Seeing things as people: anthropomorphism and theory of mind in mixed societies

Stuart Watt
Knowledge Media Institute and Department of Psychology
Open University
Walton Hall
Milton Keynes, UK

Introduction

This paper will look at theory of mind from a rather different perspective, that of inanimate objects.  At first, this might seem a little strange, but one of the peculiarities of this field is that we sometimes think of some objects -- trains and computers, for example -- as if they were people.  We begin to use intentional terms to talk about these things doing, believing, thinking things, in the same way that we talk about people.  In this paper, then, I will try to take apart some of the distinctions that are going on in our common sense distinctions between people and objects; to look at why we suddenly sometimes think of an object as an object and sometimes think of it as a person.  I also want to look at two contrasts: the contrast between a disposition and a theory which shows up a bit in theory of mind, and the contrast between the physical and the mental realms through which we tend to think of things.  The issue that is central to this paper, the glue that holds together these disparate themes, is anthropomorphism, what it actually is, how it affects us, why we sometimes see things as other than they actually are physically, and what effect this may have on our theory of mind. 

So why does the title of this paper refer to mixed societies?  Well, we all live in a world where we have other people around us, yes, but also a lot of animals and objects.  So far, the research on theory of mind has concentrated on people and to a lesser extent animals, although some have looked at objects too (e.g. Gelman, Durgin, & Kaufman, 1995).  But the world we live in is a mixed one -- we interact all the time with non-human systems as well as human systems, we interact with both animals and objects on a day-to-day basis.  How is it that we can see them as having minds, when they are so very different to us.  For example, how can people and animals can share a language.  If you interact with a cat you eventually come to a point where you can read the cat's gestural language.  Of course, this doesn't necessarily mean a language of the same scale as a human language, but even so, it does seem that we can sometimes recognise what an animal is thinking or feeling, at least to an extent. 

Figure 1 shows a cartoon which shows some of these issues; it is by Mel Calman from Humphrey (1987).  It depicts a spectrum of possibilities between things which are wholly conscious and which have a rich mental life (such as people) and things which are simply not there mentally (such as flowers).  People, animals, and objects occupy different places on the continuum of possible mental life.  But the interesting question is: why is this the case?  Why isn't there an all-or-nothing distinction between things which have minds and things which don't? 

Figure 1.  Gradations of mental life by Calman (from Humphrey, 1987).

So the main point of this paper is this difference between people and objects, between the mental and the physical.  Of course, one argument might be that it is simply that people are more complicated, both socially and biologically, and for this reason, people are intrinsically different from objects.  As I will suggest, this isn't particularly convincing; for one thing, our distinction between people and objects is only partly to do with them, with their form and their behaviour; in great measure it is also a property of ourselves as observers, and specifically with our theory of mind and with anthropomorphism.  To show this more clearly, I'll move on to look at some examples of anthropomorphism and how it can influence people to see objects as if they have minds. 

Anthropomorphism and theory of mind

What is it that makes us see things as minds rather than bodies?  Sometimes it is almost as if there's a kind of leakage from a human theory of mind to animals and to objects.  For example, think of a humble thermostat.  The mental life of a thermostat is actually one of the older chestnuts in artificial intelligence; McCarthy (1979), for instance, has said that we can say that a thermostat believes its environment is too hot or too cold, and argues forcefully that 'believes' is a valid term to use even for a thermostat -- that it is not simply being anthropomorphic.  Of course, with people, and often even with cats and chimpanzees, we'd have no problem using the term 'believes', at least on an everyday level. 

Dennett makes this idea more precise.  He suggests that terms like 'believes' are actually terms that belong to an "intentional stance" (Dennett, 1971), where we think about the behaviour of a system as if it had a mind of its own.  The intentional stance just happens to be one way of seeing and thinking about a system -- others include the "physical stance" and the "design stance", where you think about the system in terms of its physical structure or functional structure respectively.  Put simply, Dennett's argument is that for some systems, taking the intentional stance is generally the best way to guess what that system will do next, and that when we're thinking about people in particular, taking the intentional stance just comes naturally. 

