MARVIN MINSKY, THE EMOTION MACHINE ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ (Simon & Schuster, Nov. 2006) [Comments in square brackets are my own thoughts.] [The slogan of the book might well be, "People are resourceful because they are full of resources!", where we are asked to interpret "resources" as functional units. Or, putting it a little differently, the key to understanding mind lies not in principles, but in functional parts. This is very much the theme already developed in Minsky's earlier Society of Mind, but here he tries to provide a more unified account, and some more specific proposals about what the "resources" (formerly "agents") are.] 0. Introduction ~~~~~~~~~~~~~~~~ He introduces the idea that our brains contain hundreds of resources, which he draws as little nuggets within a brain-shape -- though he doesn't want to associate these with particular brain regions (he says some may be localized and some widely distributed). He suggests that what we do when we confront different types of situations is to switch some of these resources on, while suppressing others. We should think of these resources as functional units, much like program modules. What is not clear, at least at this point, is whether he regards these resources as evolved (and thus innate) mental organs, analogous to bodily organs such as the kidney, heart, digestive system, skin, etc., or learned (how?), or both. The switching on and off of various resources, he says, applies not only to different kinds of problem-solving, but also to emotions. For example, anger involves switching off methods for more prudent planning and acting, while arousing resources that allow us to act with greater strength and speed. Love, he says, involves switching off some critical abilities, and switches on a new set of priorities. Various emotions may use some of the same resources, so we can be in conflicted states, where we feel "mixed emotions". The body itself can also be thought of as providing resources to the brain (a view somewhat different from Damasio's, for example). 1. Falling in Love ~~~~~~~~~~~~~~~~~~~ He first talks about a common attitude towards love (and other emotions), which is that it ought to remain a topic for poetry, but not for analysis. He points out that someone in love typically expresses what he or she feels in terms that suggest incomprehension and loss of critical faculties and all sense of proportion. This is consistent with his view that certain faculties have been suppressed. He goes on to distinguish infatuation from parent-offspring love, friendship bonds, love of country, etc., introducing his notion of "suitcase concepts" -- words or phrases we employ commonly but that cover a whole range of different phenomena; we need to "unpack the suitcase", if we want to understand what the terms mean, in a functional sense. He then elaborates on his contention that seeking understanding of mind by looking for simple principles is futile, even though this has served other sciences (like physics) so well. Instead we need to look for the functional parts -- it is in this way that we can conquer the complexity of our minds. He points out that the seeming simplicty and effortless functioning of our most basic capabilities -- looking at an object, determining what visual regions correspond to that object, classifying the object in light of our prior experience, etc. -- are all very complex. They only seem simple because we have so little access to the "resources" doing the work. He goes on to talk about " the self", making some similar points as we've seen before (rejecting homunculus-like views). But what he wants to claim is that the notion of self is another "suitcase concept" -- we have many different conceptions of self that we use for different purposes. [Here he tends to go in a direction that I've criticized in other writings we've seen -- he seems not to allow for the (to me perfectly natural) view that in talking about oneself (or "one's self!) one is talking about the entire person: body, brain, thoughts, etc., not about some "inner" self. This is what gives the self-concept unity -- a unity that Minsky wants to deny.] He then proceeds to outline the behavioral model of animal and human functioning that has been so dominant at MIT (e.g., in the work of Brooks and Breazeal). He first says that a lot of animal behavior can be explained in terms of collections of "IF then DO" rules, such as "IF hungry, then SEEK FOOD". (He refers to Nikolaas' Tinbergen's The Study of Instinct, and he suggests that many aspects of emotion (more particularly, emotional behavior) can be understood in the same sort of way. Coming back to love, he again suggests that not only are certain faculties suppressed (critical ones, in particular), but previous goals are supplanted by new ones. [But isn't that something quite different from resource-switching? Goals are often dynamically generated things, not existing "resources". Also something missing here (perhaps discussed later) is any mention of the *attitudes* involved in many emotions; e.g., consider pride, admiration, envy, fear of ..., indignation, joyous anticipation, etc. -- these are in part conceptual/ propositional.] He then suggests that we get greater flexibility of behavior, and better control of ourselves in situations that arouse our emotions, through higher-level behavioral layers, starting with ways of recognizing problem types and selecting corresponding ways to *think* about them (not necessarily immediately act on them). This amounts to *mental* perception and *mental* action selection, so it is a level above physical perception and action selection. [It's unclear here whether he is talking primarily about a mechanism for switching to another method when one type of problem we are considering is replaced by another, or about switching to another method when we get stuck on a particular problem. The former is certainly something easy to agree with; surely we use different methods for, say, playing chess than, say, writing a computer program. But the idea that we are constantly trying different methods on a given problem is more questionable. Is Joe Citizen really good at this, or is he talking about Nobel laureates, or his own introspections?] He then outdoes other "layers-of-mind" theorists like Ortony et al., or Sloman, by sketching a 6-layer theory [with my tentative annotation]: Self-conscious emotions [e.g., guilt or pride? Via "imprimers"] Self-reflective thinking [e.g., about one's own methods/mistakes/ abilities/learning process/...] Reflective thinking [e.g., about choice of methods] Deliberative Thinking [e.g., about problems] Learned reactions Instinctive reactions The lowest level is dependent on the "instinctive behavioral system", and the highest level on "values, censors, and ideals". He also talks about arousal of cascades of resources (sec. 1.7). In 1.8 he asks a question that you might think will force him to address the issue of qualia (phenomenal consciousness): "Citizen: But machines can't feel or imagine things. So, even if we could make them think, would not they still be missing the sense of experience that gives meaning to our human lives?" However, he simply isn't interested in such (apparently) philosophical issues, only in behavior, and he suggests that we just don't know "how feelings work", and that's why our machines don't measure up to us in feelings. Just as with diagnosing what's going on inside a car in terms of the parts, you need to know what's going on inside a brain, in terms of functional parts/ programs, if you want to understand feelings. Then he comes back to self, and says that besides a simple model that divides "the self" into a body-portion and a mind-portion, he'll suggest later that we use a network of different models of ourselves. [I think we model various *aspects* of ourselves, but have a single symbol for ourselves -- in that sense the model is unitary (except of course in dissociative personalities, etc.)] He concludes by promising to explain how we come by our values, ideals, and goals through feedback from parents, friends, and other members of society. [Is there room here for moral qualia? I.e., innate values, e,g., certain altruistic dispositions?]