Introduction

In a study of the evolution of purposive intentionality in artificial life, I reviewed the historical criteria for attributing purposiveness and intention to behaving systems (Watson, 2005). I found three primary criteria for the assessment of purposiveness: equifinality (displayed equivalent outcomes by varying behavior across instances of non-equivalent situations as illustrated in Fig. 1), rationality, and perseverance. Two additional criteria appeared relevant to a claim that the purposive system was also intentional. If the system met a criterion of "equi-origin" (displayed variable behavior across instances of equivalent situations as illustrated in Fig. 2), the system was said to meet what I termed "weak intentionality". If behavior supported a claim that it was directed by a representation of the outcome (that it met the philosophical notion of "aboutness"), the system was said to meet "strong intentionality". Although the analysis would surely not satisfy everyone, I tried to accommodate important criteria that have arisen in psychology, philosophy, and jurisprudence while not straying too far from the man on the street. In this paper, I propose that equi-origin may logically trigger the intentional stance.

Taking a Stance

            Recent theorizing about cognitive-perceptual development has been influenced by Dennett's use of the concept of stance. His proposal and analysis of the intentional stance, the design stance, and the physical stance (1971, 1987) has had seminal influence on many developmental psychologists in their consideration of when children detect intentionality in others (Gergely, Nadasdy,  Csibra, & Biro, 1995; Gergely & Csibra, 1997;   Keleman, 1999, Leslie, 1995;  Premack & Premack, 1995). The more recent proposal of the infant's teleological stance ( Csibra & Gergely, 1998; Csibra, Gergely, Biro, Koos & Brockbank,1999) is a distinct concept but is clearly derivative in form. In each case, the stance is proposed as a cognitive-perceptual filter or bias that affects how the infant interprets events. Theorists have proposed specific cues that govern when the infant will assume the stance in question. For example, many have proposed that the cue of agency (self propelled motion) will invoke the intentional stance (Baron-Cohen, 1994; Leslie, 1995; Premack & Premack, 1995; and see Heider & Simmel, 1944 and Heider, 1958 for related discussions). Equifinality of action is proposed as a provocative stimulus for the teleological stance (Csibra, Gergely, Biro, Koos & Brockbank, 1999). The work of Lewis (1990) raises the possibility that emotional behavior in relation to success and failure of instrumental behavior may provide cues to intentional states.

            Dennett's original argument for why we take the intentional stance is that it is pragmatic to do so. It is an evolved strategy for coping with the behavior of others.  Evolution also provided selective sensitivity to certain cues, as noted above, regarding when it would be helpful to take the stance.  From this view, cues do not logically force us to claim the existence of internal states, they simply alert us to the potential benefit of doing so.

            The objective of this paper is to propose that the intentional stance can be viewed as logically forced by a more primitive stance: the determinist stance. From this view, the attribution of intentionality can be a logical necessity under certain conditions. What then is this more primitive determinist stance?

The Determinist Stance

            The determinist stance ¹ is a view of the world as lawful.  Events are caused by other events. An event may (or may not) cause other events to happen. When a salient event occurs, it provokes the quest to find its cause. Formulating the lawfulness of the world allows one to predict future events and to understand/explain past events. An important assumption about the determinist stance is that it frames lawfulness as complete and universal. That is, a valid or good law is not sometimes true and sometimes false. Neither is it somewhere true but untrue in other places. From the determinist stance, a law is valid only if it is universally applicable.  In the case of complete determinism: All things being equal, if a sufficient cause occurs then the effect should occur. Even from the perspective of probabilistic determinism: All things being equal, if a sufficient cause occurs then the effect should occur with specified probability over repeated instances (e.g. not .8 sometimes and .2 at other times).

            A notable feature of good laws from the determinist stance is that they are not symmetrical in the relation of cause and effect. With a good law, if you know the causal context is complete, then you know the exact effect that will occur (or, in probabilistic determinism, the probability of that effect).  But the reverse is not entailed. Alternative causes may exist for a specific kind of effect.² In such cases, knowing that a specific effect has occurred leaves open the question of which of the alternative causes was the determinant.