I suggest that the intentional stance and anthropomorphism are connected elements, and that with theory of mind, they form different aspects of human common-sense psychology.  Dennett himself admits that the intentional stance is "anthropomorphic" (Dennett, 1987).  But there is an important difference: both theory of mind and the intentional stance look at how someone can reason about other people's mental life, but neither of them looks at the other side of the coin -- how to decide whether or not they have mental life in the first place, how to decide whether or not it is an agent or an object that we are dealing with.  It is here that anthropomorphism may play a role, acting as a tendency, a disposition, to see some things (people, for example) as intentional objects with minds, and others (rocks and computers, for example) as physical objects without them.  By labelling anthropomorphism as a disposition, I want to contrast it with the idea of a theory, a contrast I will look at in more detail.  For now, all I want to suggest is that, through anthropomorphism, there are definite factors, such as physical form, which do seem to influence when we see something as having a mind in practice. 

The distinction between a disposition and a theory is not a new one (Ryle, 1949).  A theory, even a theory of mind, is usually taken as something which behaves to some extent as if it was a collection of laws, rather like Newton's laws of motion, although really the word 'theory' shouldn't be taken that literally (Goldman, 1993).  This is not to say that people are governed, or even that they can be described, by absolute laws, but that a theory of mind works, at least to an approximation, as if rules like "if I can see you, you can see me" are being used to reason about other people's minds.  A disposition, on the other hand, is more like a bad mood: there may be nothing definite in the behaviour which is 'caused' by the mood, but overall, certain definite behaviours are a bit more likely than they would be otherwise. 

Even though the distinction between a disposition and a theory was originally a philosophical one, there may still be a role for it in common-sense psychology.  We can see corresponding themes in theory of mind, for example in the contemporary debate between the theory theory and the simulation theory (e.g. Davies, 1994; Gordon, 1986; Perner, 1994; Stich & Nichols, 1992).  At some level, both the theory theory and the simulation theory contain both theoretical and dispositional aspects, but I am not convinced that either alone offers enough to describe human common-sense psychology completely.  You need something else, you need something which enables you to distinguish between when to apply a theory of mind and when not to apply a theory of mind, and that something just isn't covered by beliefs, desires and intentions on their own. 

This is all tied up with a real problem, a methodological problem of anthropocentricity.  Fundamentally, this is very simple: how can we step outside our humanity?  We all know what it's like to be a person.  We all know what it's like to be ourselves, or at least we think we do, but how can we step outside that and try to see what it is like to be animals?  Is it possible to see things from outside the point of view of our own human theory of mind?  Some people (e.g. Dennett, 1987; Gopnik, 1993) think that it is; others seem to suggest that we are stuck with some level of endemic anthropocentricity and egocentrity (e.g. Hobson, 1993; Nagel, 1974; Searle, 1992). 

This apparent endemic anthropocentricity and egocentrity certainly does seem to have caused methodological problems, particularly in the early days of common-sense psychological research, when people were trying to draw lessons from ethology and apply them in psychology.  For example, the main methodological criticism against Premack and Woodruff (1978) was that the experimenters might be ascribing their own ideas about the situation to the actors involved.  False belief tests were needed so that the experimenters' and the actors' beliefs could be distinguished. 

The same goes for thought experiments, which are particularly common both in philosophy and artificial intelligence.  There are real problems when you try and ascribe mentality to black boxes, sometimes it is very easy, sometimes it is hard, but the difference between the two depends at least as much on the observer as it does on the boxes involved.  For example, Watt (1997) shows that we 'try to get inside' models like the information processing metaphor so common in cognitive psychology, and that this can strongly influence the intuitive appeal of different models, irrespective of the actual power of the model itself.  Figure 2 demonstrates some of these effects: simply by changing the physical form, by merging it with the human physical form we change the patterns of ascription.  It becomes easier to ascribe mentality to the box because it looks partly like a person.  We are scientific psychologists, and yet we see our models through our own common-sense psychology.  There is no view from nowhere. 

Figure 2.  Superficial effects of physical form on the ascription of mentality. 

So anthropomorphism does seem to play a role in the ascription of mentality, a role far removed from the kind of 'Thomas the Tank Engine' psychology that it might seem at first.  Before we can reason about other people's minds, we need something which tells us when reasoning about other people's minds makes sense, and when it doesn't.  Anthropomorphism may be at the heart of how we distinguish between things which have minds and things which don't. 

