Intuitions about Consciousness

Jesse Prinz and I have been running some experimental studies on people’s intuitions about consciousness, and we would be very interested to hear any thoughts you all might have about the results.

Here is the basic question. Suppose that people are trying to figure out whether an entity has certain mental states. Will their intuitions only be sensitive to information about the functional roles of that entity’s states? Or will their intuitions also be sensitive to information about physical realizers? 

Our hypothesis was that there is something special about ascriptions of mental states that involve phenomenal consciousness, such that these ascriptions are sensitive to information about physical realizers in a way that most other ascriptions are not.  Thus, suppose there were an entity whose states had functional roles that were a lot like those of human mental states but whose states had physical realizers that were highly unusual. If things work out just right, our hypothesis predicts that people would be reluctant to ascribe to this entity any states that involved phenomenal consciousness but that they would be perfectly happy to ascribe various other sorts of mental states.

To test this hypothesis, we looked at ascriptions of mental states to corporations. It seems that the functional roles of states of corporations can be very similar to the functional roles of human mental states but that the physical realizers involved in corporations are radically unlike those one finds in humans.

We therefore gave subjects a variety of sentences ascribing mental states to corporations -- some that did not entail phenomenal consciousness, some that did.

Here are some of ascriptions that do not seem to entail phenomenal consciousness:

• Acme Corp. believes that its profit margin will soon increase.
• Acme Corp. intends to release a new product this January.

And here are some ascriptions that do seem to entail phenomenal consciousness:

• Acme Corp. is now experiencing great joy.

• Acme Corp. is experiencing a sudden urge to pursue internet advertising.

The results showed that people were significantly more likely to accept those ascriptions that did not seem to entail phenomenal consciousness. 

We then tried it out with pairs that were almost exactly alike except that one element of the pair seemed to entail phenomenal consciousness while the other did not. For example:

• Acme Corp. is feeling upset.
• Acme Corp. is upset about the court’s recent ruling.

Here it seems that only the first ascription entails phenomenal consciousness. And sure enough, people regard only the first ascription as unacceptable.

(If you want to read the full series of experiments, you can take a look at the latest draft of our paper.)

Finally, a quick foreshadowing. A number of other researchers have been doing experimental studies on closely related issues. We are especially excited about the work being done by Shaun Nichols and his colleagues at Arizona and by Tony Jack and Philip Robbins at Washington University.  Stay tuned!

Comments

[Rob Wilson]

Josh and Jesse,

Thanks for a cool experimental grip on this class of phenomena! For those interested in some historical and philosophical discussion of the idea of "group minds" in both the biological and social sciences, there are a pair of chapters on this (chh.11-12) in my Boundaries of the Mind. A restriction on the attribution of phenomenal states to groups can be found in those, like Gustav Le Bon and William McDougall in the social sciences, and William Morton Wheeler in the biological sciences (entomology in particular) who were happy enough--in fact, a little too pleased--to talk about groups minds and group-level mental states and traits. I suggest that there is an explanationist rationale for both the postulation of group minds and also for the limits of the range of mental phenomena embraced by group mind groupies. Now whether this is a folk bias is something I bet that you could run an ExP study on.

And for those wondering about the reason that you can find similiar kinds of cognitive metaphors in the biological sciences, check out pages 74-75 of my Genes and the Agents of Life. Again there's an idea there--the idea that these cognitive metaphors crystallize or amplify agency (yes, admittedly a metaphor ...)--that could be tested experimentally in terms of whether it captures what is going on in folk ascriptions of psychological states both to non-cognitive biological things (like genes) and to social groups (like corporations).

[jonathan weinberg]

Really nice work, you guys! I wonder, though, whether there is a corresponding distinction within the space of our moral attributions as well. Your results suggest that at least some moral concerns presuppose a phenomenal agent. But we do use at least _some_ kinds of moral talk with collective entities; at a minimum, we seem to make evaluations of fairness and justice with regard to them, and we definitely attribute moral responsibility (or its failure) to them. So it might make a nice further bit of work to tease apart which sorts of ethical attributions go with which. (It also seems like we _don't_ tend to talk so much about either fairness or responsibility when dealing with, say, cats and dogs, even though we definitely express moral concern for their well-being and hedonic states. So one might consider the hypothesis that a certain amount of complex functional cognition is a prerequisite for justice-talk, just as phenomenal states are for well-being-talk.)

