|
Simulation: Clarifying Some Confusions
July 23, 2003
The simulation theory is a direct challenge to the dominant view in folk psychology that our ability to explain and predict intentional behavior involves a tacit knowledge of a theory of mental states and their causal role. Instead of invoking a theory of intentional behavior to explain and predict what other people do, some thinkers such as Robert Gordon and Alvin Goldman(1) propose that what we do is simulate others when we try to describe, explain, and predict their actions. While there are many who claim that theory-theory and simulation theory are incompatible, others (Davies, Barnes and Thagard)(2) claim that the two are not mutually exclusive and, in fact, simulation and theory-theory may fit together necessarily in our folk psychological explanations and predictions. While I believe there is some merit to this line of thought, I think it is founded on a faulty characterization of what we do when we simulate others (at least the type of simulation put forth by Gordon).
My aim is to clarify some confusion regarding what simulation is and in the process of doing so dispel the complaints involving normativity and isomorphism regarding the move from the first-person to the third-person in simulation which threaten to collapse simulation into theory-theory. In my view, much of the problem stems from a faulty interpretation of Gordon’s example of a chess player who "tries to make adjustments for relevant differences"(3) in simulating his or her opponent.
In the third section, I will respond to Barnes and Thagard’s complaint that there has been a misappropriation of what simulation really is, and that what is really going on when we simulate is a type of empathetic analogy between the simulator and the simulated agent. I agree with their distinction between sympathy and empathy, but believe that they put forth a type of simulation through sympathy and not empathy as they had intended.
1: Overview of Theory-Theory and Simulation-Theory
First it is necessary to give a brief account of both the theory-theory and the simulation theory of intentional states such as beliefs and desires. As Peter Carruthers writes, the theory-theorist believes "that our understanding of mentalistic notions . . . is largely given by the positions these notions occupy within a folk psychological theory of the structure and functioning of the mind." (4)Thus, when we make predictions about our behavior and the behavior of others we are relying on some tacit knowledge of a theory of commonsense psychology.
Simulation theorists, on the other hand, believe that there is no such theory that is operative when we make predictions. Rather, we simulate or "put oneself in the other’s shoes" when we predict the behavior of the other. There are two main strands of the simulation alternative although they have significant overlap.
The first is the hypothetico-practical reasoning simulation asserted by Gordon. For Gordon, self-ascriptions of mental states are not inferential, rather they are "declarations of immediate intention" which take the form "I shall now X"(5). One does not introspect about inner intentional states, but is engaged in practical reasoning. When one predicts the behavior of another person one engages in hypothetico-practical reasoning in simulation. That is, one enters into an imaginative "pretend-play" as one tries to "track"(6) the other. The first thing one might do would be to exchange spatiotemporal perspectives, that is, to imagine oneself "over there" as that person, seeing what he or she sees from that perspective. One tries to enter into the other’s situation or context. One good example that Gordon himself uses is a chess match. It is advantageous for one to be able to gauge the game from the other person’s perspective. Not only does one attempt to see the board the way the other does, but it is also helpful to "adjust for the relevant differences" between one and the other. These differences may include (if you know something about the other player) differences in skill level, and certain tendencies and idiosyncracies the other player has. What one essentially does in simulation is shift into the other’s context to "see" what happens next.
The other main strand is represented by Goldman. For Goldman, simulation depends on the simulator’s ability for introspection and an appeal to the internal processes of intentional acts. A person must employ a simulation heuristic into which the person feeds pretend "initial states of the targeted agent" to "see what further states the mechanism produces"(7). This implies an underlying process determining intentional states and behaviors that is appealed to and that one seemingly assumes is operative in the simulated person. This underlying structure is used to derive predictions about the intentional behavior of others because it is assumed that others have this same structure that has become apparent to oneself through introspection.
2: Objections
Some have expressed doubts about the ability of simulation to have true predictive value. Angela Arkway believes that the problem with simulation is that the simulator must be able to correctly simulate "the ranking of value judgments attached to the behaver’s desires, emotions, and intentions"(8) and that ability depends on an enormous amount of similarity between the simulator and the agent simulated. She believes that it is not likely that there is such similarity so that one can accurately explain why someone did what he or she did because one would have to know so much. Instead, a better way to explain and predict behavior, according to Arkway, would be to pick out the most likely candidate from a range of intentional states that would serve a causal role in behavior.
While some have argued that simulation "should not be taken seriously"(9) or that Gordon is a "quasi-behaviorist" while Goldman is mired in "Cartesianism"(10), many thinkers have indeed felt the pull from simulation yet seek to incorporate it into a broader theory-theory picture of folk psychology. Davies is quick to point out that Goldman’s version of simulation falls prey to a collapse into theory-theory because of the notion that there is a "matching of structure between a causal process and a derivational process that is used on some accounts of what it is for a cognitive processing system to embody tacit knowledge"(11) and that simulation rests on this matching between the simulator and the simulated person. The same structure present in both individuals is tacitly implied by the conclusions made from the simulation process. When one projects one’s own derivational process through onto the other, one is merely taking the introspected knowledge of oneself and applying the same structure onto the other, that is that explanations and predictions are to beliefs and desires what conclusions are to premises. Davies and Stone adopt a tacit knowledge theory of folk psychology which states that the simulation process is essential to the theory-theory itself(12).
