Understanding Simulation¹

Alvin Goldman and I are on the same side of the broad debate between simulation theory and its rivals. I’ve learned a great deal from his rich and thought-provoking book, and we agree about many things. We are both simulation theorists who are attracted to the economy of simulating as opposed to theorizing (178–80), yet allow a role for theorizing within hybrid views of mindreading (43ff).2 We both believe that mindreading should be explained in ways that draw on philosophy, psychology and neuroscience and that are empirically exposed rather than purely a priori (3–5, 9). There are some points of outright disagreement, however, and some issues on which I am not clear about Goldman’s position. This commentary presents an opportunity to explore differences of approach within the simulation theory tent. My remarks are divided into four sections: Two Concepts of Simulation High-Level Versus Low-Level Simulation Mirroring, Control, and Simulation Self Versus Other: Egocentric Bias or Reality Bias? In this section, I will distinguish two ways of understanding simulation—in terms of resemblance and reuse. Although Goldman’s official account of simulation relies on the notion of resemblance, I’ll suggest that the notion of reuse may be implicit in his discussions as well. I’ll argue that the understanding of simulation in terms of reuse is the more fundamental one. First, I’ll give an exposition of Goldman’s official resemblance account of simulation. In the course of this discussion, I’ll point out two other distinctions that are related to the reuse/resemblance issue—a distinction between processes and outcomes of processes and a distinction between resemblance within one person and across persons. Second, I’ll examine Goldman’s ways of contrasting simulation and theory, which suggest an alternative account of simulation in terms of reuse. Finally, I’ll draw out the contrast between reuse and resemblance explicitly, and suggest that reuse is the more fundamental idea and should provide the generic understanding of simulation. Goldman begins by noting some ideas associated with simulation theory: of empathy, putting oneself in another’s place via pretend states, using one’s own psychology to mimic other minds, projecting from the first to the third person (17–20). He explains that a simulator uses processes of the same kind as the target uses in order to read the target’s mind: for example, uses decision-making mechanisms of the kind the target would use in decision-making, rather than factual reasoning mechanisms applied to the target’s mental states, which the target would not use in decision-making (34). He proceeds to distil a generic conception of simulation that involves copying and resemblance, which can be paraphrased as follows: Process P is a simulation of process P’ if, by definition, P duplicates, replicates, or resembles P’ in significant respects, and in doing so fulfils a purpose or function. (37) Mental simulations are simulations of mental processes. Simulation theories of mindreading hold that mindreading aims at simulation and hence at resemblance, not that mindreading always succeeds in producing a resemblance (38). Note that the official generic conception of simulation applies to processes, not the states that are their outcomes. In Goldman’s view of mental simulation, what and where are the entities that resemble one another? Both questions have dual answers. Officially, the resembling entities are processes, but Goldman also often talks of resemblance between the outputs of those processes: mental states and their contents.3 I’ll refer to this as the process/outcome duality in his discussions. For example, Goldman explains that, according to simulation theory (‘ST’), mindreading aims at simulation of, and hence at matching, the target’s mental states, although it may not succeed; pretend states may correspond only minimally to mental states of the target (38). ST’s aim is thus put in terms of matching states rather than matching processes. And we are told that the cognitive mechanisms that operate on pretend input states produce roughly the same sorts of output states as genuine nonimagined counterpart inputs produce (287; see also 283–286 on resemblance between pretend and real mental states). As Goldman recognizes (37), there is also an interpersonal/intrapersonal duality in his discussions. Simulative mindreading aims at resemblance between the processes or states of the mindreader and the target: interpersonal resemblance. But Goldman’s discussion also makes substantial reference to intrapersonal resemblance, for example, between experience of emotion and recognition of emotion by the same person. In another example of intrapersonal resemblance, the visualizing process is regarded as similar to the seeing process, and the content of visualizing is regarded as resembling the content of seeing (38). (Again, note how for visualizing, resemblance is applied both to processes and to the contents of states that are the outputs of processes.) Goldman distinguishes sharing of representations across and within individuals, and comments that sharing within individuals is the case that most interests him, as when pretend states are approximate equivalents of genuine first order states in the same person (211; see also 283–284). So far, I have highlighted two dualities in Goldman’s discussions. First, while the official generic conception of simulation is in terms of resemblance between processes, there is also discussion of resemblance between outputs of processes—mental states. Second, resemblance is applied both within one person, as between pretend and real states, or between emotion experience and emotion recognition, and also intersubjectively, between mindreader and target. Indeed, simulative mindreading appears to involve both applications of resemblance: the successful mindreader uses the same processes in himself both to experience and to recognize emotion, and his emotional processes or states are similar to those of his target. Or again, the successful mindreader generates pretend states in himself that resemble real mental states of his own, and that also resemble mental states of his target. The process/outcome duality and the interpersonal/intrapersonal duality are found throughout the book and contribute to a sense of ambiguity about the fundamental