[Chris]

Hi, first, I wanted to say that I posted on this over at Mixing Memory, and the discussion of the study in comments has been pretty good so far. They've raised some good points.

Second, the more I think about this, the less convinced I become that the data speaks to your hypotheses. I do think it's pretty clear that people have a fairly complex concept of consciousness. I suspect that there is a sophisticated hierarchy of types of phenomenal (and perhaps nonphenomenal) conscious/mental states that people are willing to attribute to different agents. A lizard, for example, can "feel hot," but most people aren't going to say that a lizard "feels depressed," while a dog can easily feel hot and depressed, but perhaps not other phenomenal states that humans can. So, I'm not disputing that, or that the data shows that one of the distinctions people make has to do with phenomenal vs. nonphenomenal states.

However, I don't think the data really say anything about physical realizers. In fact, I think the inclusion of "magic" in one condition, but not another, makes that conclusion pretty much impossible. What if we applied magic to a group agent, or, as one of my commenters suggests, a mass noun (e.g., sand)? Would people then attribute consciousness to them? It might be possible to utilize this, though. You could have the magic do a series of different types of things (related to physical realizers, e.g.) to corporations, to determine which mutation allows the corporation to be conscious. That'd be one way of more directly testing that hypothesis.

Also, I think that you've failed to take into account the complexity of people's concepts of consciousness that I just mentioned. As one of my commenters suggests, there are certain group agents that, at least when taken metonymically, are perfectly capable of experiencing phenomenal states (the White House is his example, as in "The White House was elated at today's news"). Furthermore, I think there are certain phenomenal states that people will attribute to certain group agents, but not others. I'd find it perfectly normal to say that "The New York Times is embarassed," or even that it "felt embarassed," but it would seem weird if someone told me that "The New York Times is depressed."

In order to rule these possibilities out and to draw any conclusions about physical realizers, you'd need to use a wider range of organizations and phenomenal state ascriptions in your stimuli.

Anyway, I really like the progression of the studies, but I think you need more datas to draw the conclusions you have.

[Joshua Knobe]

Chris,

These suggestions are right on target. As you say, one good way to figure out precisely what criteria people are using would be to find out what sorts of changes the sorceress would have to make in the corporation before it could have phenomenal states.

Unfortunately, there isn't (yet) any experimental research along these lines, so all I can do now is report my own intuitions. To me, it doesn't seem that there is any particular reason why a being that is spatially distributed could not have phenomenal consciousness. If I imagine that the whole corporation forms just one mind (so that the individual members do not have minds of their own), then it immediately seems to me that this one distributed mind could have phenomenal consciousness. By contrast, if I imagine that the sorceress is just casting spells that affect the emergent properties of a corporation made up of individual autonomous agents, then I find myself unable to imagine that she can make it the case that the corporation truly feels happy or sad.

If people's intuitions actually do follow this pattern, we would have reason to suppose that their criteria don't rule out beings that are spatially distributed but do rule out beings that are made up of autonomous agents.

Also, I wanted to ask your advice about how to determine whether people's statements about the corporation are literal or figurative (e.g., metonymic). Are there any experimental tests that might help one to decide between these two hypotheses?

Jonathan,

Exactly! It seems like people's intuitions about moral patiency are influenced by attributions of phenomenal states whereas their intuitions about moral agency are influenced by attributions of non-phenomenal states (beliefs, desires, intentions, etc.).

In other words, it seems like we use attributions of phenomenal states to determine whether it would be morally wrong to do certain things to a particular entity, while we use attributions of non-phenomenal states to determine whether it would be wrong for that entity to do certain things to others.

Rob,

Thanks for the references -- we'll be sure to check those out.

[Chris]

You know, there's so little research on metonymy that I can't think of anything like a standard procedure for determining whether someone interprets a sentence metonymically or not. All of the work I know on metonymy resolution is in AI and machine learning. I'll put a few references at the end of this comment.

Basically, you need a few different features:

1.) A context that mades a literal interpretation difficult.
2.) A context that fits with an existing "metonymic schema," that is, a type of part-whole relationship that people are used to using in metonomy.

