee.” So these effects are certainly operative, and they may well be a part of what is universal, but I don't think this part is specific to morality. Ingroup–outgroup distinctions arise in all sorts of contexts, some moral and some not. More importantly, perhaps, we have tried to tackle a different set of problems by holding patient identity out of the scenario, operating under a kind
of Rawlsian veil of ignorance. By doing this, we hope to uncover the architecture of the underlying psychological cause of the agent's actual action.

P
ARTICIPANT
: I wanted to shift to the musical section. I'd be very curious, if you could create a variable to your experiment – I would suspect that part of the preference for silence has to do with just the foreign nature of technologically produced sounds that we have
learned
to appreciate, like the flute or recorder sound in a lullaby. It may be quite offensive to the ear of some other primate. Referring back to your earlier experiments with quantification, one of the appeals of music is the structures that come with repetitious quantities – a performing musician learns to play chord progressions, for example, without physically counting them. You don't go “One, two, three, one one, two three two”; you
hear
the changes. And I wondered if there would be an appeal among primates that had a quantifying capacity, when using sounds from nature, that could be organized structurally to repetitively use sounds that they're familiar with rather than some kind of human technology that is used. Just an idea to look at whether the appeal of the recognition of repetitive quantities is a big factor in what we like about music. It's why young people like Techno.

H
AUSER
: In some sense I agree, but we are at such an early stage of this work that it is hard to make sense of much of it. There have been a couple of papers recently, by Smith and Lewicki, claiming that lots of the physiological firing patterns that you see to sounds have a very primitive system or structure that really taps into natural sound, and speech may simply be parasitic on this mechanism.
¹⁵
This position or perspective sets up a study that I just finished in collaboration with Athena Vouloumanos
¹⁶
– this is a slight tangent with respect to your question but it gives you an idea of what would be basic in terms of the auditory system.

There has been fifty years of research on neonates' preference for speech. The common lore is that babies are, early on, tuned to speech, preferring to listen to speech than non-speech, and showing significant abilities for speech discrimination. The problem has been that none of the studies to date have contrasted speech with other biological sounds, focusing instead on reversed speech, sine wave speech, white noise, and a variety of non-biological sounds. We therefore set out to test thirty-two neonates, less than forty-eight hours old, with a non-nutritive sucking technique where suck rate gives information about interest or attention to the material. We contrasted non-native speech with rhesus monkey vocalizations, and thus, both sets of stimuli are novel at some
level. The result was clear as can be: no preferences at all. Neonates sucked as much for non-native speech as they did for rhesus monkey calls. By the age of 3 months, however, possibly earlier, a preference for speech is in place. These results suggest that there are general auditory biases that get tuned up quickly in development. These results also rule out all the arguments that have been put forward for
in utero
experience, because by the third trimester, the baby is certainly getting some acoustic input. But whatever that is, it is insufficient to create a preference for speech over rhesus monkey calls. So that is just a long-winded way of saying that some of the preferences in music may well derive from quite general auditory preferences. It is certainly possible that if we had played more biological sounds, perhaps structured in some musically relevant pattern, then we would have seen a different pattern of responses.

Let me add one relatively new piece of data, still preliminary in terms of our analyses. We have just completed a study in which we presented marmosets with five months of exposure, twelve hours a day, to consonant chords, and then, to samples of mozart. The idea here was to more closely approximate the kind of exposure that human infants receive during early development. When we subsequently tested our subjects for a preference for consonance over dissonance, either chords or pieces of Mozart, subjects showed no preference. Interestingly, however, infants but not adults, exposed to the same materials showed a mere exposure effect, preferring the specific sounds played over novel, but matched sounds. This provides one of the first pieces of evidence in a non-human primate for a critical period effect.

G
ELMAN
: Just a bit on music. I expected you to be talking about something like harmonic principles, some principles of music that the mind treats as privileged. Consonance–dissonance is one of those, but if you are using chords, it matters whether you are in Western music or not. So the issue becomes what is universal across different harmonies. I believe it is the case that the octave and the fifth appear almost invariably in every harmony, where it is consonant, and that the transitions are such that they'll be very different, but you will find they are the fifth. That is not trivial because the physics is such that the first overtone is the octave, the second is the fifth, the ear is sensitive to these, etc. So this might mean that the principles are highly abstract and are harmonic principles, just as the linguistic principles that we are looking for are very highly abstract. And maybe it is not a question of whether the sounds are consonant are not, because what is consonant for us is not necessarily consonant in another culture. It is how we fill in the transitions that varies enormously.