Dennett's (1971) description of the intentional and other stances has a similar problem.  Along with the three stances, Dennett suggested that there was a "rationality assumption" involved in taking the intentional stance.  His argument was that, for some systems, we assume that they are 'rational', and that this means we can take the intentional stance to them.  Unfortunately, as the example in figure 2 shows, the rationality assumption is far from rational, a point made rather gleefully by, among others, Fodor (1985).  In figure 3 I show that, in practice, the overall structure of Dennett's model looks suspiciously like a 'philosophicalisation' of common-sense reasoning in the psychological realm, where the intentional stance corresponds to common-sense psychology, and so on.  And in this model anthropomorphism -- or one facet of anthropomorphism -- actually corresponds to Dennett's rationality assumption.  It is anthropomorphism that makes us take an intentional stance, rather than a physical stance, to a system. 

Figure 3.  Correspondence between the philosophical and psychological realms. 

So there is a tension between physical, functional, and psychological theories and an important part of this model is that we have to know when to use the right kind of theory.  It is possible to reason psychologically about a thermostat; to say a thermostat believes it's too hot or a thermostat believes it's too cold, but for most of us, at least for those of us that know how a thermostat actually works, that would be just pushing things a bit too far.  Alternatively, we can take a physical stance, and talk about the little bits of wire inside the thermostat being bent out of contact because of the temperature.  When we're thinking about the thermostat, we have to decide when it is appropriate to apply each stance.  This is where the rationality assumption -- and therefore anthropomorphism -- may play a role. 

Evolutionary origins of anthropomorphism and theory of mind

Anthropomorphism has scarcely been studied at all in psychology, and its role in any models of the evolution of a theory of mind will, therefore, be a bit sketchy.  Even so, there are a few biological and evolutionary models for altruism which seem to give a few hints, and even though there may only be the vaguest shadows of them left inside us today, they may still form some of the essential precursors to our human ability to ascribe mentality to others. 

Dawkins (1989) following (Hamilton, 1964), discusses the "green-beard effect".  The green-beard effect is one possible genetic model of altruistic behaviour.  It works like this: there is a gene which has two effects, first it makes you grow a green beard, and second, it affects your behaviour so that if you see somebody else with a green beard, you behave altruistically towards them.  There is a label which is either absent or present, and if you see someone and recognise this label, you act altruistically to that person.  And it turns out that a gene with these two effects will be successful in a population, but that it is vulnerable to "outlaws" which display the label but which do not behave altruistically to others who also show the label.  Instead, Dawkins suggests that very similar altruistic behaviour can be caused by an "armpit effect" where, instead of recognising a label, you have a tendency to sniff people's armpits, and according to the similarity between your own armpit smell and somebody else's armpit smell, you act altruistically towards them.  This isn't vulnerable to outlaws in the same way.  This principle is a general one, and any measure of similarity between you and another will do as well. 

How does this relate to anthropomorphism?  Well, it suggests that anthropomorphism -- if it based on a biology of reciprocal altruism -- could be based on a measure of similarity between the ascriber and the system, rather than on any 'absolute' features of the system.  Building on this idea I have developed a more complete model of anthropomorphism, a model which consists of these six primary factors:

• Similarity; how similar are you to the system, physically and behaviourally?
• Familiarity; how familiar are you with the system?
• Animation; to what extent is the system an animate object rather than an inanimate one?
• Structure; what you know about the structure of the system and how it works?
• Context; what is the current social context?
• Medium; through what medium are you interacting with the system?
  

Of these, perhaps the most dramatic factor is the similarity factor, based purely on the similarity of the physical form.  Not only is this consistent with the possible model of anthropomorphism as an armpit effect with an origin in a kind of reciprocal altruism, it also shows up strongly in experimental results.  Figure 4 shows results from Eddy, Gallop and Povinelli (1993), who carried out one of the very few experiments on anthropomorphism in the psychological literature.  In figure 4 there is a clear correlation between estimated similarity and ascribed mentality -- the more similar the system, the more likely it is that people will ascribe mentality to it.  As a subsidiary theme, though, people consistently tend to underestimate similar experience, or consciousness, when compared to determining intent, or the cognitive power of the system.  A consistent tendency to underestimate consciousness compared to cognition for the same system is an interesting phenomenon in its own right, and one that deserves fuller study. 