The best way to do this is probably to provide people with a contrast, so that one use of a noun is obviously literal, and the other is different from the literal use in a way that highlights is metonymic function. Say, using "New York Times" to refer to the actual printed paper in one context and "New York Times" to refer to the editors of the paper in another.

I suspect that in your first experiment, where they don't really have any context to go with the sentences, participants aren't quite sure what to do with it -- should they interpret it literally or metonymically? Corporations are strange, too, as agents. I think most people in today's culture understand that corporations can take on a life of their own, so to speak, and do things that most of their employees wouldn't want. So, referring to a corporation as an agent can be done both literally and metonymically. You might try running versions of the experiment using entities that, when referred to as agents, are generally used metonymically. News papers referring to its editors might work. The references below might provide some examples that would work.

Frisson, S., & Pickering, M.J. (1999). The processing of metonymy: Evidence from eye movements. JEP: LMC, 25(6), 1366-1383.

Nissim, M., & Markert, K> (2003). Syntactic features and word similarity for supervised metonymy resolution. Proceedings of the 41st Annual Meeting on Association for Computational Linguistics, 56-63.

Markert, K., & Hahn, U. (2002). Understanding metonymies in discourse. Artificial Intelligence, 135(1-2), 145-198.

Murphy, G.L., Klein, D.E. (2002). Paper has been my ruin: conceptual relations of polysemous senses. Journal of Memory and Language, 47(4), 548-570.

[Brandon N. Towl]

Mad pain vs. Martian Pain anyone?

This was a neat survey to try. My concern, which I think was raised already, is that it is in part an artifact of the language. It might be that, across the board, we do not attribute mental states to corporate entities, aliens, etc. But we often interpret their behavior "as if" they had intentional (functional?) states. If such "as if" attributions is just metaphorical, though, it is not surprising that the causal/functional states are "attributed", but the phenomenal ones not.

But here is potential foil for you... Dennett would say that content is necessary for consciousness (or, at least that content is needed for individuating conscious mental states-- though I admit its been some time since I read Dennett). But content is a matter of adopting a certain kind of explanatory stance toward a cognizer (the intentional stance). These experiments might just show that there is a break between attributing contentful thoughts and attributing consciousness. Or some such.

Hope this helps!

[David Pizarro]

Hey Guys!

I love this stuff...

I just had a quick question that actually has little-to-no bearing on your studies (but I wanted to ask it anyway). I often read statements where people describe corporations using the plural. So, on Slashdot.org, it's often the case that when people are ranting against Microsoft (for instance), they say "Microsoft ARE planning on bombing the Mozilla foundation". Is this a reflection of their lay-theory that corporations aren't singular entities, but collective individuals? Would this make them reluctant to ascribe any of the sort of phenomonenal states that they would to individuals? Is it a social convention only, and if so would it affect their ascriptions anyway?

On another note, Daniel Wegner at Harvard and Heather Gray (a grad student there), both social psychologists, have been doing work on lay theories of mental/phenomenal states. They are actively collecting data online, although as far as I've seen none of it is about collectives/groups. http://mind.wjh.harvard.edu/

So, two barely related comments. Thomas, you can kick me off the message boards!

[Jonathan Weinberg]

David, my folk-academic understanding is that there is a UK/US difference in usage here, where folks in the US more frequently use the singular and on the other side of the Atlantic, the plural. E.g., here we almost always say "the team is working very hard this season", but in Britain they are more likely to say "the team are working very hard this season".

But I don't have any data I can cite on the matter, and I don't know whether it applies to the likes of "Microsoft" as well as "team".

[Pete Mandik]

Josh and Jesse,

I think this is interesting work but question the degree to which it reflects on (concepts of) consciousness, phenomenal or otherwise.

It looks like the folk are eager to distinguish thoughts from experiences or thoughts from feelings. But lots if not all of that stuff can occur both consciously and unconsciously. I worry that you are conflating distinctions that actually cross-classify mental states.

Cheers,

Pete

P.S. I once (once!) let my daughter watch an episode of Barney. In this episode they were explaining to the kids that robots don't have feelings. The outrage I felt paled in comparison to the outrage felt by my robots.

[Timothy Tang]

Here are my intuitions of consciousness.