Richard Moran argues for "interpretation theory"(13); and denies that simulation escapes assumptions of normativity and rationalization of the simulated agent because without these assumptions there is no way of understanding others if we engage in simulation. He claims that to construe simulation any other way is to misunderstand it. However, it seems that Moran misunderstands simulation when he writes that "projecting one’s own likely beliefs onto another person is normally equivalent to interpreting him as rational, because it is a logical truth that a person takes his own beliefs to be true"(14). Simulation, however, does not necessarily involve "projecting one’s likely beliefs onto another". That would be like the chess player who only sees the board from the other person’s point of view instead of also taking into account the relevant differences of the other player. It is not that one projects his or her beliefs onto the other player, instead he or she tries to get inside the other’s head to predict what the other will do. To project one’s beliefs onto the other without taking into account the other’s relevant differences in chess playing could be disastrous and wholly counter-productive to the goal of winning the chess match. To do such projection that Moran describes that is, to say "Let’s see. What would I do if I were her?" is in fact normative and may involve and isomorphic process, but that is not what one does in simulating the other. Instead one says something like "Let’s see. She tends to bring the queen out early, and is willing to sacrifice some pawns in the center. This situation may call for that, so, if I pretend to really be her, what will "I" do?" This is still a question about what she will do, not what I will do. Simulation is not about moving oneself into the other’s situation. Rather one pretends that one actually is the other. Moran’s characterization is devoid of any attempt to really get into the other person’s head, and "become the other", instead, on his view of simulation one just projects one’s own thought processes onto the other.
3: Simulation as Empathy as Analogy
This last section deals with a final objection to simulation as an independent account of folk psychology. Not only do Barnes and Thagard argue that some amount of theory may be required in simulation, but they also claim that their account of the "analogical processes of empathy and their constraints"(15) can show just how much theory is required to simulate/empathize in particular situations. Furthermore, they claim, like I do, that Goldman’s description of simulation is lacking, but it is interesting that Gordon’s version of simulation is not mentioned except in an endnote referring to Gordon’s view that simulation theory and theory-theory are mutually exclusive. In the notes they also refer to simulation theory as having "slightly different versions" of simulation theory(16), so it is apparent that, although Barnes and Thagard do not deal directly with Gordon’s "slightly different" version of simulation they believe that their paper addresses Gordon’s version as well(17).
Barnes and Thagard make a clear distinction between sympathy and empathy and intend to employ empathy in their version of a new and improved simulation, but, in fact, they seem to confuse sympathy and empathy shortly after distinguishing the two, and end up giving a simulation-as-sympathy version which is an inaccurate portrayal of what we do when we simulate others. They note that the prefix em-means "in" or "within", while sym-means "along with" or "together"(18). To empathize with the other is to know the other by putting oneself into his or her position as best one can. To sympathize is relate to the other and try to share an experience by going along with, or trying to be together with the other in his or her state of mind.
Barnes and Thagard characterize empathy (and thus simulation) as an analogous process as follows (first to determine emotional state):
"The person P is in situation S. which is like your situation S*. P has goals G, which are like your goals G*. When you faced situation S* which affected your goals G*, you felt emotion E* (S* and G* caused E*). So P is probably feeling emotion E, which is like your E*, caused by S and G."(19)
To predict resulting action:
"The person P has emotion E in situation S. When you were in a similar situation S*, with a similar emotion E*, you did action A* (S* and E* caused you to do A*). Hence P’s emotion E may lead to similar action A."(20)
Barnes and Thagard then tell us that empathy can easily fail because one may not be able to find situations, emotions, or goals that are similar to the simulated agent. That is, one cannot find the analogous conditions in oneself that would go along with, or run parallel to, the other’s conditions. The varying degrees of similarity to the other’s conditions as such determine whether or not one is able to simulate automatically, on one end of the spectrum, or more deliberately through rule-based reasoning on the other end. This rule-based reasoning entails using general normative knowledge about people and is necessarily theoretical.
Barnes and Thagard maintain that to empathize and know the other, one must first appeal to one’s own experience to find analogous similarity. This seems like sympathy to me, not empathy, in that one calls on one’s own experience to try to mimic and process together along side of the other’s processing. On their view of simulation, one tries to share in another’s experience, and the only way one can is to appeal to one’s own experience. On the other hand, simulation or empathy means being in the other’s shoes and "seeing the world through the other’s eyes"(21). This is not what Barnes and Thagard suggest we are doing when we simulate. What they propose as empathy seems like very little more than a return to the self, not a "movement" towards the other. Simulation means that we try to know as best we can, or as much as is necessary to predict what the other will do, the other’s internal and external conditions by actually imagining ourselves as the other.