The results in figure 5, from the same study, are perhaps even more startling.  Here, once again, there is the clear correlation of estimated similarity with ascribed mentality, but this time, the correlation with phylogenetic similarity also shows up strongly.  Estimated similarity turns out to correlate well with the genetic similarity between the systems, shown by the different phylogenetic groups. 

Figure 4.  Correlation of estimated similarity with ascribed mentality. 

Figure 5.  Correlation of ascribed cognition with estimated and phylogenetic similarity.

There are some important points to be drawn from these results.  First, it really does look as if people ascribe mentality to systems on a continuum rather than according to a simple all-or-nothing criterion.  Second, similarity -- even genetic similarity -- does seem to play a significant role in people's common-sense ascription of mentality.  It really does seem plausible that something as simple as physical form does strongly influence when people see something as having a mind.  Certainly, deciding whether or not something has a mind is not as simple as looking to see if it seems to move all by itself (Gelman et al., 1995; Shultz, 1991). 

The other factors in this model of anthropomorphism deserve a little more discussion.  First, familiarity seems to be a kind of learned factor, and also showed up in Eddy et al.'s (1993) study, where they found (in figure 5, for example) that familiar animals, such as cats and dogs, were ascribed mentality rather more readily than ought to be the case.  Animation, often taken to be the really important factor in ascribing minds to something (Shultz, 1991), is, of course, still an important factor.  And animation is also important in anthropomorphism; the objects we tend to anthropomorphise usually move to some extent -- trains, aeroplanes, clouds, and so on.  Structure seems to be a rather strange factor; the more you know about how and why something behaves the way it does, the harder it may be to ascribe mentality to it.  The theory behind this is still unknown, but it does seem to be an important effect, and it may play an important role in some of the occasional entanglements of common-sense and scientific theories (Watt, 1997; Wilkes, 1991).  Next, the context matters; if you're reading Thomas the Tank Engine books you can ascribe mentality to trains, but the very next moment, you won't ascribe mentality to a train you've been waiting for, when it is late at the railway station. And finally, the interaction medium also has an effect; a richer medium of interaction medium tends to accentuate the differences in ascription, a poorer one will tend to obscure them. 

So, very tentatively, we can suggest a number of steps which could lead to the evolution of a theory of mind.  First, there is species recognition -- the ability to recognise other members of your own species -- this is needed for reproduction, anyway.  This provides enough of a trampoline to the next stage, to providing a measure of similarity; it provides the perceptual capabilities that would enable armpit effect reciprocal altruism to work.  And through this reciprocal altruism, there will inevitably be a disposition to increase interaction with those that are similar to you.  This might then lay a solid enough foundation for the evolution of a full theory of  mind.  All in all, then, there does seem to be an evolutionary story for theory of mind which is roughly Darwinian; it didn't need to come together in one big lump, and it could easily have happened in lots of little stages throughout our evolutionary history. 

According to this model, then, anthropomorphism is pre-social in that it forms a disposition to act altruistically to other things that look like you.  This could then act to bring together societies which didn't exist before.  But more than this, anthropomorphism is genetically adaptive as a kind of reciprocal altruism.  So this kind of 'anthropomorphic altruism' story does seem to have a measure of biological plausibility, while being corroborated by Eddy et al.'s (1993) results.  Certainly, the role it plays in a theory of mind -- the role of deciding when to ascribe mentality -- is one that deserves more study; it may offer very important clues to the underlying nature and evolutionary history of theory of mind. 

Conclusions

There are some conclusions we can draw from all this.  First, anthropocentricity is still a big problem.  We haven't really addressed it; we're still stuck inside our human kind of thinking.  At the end of the day we've really no definite idea what it's like to be anything -- or anyone -- else.  So 'theory of mind' means 'theory of human mind' with a kind of ominous inevitability. 

Second, anthropomorphism seems to form part of theory of mind in one sense, in that it enables us to decide when to -- and when not to -- ascribe mental states to things.  This is a fundamental issue that should be addressed within any useful or complete theory of common-sense psychology.  Anthropomorphism, however, shows clear patterns of being dispositional rather than theoretical in structure -- the difference between having a mind and not having a mind is not all-or-nothing, at least as far as human common-sense psychology is concerned.  Instead, it is a tendency to ascribe mentality, a tendency that may be influenced by many possible factors. 