When I have this feeling of being stared at, I turn around and see the person staring at me. It confirms my intuition of what is taking place in the mind of the starer.

This is not psychic phenomenal, but the interference of a person's interpretation in his consciousness unto mine.

[Joshua Knobe]

Pete,

We do make some attempt to address that issue, though I'm not sure whether you'll be satisfied with our approach.

The basic claim is that people do think group agents have emotions but that they don't think group agents have phenomenal consciousness. The evidence for this claim is that people find it acceptable to say that a group agent 'is upset' but not that a group agent 'feels upset.' So it seems that the emotion itself isn't the problem. The problem appears to be something in the ordinary concept of *feeling* (which, we suggest, is where phenomenal consciousness comes in).

Does that help to address your worry?

[Pete Mandik]

Josh,

Thanks for the response. I think the 'is upset' v. 'feels upset' distinction is highly promising and may be very relevant to my concerns, but I'm not exactly sure we are on the same page about this stuff yet. Let me say a bit more.

I suspect that there is such a thing as a folk concept of consciousness, as well as a concept of a distinction between conscious and unconscious mental states. I also suspect that the folk grasp that there can be thoughts (like the memory of what your telephone number is) as well as feelings (like a feeling of anger at getting passed over for a promotion) that can occur consciously as well as unconsciously.

I also suspect that the folk don't have a concept of so-called phenomenal consciousness, that it is instead a concoction of theorists like Block and others that doesn't map particularly neatly onto any folk notions.

Now, I think the work you and Jesse present is highly promising for shedding light on whether my suspicions are correct or just more theorist's concoctions. I nonetheless worry that describing the results in terms of phenomenal consciousness reads a lot into the data.

[marc moffett]

Just an observation, but most of the intentional states by there nature contain information about the conditions of application. For instance, if the sentence is "Acme Corp. believes that its margin will soon increase" we are providing more context than if we simply say, "Acme Corp is upset." I wonder if this isn't part of the explanation. Would the results be different if we said something like, "Given the recent preciptious drop in its margins, Acme corp is very upset?" To my ear, at least, this latter attribution is far better.

Incidentally, I disagree that "Acme Corp. is upset about the court's ruling" doesn't entail phenomenal consciousness. It is true that sentences of the form "X is upset about Psi" don't generally entail "X is upset". However, it doesn't follow from this that "X is upset about Psi" could be true of an individual X even if X did not have phenomenal states at all (putting metaphor to one side).

[Joshua Knobe]

Marc,

This is a very good point. Prior to seeing the experimental results, it definitely would have been reasonable to hold the view that people don't apply phrases like 'is upset about the court's ruling' to entities that do not have phenomenal consciousness. The key question now is whether there is any good way of explaining the experimental results while still holding on to a view like this one.

The explanation we offer is that (1) people think that corporations do not have phenomenal consciousness and (2) they see a difference between phrases like 'is upset about the court's ruling' and 'is feeling upset' such that they are willing to apply the former but not the latter to entities that do not have phenomenal consciousness.

Now, suppose one were to say that people actually don't think an entity can be upset about the court's ruling unless that entity is capable of phenomenal consciousness. This does seem to me to be a reasonable view. The key question is whether one can now offer some alternative hypothesis that explains all of the experimental results. Do you have any suggestions?

[marc]

Joshua, I am sorry, but I haven't had time to read the paper in full. My thought was that the results could be interpreted as suggested previously--intentional states carry with them a greater amount of context because of their content. Respondents aren't keying in on phenomenal consciousness per se; rather, the intentional content gives them a better handle on why someone would be attributing a mental state to a corporation in the first place. This interpretation would be supported if sentences like "Acme Corp. is upset about such and such" do entail phenomenal consciousness. But it seems to me that the real test to control for this issue is to give "pure" (rather than "mixed") phenomenal state attributions together with about as much background as is provided by the intentional state attributions. For instance, "Due to the recent precipitous drop in its (their?) margins, Acme Corp. is very upset." I doubt the numbers will be as strong as with intentional states, but I would be willing to bet that a significantly larger number of respondents would find these sorts of contextualized phenomenal state attributions acceptable than would find bare attributions such as "Acme Corp. is very upset."

