Over at the Experimental Philosophy blog, Joshua Knobe has a post about a series of experiments that he has run with Jesse Prinz on people's intuitions about consciousness, and he includes a link to a draft of the paper they're writing on the experiments. The experiments were motivated by two hypotheses. The first says that people have a concept of "phenomenal consciousness," even if they don't know it. As a result of this concept, people will be willing to attribute phenomenal states to certain kinds of agents (you know, like people or animals), but not to others (like group agents, like businesses or nations). As evidence for this hypothesis, they present data showing that people rate sentences that attribute non-phenomenal states to group agents (e.g., "Acme Corp. believes that its profit margin will soon increase.") as being much more natural that sentences that attribute phenomenal states (e.g., "Acme Corp. is now experiencing great joy.") to those same group agents.
Their second hypothesis is that people's intuitions about phenomenal consciousness include restrictions on "physical realizers." That just means that, while people aren't all that concerned about the physical states in which nonphenomenal ascriptions (like "beliefs" or "intentions"), but they do pay attention to physical realizers when making phenomenal-state ascriptions (feelings, experiences, etc.). As evidence for this hypothesis, they present a study in which people are willing to attribute phenomenal states to a magic chair, but not to a company. They use this vignette to describe the magic chair:
Once there was a powerful sorceress. She came upon an ordinary chair and cast a spell on it that endowed it with a mind. The chair was still just made of wood, but because of the magic spell, it could now think complex thoughts and form elaborate plans. It would make detailed requests to the people around it, and if they didn't do everything just as it wanted, it would start complaining. People used to call it the Enchanted Chair. (p. 10 of the draft)
From this data, Knobe and Prinz conclude:
The moral here is clear. From the standpoint of physical realizers, a wooden chair is extremely different from a human being. Yet people were perfectly willing to ascribe phenomenal states to the chair. It therefore appears that people do not simply refuse to ascribe phenomenal states to any agent that differs from human beings in its physical realizers. They must be making use of some more specific restriction that rules out group agents on independent grounds. In philosophical jargon, our respondents were committed to the multiple realizability of phenomenal states, but they were also willing to impose certain specific restrictions on realization. (p. 10)
In other words, something about group agents makes them unfit for phenomenal states. It's not clear, from these studies, what that is, aside from number.
Finally, Knobe and Prinz present a study demonstrating possible moral implications of our concepts of phenomenal consciousness. In this study, they again present the magic chair and corporation, but in this case, the magic chair and corporation don't do what their creators wanted them to do, so they're dismantled. Participants are then asked whether it was wrong to dismantle the chair/corporation. Participants were much more likely to say that it was wrong to dismantle the chair than to dismantle the corporation. Apparently when an agent possesses phenomenal consciousness, our "moral concern" for that agent increases.
So that's the work. You can read the paper for more of the philosophical background. In concluding this post, I want to say something about this work as science, and little bit more about the data. First, I've said several times that I think Dr. Knobe does the best work in experimental philosophy, and I think this study is evidence that with practice, his empirical work is getting even better. At first, as I read through the draft, I was a bit worried because it seemed that they were relying too heavily on linguistic evidence. Anyone who's done work on concepts (as Jesse Prinz has) knows that linguistic data isn't enough to make strong conclusions about people's conceptual representations. But the last experiment on the moral status of the magic chair show that the implications of the linguistic distinctions people make about phenomenal and non-phenomenal states go beyond the words they'll use to describe different types of agents. People are using these distinctions to make judgments, and that means that the distinctions are conceptual. So, I'm very impressed with the designs of these studies.
Still, I'm not entirely convinced by their data. I wonder whether their second hypothesis is really addressed by their studies. The use of magic in the magic chair experiments, in particular, is problematic I think. What would happen if we used magic on a corporation? Might people then be more willing to attribute phenomenal states to it? And how does magic fit in to people's concepts of phenomenal consciousness. I mean, ordinarily, we don't attribute any sort of mental/conscious states to chairs, while we readily attribute nonphenomenal mental states to corporations. How does the use of magic change that? What the hell are people thinking? Clearly, people are easily convinced that phenomenal consciousness-inducing agency is involved, and I see nothing in their data to imply that such convincing couldn't be made for group agents. It may not be physical realizers, then, but something else having to do with our agency concepts that is responsible for much of their data. And if it is "physical realizers," might the hypothesis be simplified to one concerning only number (i.e., single agent = phenomenally conscious; group of agents = not phenomenally conscious)?