If the anthropomorphic model has anything to offer, it would suggest that people have an innate sensitivity to agency -- an innate sensitivity that is not simply based on uncaused motion.  It is through this that social relations can emerge, and which might enable the more complicated, the more symbolic and re-described theory of mind to come to full flower.  Anthropomorphism is certainly an intriguing area for future research and experiment on theory of mind; much of the area is still unexplored, and it may yet have a lot to offer -- certainly a lot more than is apparent at first glance.  Who knows: Thomas the Tank Engine may hide many secrets of the human mind. 

References

Davies, M. (1994). The Mental Simulation Debate. In C. Peacocke (Ed.), Objectivity, Simulation and the Unity of Consciousness (pp. 99-128). .

Dawkins, R. (1989). The Selfish Gene. Oxford: Oxford University Press.

Dennett, D. C. (1971). Intentional Systems. Journal of Philosophy, 68, 87-106.

Dennett, D. C. (1987). The Intentional Stance. Cambridge, Massachusetts: MIT Press.

Eddy, T. J., Gallup, G. G., & Povinelli, D. J. (1993). Attribution of Cognitive States to Animals: Anthropomorphism in Comparitive Perspective. Journal of Social Issues, 49(1), 87-101.

Fodor, J. A. (1985). Fodor's Guide to Mental Representation: The Intelligent Auntie's Vade-Mecum. Mind, 94, 55-97.

Gelman, R., Durgin, F., & Kaufman, L. (1995). Distinguishing Between Animates and Inanimates: Not By Motion Alone. In D. Sperber, D. Premack, & A. J. Premack (Eds.), Causal Cognition: A Multidisciplinary Debate (pp. 150-184). Oxford: Oxford University Press.

Goldman, A. I. (1993). The Psychology of Folk Psychology. Behavioural and Brain Sciences, 16, 15-28.

Gopnik, A. (1993). How We Know Our Minds: The Illusion of First-Person Knowledge of Intentionality. Behavioral and Brain Sciences, 16(1), 1-14.

Gordon, R. M. (1986). Folk Psychology as Simulation. Mind and Language, 1(2), 158-171.

Hamilton, W. D. (1964). The Genetical Evolution of Social Behaviour. II. Journal of Theoretical Biology, 7, 17-32.

Hobson, P. (1993). Understanding Persons: The Role of Affect. In S. Baron-Cohen, H. Tager-Flusberg, & D. J. Cohen (Eds.), Understanding Other Minds: Perspectives From Autism (pp. 204-227). Oxford: Oxford University Press.

Humphrey, N. K. (1987). The Inner Eye. Channel Four.

McCarthy, J. (1979). Ascribing Mental Qualities to Machines. In M. Ringle (Ed.), Philosophical Perspectives in Artificial Intelligence (pp. 161-195). Brighton: Harvester Press.

Nagel, T. (1974). What is it Like to be a Bat? Philosophical Review, LXXXIII.

Perner, J. (1994). The Necessity and Impossibility of Simulation. In C. Peacocke (Ed.), Objectivity, Simulation and the Unity of Consciousness (pp. 145-154). Oxford University Press for the British Academy.

Premack, D., & Woodruff, G. (1978). Does the Chimpanzee Have a 'Theory of Mind'? Behavioural and Brain Sciences, 4, 515-526.

Ryle, G. (1949). The Concept of Mind. Hutchinson.

Searle, J. R. (1992). The Rediscovery of the Mind. Cambridge, Massachusetts: MIT Press.

Shultz, T. R. (1991). From Agency to Intention: A Rule-Based Computational Approach. In A. Whiten (Ed.), Natural Theories of Mind: Evolution, Development and Simulation of Everyday Mindreading (pp. 79-95). Oxford: Basil Blackwell.

Stich, S., & Nichols, S. (1992). Folk Psychology: Simulation or Tacit Theory? Mind and Language, 7(1).

Watt, S. N. K. (1997). The Lion, the Bat, and the Wardrobe: Myths and Metaphors in Cognitive Science. In S. O Nuallain, P. Mc Kevitt, & E. Mac Aogain (Eds.), Two Sciences of Mind: Readings in Cognitive Science and Consciousness . Amsterdam: John Benjamins.

Wilkes, K. V. (1991). The Relationship Between Scientific Psychology and Common Sense Psychology. Synthese, 89, 15-39.