Gordon himself addresses the issue of analogy and straightforwardly denies that simulation involves such analogy. He says that the "old argument from analogy" is what some writers, both in favor and not in favor of simulation, "take simulation to involve an implicit inference from oneself to others", and this is why he wishes to "emphasize transformation"(22). Gordon tells us that we should think about actors playing roles and how an actor will transform oneself into the character in order to play the character. This is what simulation is similar to. One gives a characterization of the other based on what one knows about the other and then changes into the other during pretending. We as simulators engage in pretend play where we imagine ourselves as other and then play out the role.
Conclusion:
An unfortunate result of our everyday life as simulators is that one often does not really learn as much as possible about the other before making judgments about the person. Stereotyping is everywhere, and this stereotyping is a "sizing up" of the other and assuming certain things based on the other’s appearance and imagining what that person "must" be like. It is hard for me to believe that in such cases one predicts the other’s behavior based on what the other believes and desires from premises to conclusion. This claim may turn on some controversial assumptions about the speed of mental processes however, that is, how quick could one follow from premises to conclusion about the other’s behavior? Goldman’s introspectionist account of simulation as an inner awareness of functional processes that are matched with the simulated person’s processes would require more time than Gordon’s view of simulation, and thus falls prey to the belief-desire-action derivation. Gordon’s version of simulation, although requiring one to determine relevant differences, seems to happen much quicker, at times before one realizes that he or she is thinking as if one were in the other’s shoes.
What simulation involves has been a matter of confusion throughout this debate regarding mental states and intentional behavior. There seems to be a general lack of understanding of simulation and that has proved detrimental to simulationists like Gordon who stand by simulation. Goldman’s view of simulation does seem to collapse into theory-theory as it involves processing that runs from the simulator in an isomorphic manner to the simulated. While there are obviously more than two strands of simulation accounts, I recommend that we adopt Gordon’s view as the correct view of simulation and of folk psychology as a whole when it comes to explaining and predicting the behavior of others. I do not agree with Barnes and Thagard that there are "slightly different versions" of simulation theory, at least with respect to Gordon and Goldman. There are key differences regarding the way one simulates in both versions.
A lot of work still needs to be done to sketch a complete picture of simulation to convince at least some of the doubters. Such work will draw on studies of historical interpretation, empathy, and pretend-play capacities (autism and schizophrenia)(23). In general, the mischaracterization regarding simulation centers around the question of isomorphic functioning and normativity assumptions. I believe that Gordon’s view draws us away from such mischaracterization. Simulation is an intriguing---and I think potentially very fruitful---alternative to the dominant theory-theory view of folk psychology.(24)
References
Arkway, Angela J. (1999) "The Simulation Theory and Explanations That Make Sense of Behavior". Paper presented at the 20th World Congress of Philosophy, August 1998.
Barnes, Allison and Thagard, Paul (1998) "Empathy and Analogy" in Dialogue: Canadian Philosophical Review, 37 pp. 705-720
Carruthers, Peter (1996) "Simulation and Self-knowledge: A Defense of Theory-Theory" in Carruthers, Peter and Smith, Peter K. (Ed.) Theories of Theories of Mind chapter 3, pp. 22-38 (Cambridge: Cambridge University Press).
Davies, Martin (1994) "The Mental Simulation Debate" in Lycan, William (Ed.) Mind and Cognition p.414-428 (Malden, MA: Blackwell)
Davies, Martin and Stone, Tony (1998) "Folk Psychology and Mental Simulation" in A. O’ Hear (Ed.) Contemporary Issues in the Philosophy of Mind: Royal Institute of Philosophy Supplement 42 pp. 53-82 (Cambridge: Cambridge University Press)
Goldman, Alvin I. "Simulation and Interpersonal Utility" in Ethics 105 (July 1995) pp. 709-726 (Chicago: University of Chicago Press)
Gordon, Robert M. (1986) "Folk Psychology as Simulation" in Lycan, William (Ed.) Mind and Cognition pp. 405-413 (Malden, MA: Blackwell)
--------(1995) "Sympathy, Simulation, and the Impartial Spectator" in Ethics 105 (July 1995) pp 727-742 (Chicago: University of Chicago Press)
Moran, Richard "Interpretation Theory and the First Person" in The Philosophical Quarterly Vol. 44 No. 175 (April 1994) pp. 154-173 (Malden, MA: Blackwell)
Stich, Stephen and Nichols, Shaun (1992) "Folk Psychology: Simulation or Tacit Theory?" in Mind and Language Vol. 7 No. 1 pp. 35-71 (Cambridge, MA: Blackwell)