The use of magic in the magic chair experiments, in particular, is problematic.
That struck me immediately, too -- the concept 'magic' seems to be a restriction-eliminator (saying that something is magical seems to imply that some sort of ordinary restriction doesn't apply to it), so calling something magical is usually a signal to ignore things we would otherwise consider important, or even essential. And I doubt we learn much from finding out that people don't restrict a concept to similar 'physical realizers' when we've already suggested that they can ignore any ordinary restrictions.
I'm not sure whether this has much relevance to the Knobe-Prinz study in particular, but I also wonder whether people are influenced by the fact that they are fairly often exposed to chairs with minds -- not IRL, of course, but in stories and movies -- so they know how stories like that typically go. This actually worries me a lot about cognitive science studies that use narrative scenarios as part of the set-up -- it does seem to be the case, at least at first glance, that people will often concede and admit things for the sake of the story that they wouldn't anywhere else, so I always wonder how much of the conclusion is about what the study's about, and how much is really about how people think when faced with stories. Are there any tricks cognitive scientists typically use to deal with this worry, or have there been studies done that suggest that it's not actually a significant worry?
Hey Brandon, I'm glad it wasn't just me re: magic.
As for the use of stories, there are a few tricks you can use. One is to use several different stories, so that the idiosyncracies of one story don't affect those of other stories. One of the problems with experimental philosophy studies, to date, is that they tend to rely pretty heavily on one story or type of story. Another trick is to use possibly less direct paradigms to confirm the results you get with studies. Priming studies are common, for example. Where stories are used most frequently is in psycholinguistic research, and there, they're often concerned primarily with how people use/interpret linguistic constructions, so it's less of a problem there as well (magic, in this case, would be relevant, because it would provide a clue into the use of phenomenal consciousness constructions).
In my own studies, I've used stories pretty frequently, but in general, I'm less interested in the concepts people employ when reading them. In fact, I write really strange stories specifically so that people won't bring in too much background knowledge, because what I'm looking at is how differences in the structure of the stories affect what they remember, or how they form new concepts.
Could this notion that groups don't have phenomenal states be one of the reasons Searles "Chinese Room" anti- Artificial Intelligence thought experiment is as powerful as it is? We are forced to ascribe to the group of people in the room the phenomenal states of the "brain" they are emulating and we don't like to do that. Just a thought.
Or perhaps the reason why we don't like to think of large groups as possessing consciousness is that we have effective mental models of the behavior of large groups, and it's nothing like our models of things with minds.
I always considered corporations and the like to be akin to amoebas - capable of stimulus-response patterns, but not thought. The relationships between individuals that made up the group were far less complex than the individuals themselves.
Caledonian, that's essentially the position that Knobe and Prinz are arguing for (that our mental models, or whatever type of conceptual representation you refer, are different for the two).
Great comments all. As a recovering psycholinguist, I'm most intrigued by some of the manipulations here (believe me, I'm no philosopher). Animacy seems to be a confound in the above study. Has anyone attempted to control animacy and look at ascriptions of phenomenal states to simple count and mass nouns? So say, comparing the ascription of phenomenal states after being exposed to magic for a rock and for sand? Is sand a bunch of moral agents we have to be concerned about or one?
Couldn't you just say that "Acme Corp" is a metonym? Would that change anything? (ie. "The White House was pleased today..." sounds fairly natural to me.
Michael, yeah, you're right, there's just something off with the single vs. group agents contrast. It could be about animacy. It could be about many things. They need to run more conditions with different stories.
Mark, you know, that's a great point. I hadn't thought of any examples when I was reading the paper. And the phenomanal state ascriptions can be pretty group agent-specific. For example, it would be strange to say that the New York Times was depressed, but I bet you could say that the New York Times was embarassed.