idea of simulation employed. I’ll now suggest that there are in fact two different concepts of simulation in play, which should be distinguished: reuse and resemblance. Both are in play, for example, in the general idea of using one’s own mechanisms on pretend states to mirror or mimic the minds of others, which Goldman regards as ‘[a] fundamental idea of ST’ (20). Ultimately, I’ll argue that the fundamental generic concept of simulation is reuse, not resemblance. The contrast between reuse and resemblance emerges from attention to another contrast, that between simulating and theorizing. My claim will be that reuse rather than resemblance is really the core idea that distinguishes simulating from theorizing. How does Goldman distinguishing simulating from theorizing? I’ll begin to answer this question by explaining how he does not distinguish them. He rejects the ideas that any body of information is a theory (25) or that any appeal to information, information–processing, or computation implies theorizing (34, 150, 175). While theorizing involves applying laws and making inferences from them, Goldman does not rely on a merely negative characterization of simulation as not involving laws in these ways (34). He aims for a positive characterization of simulation. ST typically invokes the use of mental processes by the mindreader, while theorizing invokes factual reasoning or information about mental processes (28–29, 40). This suggests that something like a use/mention distinction is fundamental to distinguishing simulating from theorizing (see also 177, 181). In response to Daniel Dennett’s argument that simulation collapses into theorizing since theorizing is needed to drive simulation, Goldman draws the important distinction between and simulation may a by applying the laws of a theory to a or that is about the target. In doing it does not use or the same of processes as its target but rather the target processes, using another This is not simulation in the sense that with it is by theorizing contrast, a simulating may resemble a target by using processes with the it theory about processes in order to use them. simulating is what with theorizing. For example, Goldman ST that use the same mental in mindreading emotion that use when emotion, theory theory that mindreading emotion information about emotion As of simulation in Goldman other the between for and for and the of in and This is that processes that are for one purpose are for to the reuse conception of simulation, what distinguishes simulating from theorizing is that information about a process is by that process, as opposed to by theorizing about the The of simulation as reuse is processes for emotion are for emotion, processes for seeing are for of processes is from resemblance between mental states, interpersonal mindreader and or intrapersonal real and pretend states in in the same process not produce similar states it is or and similar states can be by different There are two of in Goldman’s discussions of simulation. The implicit reuse conception on intrapersonal reuse of processes while the official resemblance conception on interpersonal resemblance of mental states that are the outputs of processes The reuse/resemblance distinction is by the fact that we may also appeal to intrapersonal resemblance between the outputs of processes as when Goldman that the content of visualizing resembles the content of And we may appeal to interpersonal of processes as when Goldman that mental processes for mindreading are of the same kind as those that produce the mental states of the target. these two are for of simulation. resemblance between a and a would not as simulation processes by that same person for reuse of process, be like simulations of a involve of a theory of to produce a the content of which is similar to that of And interpersonal of processes between mindreader and target does not for simulation intrapersonal reuse of the own mental processes is simulation that interpersonal between mental processes of mindreader and target from intrapersonal reuse of the own mental processes, not in some other the processes not reuse his own processes, are not the mindreader processes of the same kind as his I’ve that the two of simulation and which may to between the of reuse and resemblance, rely on simulation as reuse is a to argue that of simulation involving interpersonal resemblance of states also rely on simulation as reuse points of states between mindreading and target does not for simulation intrapersonal reuse of the own mental processes is The states of the same kind as his target’s is not his processes not reuse his processes. The case for the resemblance involving interpersonal resemblance of states, thus also relies on simulation as reuse. My point in this that reuse rather than resemblance is the core generic sense of simulation. is to distinguishing simulating from theorizing. resemblance, between processes or their output states, intrapersonal or does not for simulation the of intrapersonal reuse of this is implicit in Goldman’s own discussion, and his own concept of simulation as opposed to theory is in terms of reuse rather than resemblance. Goldman distinguishes and simulative I not have a clear view of this however, and would like to and for Goldman’s of simulation involves of or while his of simulation involves the use of to produce pretend states How is simulative mindreading is as and simulative mindreading one or more of the it mental states as of it are to at and and it of to processes are to be an in In simulation, Goldman resemblance or rather than How should we resemblance in it involve or Goldman with when I that I am I merely or while of involves that rather like a or of Goldman does not to rely on resemblance to simulation, but regards and resemblance as more mindreading is and can that is or as as and While Goldman’s distinction with the distinction between and processes, it it does the distinction Goldman’s distinction does not with or he processes can be or while processes are only and to be differences in rather than in I the of the or that is this have to with a distinction between representations with and be that of mindreading can on argue this in but Goldman does not a We have two of and but how a I of in and mirror as the of Indeed, response to Goldman’s distinction to it as a of the But it to see that this