On reflection, I also think you have to be careful about which phenomenal state predicates you choose because you might end up with some stereotyping effects. For instance, it might be true that when we think about the stereotypical business leader we think of them as emotionally staid. If so, then we might get some bleed-over to the corporation. As a result, respondents might find phenemenal state attributions like "concerned" or "optimistic" more acceptable than "upset" or "happy". And there might be some predicates that we feel go natural with corporations. For instance, what about "Acme Corp feels nearly unstoppable"?

[marc]

It occurred to me that the following might also support the claim I am making. First, it is odd to attribute quantified belief statements to corporations.

1. *Acme Corp. believes something.

Second, it is odd to attribute beliefs claims with content for which it isn't easy to fill in the context.

2. ?Acme Corp. believes that Merriam's birthday is in early August.

[Joshua Knobe]

Marc,

These are really excellent points, and I'm not exactly sure how to account for the intuitions you point to here. It certainly doesn't seem that people are generally unwilling to ascribe mental states to corporations in the absence of context. Consider: "There's something important that Microsoft doesn't know here." "Microsoft just made a big decision." "That is what Microsoft wants more than anything else."

Similarly, it doesn't seem like one can ascribe phenomenal states to group agents just by adding context. It sounds very wrong to me to say, e.g., "Because of its declining profits, Microsoft is now feeling extremely depressed." (But, of course, we haven't tested this experimentally -- it could be that I'm mistaken here.)

Still, you are right to say that there is something odd about sentences like: "Microsoft believes something." I really don't know why that is. It could be that these sorts of sentences can't be applied even to individual agents. (Compare: "George believes something.") But it also might be that there is some more complex explanation (e.g., something about focus) that is now eluding me.

[marc moffett]

Joshua, the Microsoft example is interesting, but I am not sure that there isn't enough default context available. We are pretty familiar with the idea of corporations not having sufficient market information and thereby making poor business decisions.

I had two further, but related, thoughts. One is that many of these attributions sound to me a lot like kind-referring generics. Kind referring generics don't admit of accidental generalization. So, for example, (1) is true while (2) is false even in a situation where every lion is asleep.

1. The Lion eats meat.
2. The Lion is asleep.

By analogy, it might the case that phenomenal state attributions are odd because our "theory of business" is highly intentional: we tend to think of business in terms of decision, planning, information, etc., and not in term of emotional responses. If so, then from the point of view of business theory, the phenomenal state attributions look more accidental (in the sense of not obviously falling within an explanatory theoretical framework).

This makes me wonder if other collective entities might take phenemonal state attributions more readily. This certainly seems to be the case for sports organizations:

3. The Yankee Organization has been embarassed.

Similar points hold for Universities and Nations:

4. The University of Miami has been embarassed.
5. The United States feels besieged.

Even investment organizations also have a sense of "feel" for the market which makes some phenomenal state attributions pretty natural:

6. Merril-Lynch is feeling bullish.

Well, at any rate, I hope these reflections prove useful.

Best, Marc

[John K. Alexander]

2 comments

1) re Enchair, why not offer a second situation where the business person does not find good jobs for his employees and compare results?

2) Also, I do not know what your sample group is, but it would be interesting to test a group made up of businesss people and members of business organizations (do not include educational institutions) and compare the results.

[Joshua Knobe]

John,

With regard to your second point, I think you're right to suggest that people's responses are shaped not just by their recognition that corporations are group agents but also by the attitudes they hold toward corporations in particular.

Thus, it seems like people's intuitions might have been different if we had used names of countries instead of names of corporations. To me at least, it sounds a lot better to say 'The nation of Israel is feeling very upset' than it does to say 'Microsoft is feeling very upset.'

For some reason, it seems that a phrase like 'the nation of Israel' can be used to refer to the various individual people who inhabit the nation but a word like 'Microsoft' can't be used to refer to the various employees of the corporation. I am not exactly sure what accounts for this difference, but it might have something to do with the way we ordinarily think about the relationships between people and the different types of group agents they can be affiliated with. Perhaps we think that being Israeli is a truly deep and essential feature of Israeli people but that being a Microsoft employee is not a deep or essential feature of Microsoft employees. (People frequently change employers without changing their identities in any very important way.)

If this hypothesis turns out to be correct, then (just as you suggest) we might find different intuitions among people who have different attitudes toward corporations.