Fixing a Failure of Universality

            When a determinist formulates a law of behavior, the objective is to identify the situational event (S) that causes the behavior (B).  This lawful relation is often rendered graphically as:

S ->B, or by the equation B = f S. If this law is meant to apply to people, then it needs to be applicable to all people all of the time. There are two ways this assumption of universality can fail, however. It might turn out that it seems valid for some people but not others. Or it might seem valid sometimes but not at other times for any given person. The determinist stance can not tolerate either failure because there is no wiggle room in this stance for free will or any other indeterminacy. Historically, determinists have found a conceptual cure for both of these potential failures of the universality principal. The cure comes in the form of dispositional property attribution (Armstrong, 1968,1969, 1993; Carnap, 1938; Prior, 1985; Ryle, 1949; Watson, 1995).

Regaining Universality with Dispositional Properties

Dispositions of Kind. When a behavioral law appears to vary across people, psychologists introduce dispositional constructs of individual difference or personality.  These are dispositional properties of kind. An example is given in fig. 3.

Disposition of State. When a behavioral law appears to vary across time for particular people, psychologists introduce dispositional constructs of state (as in appetitive and emotional states) or stage (as in stages of learning and development).  An example is given in Fig. 4.

A logical trigger for mental state attribution

A determinist tries to find a situational contrast to explain an individual's change of behavior from one time to another, but failing that, is logically forced to introduce a dispositional property of state. The logical pressure derives from a judgment that the situations are equivalent, and thus, by exclusion (or successive negation of disjunction), a non equivalence in individual is implied. The principle of universality requires that this claim of non equivalence be framed so as to be, in principle, a claim about any individual. The dispositional property claim meets that requirement. All people are comparable in reference to the dispositional property variation just as all situations are comparable in reference to their stimulus variation.  Thus, when determinists are lead to believe that in equivalent circumstances the actor has or could have acted differently, they are forced to introduce a dispositional factor in order to avoid the failure of determinant lawfulness.

To believe that an individual could have acted differently in the same circumstances is a counterfactual assumption (in jurisprudence, support for this assumption is central to assessment of intent).  If such counterfactual claims were derived solely from imagined alternatives, then there would be little justification for introducing dispositional properties as additional factors in empirical laws that relate situations to behavior.  Empirical laws are not obliged to explain imagined events. But the case is very different when the alternative behaviors have been observed in equivalent situations.  These alternatives, if they are to be accounted for in a single law, force that law to include reference to a dispositional variation across the equivalent situations.

Implications

            Einstein's Baby.  Bloom (2004) has proposed that human infants appear to be innately committed to a view that there are two kinds of lawfulness in the world: determinant laws of the physical bodies in the world and self-determinant (or indeterminant) laws of humans and possibly other animate beings that are affected by the intentions of their "soul". This causal dualism is aptly captured under Bloom¡'s metaphoric book title Descartes' Baby. Bloom is not arguing for the truth of such dualism, only that it appears to be an innate bias.  To highlight the alternative view that I am proposing, I have introduced the above heading. Einstein's baby is a determinist, and by my reckoning, she will thereby be logically forced³ to attribute intentions (and other mental states) as dispositional properties of state to cope with the behavior of others. In my view, she does not need a second form of law (e.g., God's dice) to detect and cope with the mental states of others.

             Consider the case of an infant learning to adapt to his mother's interactive style.  If he is to gain even a small degree of control of the interaction sequence, he will need to anticipate her reactions to his behavior. Now if he views his mother's behavior as independent of external causes, then there is nothing to learn save perhaps her general tendencies of behaving one way or another. Alternatively, if he views her behavior from a determinist's stance, he is set to relate her behavior to potential causal events in her environment including his own behavior. His task will be to uncover the causal factors. He will discard any notion that she might be behaving indeterminately. He may use this stance to refine his classification of what he should record as equivalent behavior on his part and what he should record as equivalent behavior on his mother's part (see Watson, 2001). Once his judgment of equivalence is settled, he is in position to feel the logical pressure for attribution of mental state change in his mother. This will arise whenever he observes her behavior fulfilling the criterion of equi-origin: her behavior varying (nonequivalent) across instances of equivalent situations (equivalent in terms of the setting and his behavior).