Goldman makes appeal to the distinction in does he for to across Goldman’s for example, and understanding are processes but can be at the For these I’ve found Goldman’s distinction and would I’ll now on which Goldman views as a of simulation, and its to I’ve this in and there to be substantial between Goldman’s account of I’ll argue that one of what I as Goldman’s views on between and is his on for emotion recognition rather than for is of or rather than that the fundamental between and this to the case of is a on which the fundamental does not I’ll also argue that there is a between Goldman’s views on this and his views on between ST and Goldman four of simulation for emotion of the are similar in that the seeing another’s of emotion to his a similar only in the between will not Goldman’s other which he and for other emotion, which to which to seeing which to similar which to similar emotion, the like of which mindreader to the target Goldman this the the to his emotion that emotion to simulation would involve to the mental that we as that is is not the of simulation, but the of from mental to There is some for this in the case of Goldman regards as of mental see also for other emotion, which to which to seeing which to similar to that which would the mindreader to have a similar which to the similar emotion, the like of which the mindreader to the target Goldman this since it an similar in the mindreader and via of what it would like the mindreader that for other emotion, which to which to seeing which to similar emotion, the like of which the mindreader to the target the between the seeing the and his a similar emotion is and The mindreader uses his own emotion to recognize the target’s Goldman about by the simulation and as simulation of emotion recognition but a case can be for the for mindreading of and involves a from another’s mental to by the other and from an of that to a similar mental in the The of is the between the another’s and his a mental similar to that which the a resemblance account of simulation, is simulation of the interpersonal resemblance of mental states. a reuse is simulation of the intrapersonal reuse, in another’s as of processes that to The intrapersonal reuse of in that as when one a and one’s this point to and but a is to see also Note the implicit in this account of one’s to seeing one’s own is thus between processes that to and processes that of As a seeing another’s similar can a process that to one’s own may be This between processes for and processes that to is an of should not be with the by what Goldman and rejects on the that processes are in the case of emotional The is the of is between processes for and processes that to while the of is between and states that to another’s as may a similar merely in a does not an to in that I suggest that Goldman’s of the for simulation that from to its is it to with the to I this in the account of Goldman in terms of that mental processes can be to his general about this Goldman on in mindreading of but also recognizes the I and have for which to the of on mirror My is at the is how Goldman’s would to for other which to which to seeing which to similar the like of which the mindreader to the target Goldman’s and are not to the issues that about and Goldman’s conception of we can now between and As Goldman simulation and in a generic sense that he within As I read he as a of simulation in this Goldman comments that this is not clear to him, and that is how the approach would to this case in does not the between and in for this that I the which does out the in a that on and is with Goldman does not account of what in is a generic account of the of and when I I see and it or and thus that is to a and this is as target inputs associated with As may be with the and to output target. This is can be by use processes are not to but to associated can now be to output in order to the target. process that uses in this can be and than one that for Note that a process involves simulation in the sense of intrapersonal reuse, which I is the core generic sense of simulation. This is also sense of simulation when he it to Note also that it is not to simulation in this sense that it be simulation can be to Goldman to ideas from theory to his kind of in emotion He recognition of from a of and to He it to of simulation with the sense of this is Goldman only how to emotion recognition on not how to on he that there is also a case for mindreading for as and provide the in which to not and emotion are rather than or the idea of in order to a to a out of place and So does the idea of simulating in order to contrast, is to as in order to a does Goldman of emotion but not in the between and he the idea that are But this provide a for the of where are may be fundamental to between and not is is the the to simulation in a involves simulation when processes that to are to the or from processes. As I explained is simulation processes that to are to another’s as is the between from one’s own processes and another’s similar as input from one’s own and another’s of the same kind can be similar for or to to distinguishing them. the same can simulative in and in can be in to the of processes and in to the processes that to similar in The is this is the between processes for as and processes that to which I explained is to the of found in the of while the of Both involve simulation in the core generic sense of intrapersonal reuse. The interpersonal of as an of to distinguish between the of one’s own processes own and inputs from similar another’s similar to the distinction between and the interpersonal of can be for cognitive simulation does not an interpersonal simulation in is not is visualizing, which Goldman regards as a of simulation. ST to simulation for it an that many other There is more to the than of the of in and their use in to of ways of with the of to understanding of the of But I to any more of this but rather to the I The I’ve told about the between and applies to or it is a about the simulative processes that This is the by theory, which and or the processes by which are is also the by a body of on of is as a for and provide the most to not and emotion As however, Goldman to the between and in the of emotion and the of So it that he How and are related the of and is a I it can be one their fundamental in and also a view about how and are In for on the more for So in this section, I have Goldman’s views about I have also explained how simulation in the sense of intrapersonal reuse is in both and in the between and the the interpersonal of from intrapersonal And I have explained what I as the of Goldman’s of these one more I to in this section, Goldman’s that simulation in theory is more like simulation or theorizing in than like simulation in ST Goldman begins his on simulation and theory by simulation can be within a as suggests Goldman recognizes in a note that Daniel simulation with a on and that may have this in Goldman comments that from this that the general should be in And he the by that theory can be in in the sense he with ST Goldman thus to from simulation can be explained in a to can be explained in a simulation Goldman suggests that the sense of simulation to is not the sense to simulation for He that simulation for is more like theorizing than simulating in the sense to ST that simulation, of a as not simulations or are using representations or of their Goldman suggests are in the that simulations not simulations in the resemblance sense he uses to He of or to this I will this by that this of Goldman’s is in with his own view of the between ST and theorizing. I that simulation as reuse is the generic sense of simulation that emerges from the contrast between ST and not simulation as resemblance. So I with Goldman’s for simulation for is more like simulation or theorizing in than like simulating in This into account of how the reuse sense of simulation applies both in processes and in processes. I not the that processes or simulations are I to Goldman’s on his own the reuse to his resemblance account of simulation. is that simulation for on theorizing in a that is with simulation in My claim is that this view is in with Goldman’s of the of theorizing and The of theorizing and ST two general in Goldman’s discussions. First, an account of mindreading can be a in which the of simulation and theorizing I agree on this Goldman an example of a hybrid although he does not this account as I the and of simulation for emotion recognition Second, simulation and theory not theorizing simulation although there is a between simulation and theory at the there is between at different There is in that mindreading is at the by simulation, which is in at the by an This suggests an to the a although now theorizing would be doing the The differences between the of simulating and theorizing not be by fact that are (34). ST is not it is by theorizing Simulation be for mindreading theorizing is also at theorizing simulation rather than the positive simulation and theory are ST not or theorizing to its Goldman in fact the hybrid rather than the view of but that the of simulation by theory would not The that is now the of simulation by theory is with should the of simulation by processes, as a kind of be with on information does not ST in approach to ST in this book does not to information out of the see also does theory of simulations as or information the that it is with the ST sense of simulation may from but Goldman the is a of and it for to be in theorizing would not its as simulation also in as And would the of simulation in theorizing its as simulation, any more than it would the of Goldman’s own his that simulation in involves theorizing in a that is with ST does not hold this I’ve Goldman’s account of explained the between and that Goldman it of his on emotional and of and and his that processes are in a with on the that it is in with his claim that theory can simulation with My and the of that Goldman as a He this to an between first person and third person mental and the of first person Simulative mindreading in his view involves from the first to the third person, which makes first person to third person I will suggest another of the in terms of which is with a different view of between the first and third person. of these is is an for to Goldman views from the first to third person as one of the general ideas ST he recognizes that In projecting a pretend of one’s own to one’s real states to be and so that are not to is in or the some to or it does in the would be of one’s own mental states to Goldman that is found in that an as the of and to that the same as The of can operate when one that are is found for and to the first person mental states Goldman the in terms of an like to suggest that a of it is to another in terms of a which would not from the first to the third person. In some of Goldman’s both are in I it is distinguishing The that mental states are and to explains the The that is and to explains the to and as But to to something in the one does not to a mental to oneself of of mental states would also be of how these would to in for In an it be the in order to a to it is not to that the any to the the in order to a to the As Goldman in and In the a as a an in one of two the see The also a and that the can see which of the two the is The and while is the the The and points to a The should that this will be the as does not that the have So the should that the is in the other This is a the does not which the In the is the same that the does see which the is in by the the but the one the does not of a about where the while the But on the other the does not of a but rather of a while the does while a distinction in mental states. that of does not of these distinguish between and of and to them, some who these may a distinction between and mental states. may be for rather than the first may to reference rather than may distinguish these The is and of as those of and not the of first person third person. For mirror information that is to distinctions both and and is in and My account in this information that does not distinguish the first and third person. This distinction only with the to and to that one is is not of a first person, but is of what the distinction between and to be on information in which and other are not is not that these of ST would an are with a My comments have into four I’ve Goldman’s resemblance account of simulation from a reuse account and that the is the core generic I’ve the distinction between and simulation for I’ve Goldman’s view of the simulation, and and put a different And I’ve from While I’ve on many points with this the fact that we are on the same side of the broad I’ve found it to on his book in to