            As a simple concrete illustration, I will shorten an example I have used previously in a related discussion (Watson, 1995). Imagine an infant who enjoys extended interaction with his mom. Sometimes when she tucks him in and he smiles at her, she turns out the light and leaves. Sometimes, however, she tucks him in and if he smiles, she picks him up again and interacts with him for a while longer. Alternatively, if he pouts when she tucks him in, sometimes she leaves but other times she picks him up and interacts a little longer. As a determinist, our infant rejects the notion that his mother is behaving randomly. The fact that she might be caused to stay (or likewise caused to leave) by a smile or a pout presents no special problem, since lawfulness allows alternative causes for a specific kind of effect. However, the fact that his smile (and likewise his pout) appears to sometimes cause her leaving and sometimes cause her staying in an otherwise equivalent situation calls for remedy of the infant¡¯s formulation of the lawful efficacy of his behavior. He needs to posit a disposition of state in his mother in order to maintain his commitment to determinism. With additional experience on his part, we might imagine (as in Watson, 1995) that he will eventually uncover cues to her state variation that he has been logically forced to assume. If he is successful, then he will cope well with the fact that, in this example, his mother sometimes has rewarding days at work and sometimes has days that have depleted her reserves of nurturance.  On rewarding days she views his smile as a sign he is content (so she leaves) but his pout as a sign that he needs additional comforting (so she picks him up). On her difficult days, she views his smile as a sunny reprieve (so she picks him up) but his pout as a sign that he is asking for unnecessary support that she is not prepared to give at this time.

            Animacy as equi-origin.  Animacy (self generated motion) has been proposed as a prime cue for eliciting the intentional stance and/or purposiveness (Baron-Cohen, 1994; Leslie, 1994; Premack,1990, Premack & Premack, 1995).  Csibra, Gergely, Biro, Koos, and Brockbank (1999) provide evidence that animacy is not necessary for an infant¡'s interpretation of behavior as purposive by the criteria cited above. Cues of equifinality appear to be sufficient.  However, if the attribution of intentionality stands on some evidence of equi-origin as proposed above, then animacy may have a special relevance to that attribution. The standard display of agency introduces evidence of equi-origin because the focal object's behavior varies in an otherwise constant situation. The present argument for equi-origin as a logical elicitor of the intentional stance would suggest that if change in motion of an object where immediately preceded by a salient change in the situation, then the behavior would be resolved as a simple situational determinant law. By contrast, if the change in motion were not accompanied by a situational stimulus, as is usually the case in animacy displays, then the behavior would trigger the inference of a dispositional state change in the object and the essential basis for attribution of intention would be provided.

            Equi-origin and theory of mind. If equi-origin is a logical trigger for the infant's attribution of dispositional states, then attribution of belief should be affected by this trigger.  A study by  Onishi & Baillargeon (manuscript under review) for assessing sensitivity to false belief in infants would lend itself nicely to testing the triggering effect of equi-origin. The false belief test in theory of mind  research (Bartsch & Wellman, 1989; Dennett, 1971; Gopnik & Astington, 1988; Perner, Leekam, & Wimmer, 1987) was presented in a manner allowing for assessment of much younger subjects  (15-month-old infants versus 4-5 year-olds). With the violation-of-expectation method, infants were shown an object being hidden within one of two boxes while a person was present. In the focal test sequence (versus a number of control sequences), the hidden object changed place from one box to the other while the person was absent.  The person returned and reached into one or the other of the boxes. Extended fixation (the criterion of surprise) was observed when the person reached to the true hiding place. This is taken as evidence that the infant assumed the person held a false belief regarding the whereabouts of the hidden object and expected a reach to the wrong box.

Onishi and Baillargeon did not manipulate equi-origin and they used a person as their focal object. A small change in their method could assess the roll of equi-origin in infants' attribution of this dispositional/mental state to even novel non-human objects.   As depicted in left panel of Fig. 5, an infant could be shown the focal object making a choice between two occluders. At the moment of choice, there is a stimulus equivalence of situation from one trial to the next (1b versus 2b and 3b versus 4b), the only difference being an historical reference to where the hiding took place. If the infant understands this choice as based on the focal object's dispositional state of goal representation, as forced by the equi-origin evidence, then a slight variation in procedure should interfere with the infant taking this intentional stance. The variation involves introducing a signal (e.g. appearance of a black oval or a white oval at the moment the focal object begins to move) that is consistent with where the object is hidden. This stimulus can be viewed as a simple external determinant of the focal object's behavior as depicted in the right panel of Fig. 5.  In this case, evidence of equi-origin is not present. By the view I am proposing, one should expect the infant to show surprise when the object chooses the incorrect occluder (the opposite result from the standard task as shown in Fig. 6).

Conclusion

            The infant's intentional stance may be derived logically from a primitive stance on causal determinism and evidence of equi-origin of behavior. If so, evolution need not have designed a dualistic view of lawfulness for coping with animate and inanimate objects in the world. Schulz and Gopnik (forthcoming) have recently and independently developed a similar analysis for how young children may detect "hidden causes" in the physical domain.    As in the proposal of this paper, the assumptions that young humans are capable of sensing logical implication, at least unconsciously, and are committed to determinism are viewed as providing cognitive leverage for taking unobservable causes into account.

Notes

1)    It is possible that my "determinist stance" is equivalent to what Dennett meant to convey with respect to his "physical stance", but I am not sure enough to borrow his term. If it is, it should yet be clear that Dennett does not propose that the intentional stance is logically derived from the physical stance.

2)   Historically, there has been disagreement about whether a particular kind of effect should be conceived as having the possibility of more than one kind of cause. William James (as cited by Copi, 1953, p 331) was apparently against this idea of a "plurality of causes" and made the counter claim that "every difference must make a difference". However, it is a more commonly held position that, in many causal sequences, the effect may carry no distinction as to alternative sufficient causes (e.g. the light comes on the same whether I through the switch with my fingers or my thumb.)

3)   The idea that very young humans might respond to logical implication is not new. Piaget (1954/1937) framed the infant's capacity to pass stage IV of object permanence (around 8 months of age) in terms of underlying capacities for representation and deduction. The contemporary research on physical and social knowledge in infants (e.g. solidity, gravity, numerical object relations, implications of desire) that relies on manipulation of expectancy (e.g., use of the habituation/recovery method) is at least implicitly assuming logical processing on the part of the infant. Gopnik and her colleages (Gopnik, Glymour, Sobel, Schulz,  Kushnir, & Danks, 2004) are explicit in the assumption of the young child's sensitivity to logical constraint in their studies of how "causal power" is inferred. Gergely and Csibra and their colleages argue for a view of young infants employing "principle-based" reasoning about rational action when engaged in the teleological or intentional stance (Csibra, Gergely, Biro, Koos, and Brockbank; 1999). When making this assumption of a young child's sensitivity to logical implication, I believe most of the cited theorists would concur with Gopnik et als (2004) explicit proposal that these logical inferences occur at an unconscious level.

References

Armstrong, D. M. (1968).  A materialist theory of the mind. New York: Humanities Press.

Armstrong, D. M. (1969).  Dispositions are causes. (Reply to Roger Squires). Analysis, 30, 23-26.

Armstrong, D. M. (1993). Reply to Martin. In J. Bacon, K. Campbell, and L. Reinhardt, (Eds.) Ontology, causality and mind. New York: Cambridge University Press. Pp. 186-194.

Baron-Cohen, S. (1994). How to build a bsaby tht can read minds: Cognitive mechanisms in mindreading. Cahiers de Psychologie Cognitive/Current Psychology of Cognition, 13, 1-40.

Bartsch, K. and Wellman, H. (1989).  Young children's attribution of action to beliefs and desires. Child Development, 60, 946-964.

Bloom, P. (2004). Descartes' Baby. London: William Heinamann.

Carnap, R. (1938). Logical foundations of the unity of science. In International encyclopedia of unified science (Vol. 1, Part I). Chicago: University of Chicago Press. Pp.42-62.

Copi, I. M. (1953) Introduction to Logic. New York: The MacMillan Company.

Csibra, G., and Gergely, G. (1998). The teleological origins of mentalistic action explanations: A developmental hypothesis. Developmental Science, 1, 255-259.

Csibra, G., Gergely, G., Biro, S., Koos, O., and Brockbank, M. (1999). Goal attribution without agency cues: The perception of 'pure reason' in infancy. Cognition, 72, 237-267.

Dennett, D C. (1971).  Intentional systems. Journal of philosophy, 8, 87-106.

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

Gergely, G. and Csibra, G. (1997). Teleological reasoning in infancy: The infant's naive theory of rational action: A reply to Premack and Premack. Cognition, 63, 227-233.

Gergely, G., Nadasdy, Z., Csibra, G. and Biro, S. (1995). Taking the intentional stance at 12 months of age. Cognition, 56, 165-193.

Gopnik, A. and Astington, J. W. (1988). Children's understanding of representational change and its relation to the understanding of false belief and the appearance-reality distinction Child Development, 59, 26-37.

Gopnik, A. Glymour, C., Sobel, D., Schulz, L., Kushnir, T., and Danks, D. (2004). A theory of causal learning in children: Causal maps and Bayes nets. Psychological Review, 111, 1-30.

Heider, F. (1958). The Psychology of Interpersonal Relations.  New York:  John Wiley & Sons.

Heider, F. and Simmel, S. (1944). An experimental study of apparent behavior. American Journal of Psychology, 57, 243-259.

Hull, C. L. (1951).  Essentials of Behavior. New Haven: Yale University Press.

Kelemen, D. (1999). Beliefs about purpose: On the origins of teleological thought. In Corballis, M. and Lea, S. E. G. (Eds.), The Descent of Mind:Psychological Perspectives on Hominid Evolution. (pp. 278-294).Oxford: Oxford University Press.

Leslie, A. M. (1994). ToMM, ToBy, and agency: Core architecture and domain specificity. In L. Hirschfeld and S. Gelman (Eds.), Mapping the mind: Domain specificity in cognition and culture. New York: Cambridge University Press. Pp. 119-148.

Lewis, M. (1990). The development of intentionality and the role of consciousness. Psychological Inquiry, 1, 231-247.

Onishi, K. H. and Baillargeon, R. (manuscript under review). 15-month-old infants understand false beliefs.

Perner, J.; Leekam, S. R.; and Wimmer, H. (1987). Three-year-olds' difficulty with false belief: The case for a conceptual deficit. British Journal of Developmental Psychology, 5, 125-137.

Piaget, J. (1954/1937). The construction of reality in the child. (M. Cook, Trans.). New York: Basic Books.

Leslie, A. M. (1995) A theory of agency. In D. Sperber, D. Premack, & A. J. Premack (Eds.), Causal Cognition: A Multidisciplinary Debate. Oxford: Clarendon Press. Pp. 121-149.

Premack, D. (1990). The infant¡'s theory of self-propelled objects. Cognition, 36, 1-16.

Premack, D. and Premack, A. J. (1995) Intention as psychological cause.  In D.

Prior, E. (1985). Dispositions. Aberdeen: Aberdeen University Press.

Ryle, G. (1949). The concept of mind. London: Hutchinson.

Sarason, S. B., Lighthall, F. F., Waite, R. R., and Ruebush, B. K. (1960). Anxiety in elementary school children: A report of research. New York: Wiley.

Schulz, L. and Gopnik, A. (Forthcoming) Sight Unseen: Causal Determinism Allows Preschoolers to Learn about Unobserved Causes.

Watson, J. S. (2005) The elementary nature of purposive behavior: Evolving minimal neural structures that display intrinsic intentionality. Evolutionary Psychology, 3, 24-48." Link is at http://human-nature.com/ep/

Watson, J. S. (1995). Mother-infant interaction: Dispositional properties and mutual designs. In N. S. Thompson (Ed.) Perspectives in ethology, Vol. 11, Behavioral design. New York: Plenum Press. Pp. 189-210.

Watson, J. S. (2001). Contingency perception and misperception in infancy: Some potential implications for attachment. Bulletin of the Menninger Clinic,  65, 296-320.