From Aaron Sloman Wed Aug 12 10:01:19 BST 1998 To: PSYCHE-B@LISTSERV.UH.EDU Subject: Re: Emotion, attention and consciousness (and global villages) Cc: lnielsen@u.arizona.edu,drwatt@classic.msn.com On: Fri, 17 Jul 1998, Lis Nielsen reminded us of the importance of emotion, writing: > ... > Despite early attempts by Wundt, Freud, and James to integrate the study > of emotion and cognition, subsequent research programs in these areas have > taken quite divergent paths. Recent trends suggest that their streams are > again converging. I agree with Lis about the importance linking these topics, but it's worth noting that the need to integrate studies of emotion (or more generally *affect*, including motivation, pleasure, pain, moods, attitudes, values, ideals, standards...) with cognition has been evident to many people for some time, though I agree that recognition of the need has recently become more widespread among certain scientists. [Digression: It's interesting that despite all the recent talk of a "global village" and the apparently wide-spread belief that ever improving communication technology will cause more communication, there remain, and probably always will be, various groups of people who unwittingly pursue the same questions in ignorance of one another, sometimes re-inventing each other's wheels, and often using different terminology -- which can defeat literature searches. I guess that is inevitable, partly because any one busy person can cope with only N regular contacts, where N is at most a few hundred, and since a fairly high proportion of communication is two-way there will normally be clusters or cliques who know about one another's work and remain ignorant of other things going on. (I am talking about active researchers, not the much larger numbers who buy popular science books, etc., where communication is one-way.) Once upon a time such a cluster consisted of people in the same village, separated from others by rivers, mountains, sheer distance, etc., whereas now such an intellectually inbred "village" can be spread all over the globe. Thus we have many little villages, all global, but not talking to one another (much). End Digression.] I wonder how many people on this email list know that there has been a journal called Cognition and Emotion since the mid 80s, and that ISRE (International Society for Research on Emotions) holds meetings regularly discussing relations between cognition and emotion. (I used to subscribe to Cognition and Emotion, but cancelled my subscription recently because the price went up while the contents became more and more driven by blind empiricism.... a common effect of taking a degree in psychology, it seems.) Are many people on this list aware of Herbert Simon's 31 year old paper `Motivational and Emotional Controls of Cognition' reprinted in his collection: Models of Thought, Yale University Press, (1979). It's based on his 1963 William James lectures, and was originally in Psych. Review vol 74. In it he responds to criticism (in an article in Science 1963), by Ulric Neisser, a cognitive psychologist, that information processing theories cannot account for motivation and emotion. Although I don't agree with everything in Simon's paper (in particular his tendency to ascribe too much to a central serial process) I found his paper very stimulating when I first read it about 18 years ago, and much of my own work on architectural underpinnings of emotions, is largely a development of Simon's ideas. There are various papers on this in the Birmingham Cognition and Affect FTP repository: ftp://ftp.cs.bham.ac.uk/pub/groups/cog_affect/0-INDEX.html (I can now provide PDF versions of our postcript files, on request). I've recently criticised some of Damasio's discussion of secondary emotions on the basis of this framework. (More about that in a later message.) Do people know that for nearly 20 years there has been work in AI attempting to link cognition and emotion, e.g. people who were at Yale around 1980 including Bob Abelson, Mike Dyer, Wendy Lehnert. With a student, Monica Croucher, I published a paper called "Why robots will have emotions" in proceedings of the International Joint Conference on AI in 1981, which also included other stuff on emotions e.g. from the Yale group. Our 1981 paper is in the above ftp directory. Marvin Minsky has also been attempting to include emotions, aesthetic experiences, and humor in his theories of mind for some time. See http://www.ai.mit.edu/people/minsky/minsky.html I like this quote from one of his papers from around 1981: The old distinctions among emotion, reason, and aesthetics are like the earth, air, and fire of an ancient alchemy. We will need much better concepts than these for a working psychic chemistry. Oatley and Johnson Laird, whose recent work is mentioned by Lis have been working on this (separately and together) at least since the early 80s. (I think their global interrupt signal theory handles only a subset of the cases discussed by Simon. I haven't read their recent work.) Andrew Ortony, along with G. Clore, and A Collins wrote a book: The Cognitive Structure of the Emotions, New York, Cambridge University Press (1988) which has been the basis for a lot of subsequent work, including computational models. (I think their taxonomy is primarily a taxonoy of affective attitudes, not emotions, but that's another matter.) Interestingly, there has been a rapid recent growth of interest in emotion among AI researchers. E.g. At the 1998 conference of the American Association for Artificial Intelligence in Madison a couple of weeks ago there was a session on motivation and emotion, including these papers: A Motivational System for Regulating Human-Robot Interaction Cynthia Breazeal (Ferrell), Massachusetts Institute of Technology Emotion Model for Life-Like Agent and Its Evaluation Hirohide Ushida, Yuji Hirayama and Hiroshi Nakajima, OMRON Corporation When Robots Weep: Emotional Memories and Decision-Making Juan D. Velásquez, MIT Artificial Intelligence Laboratory with abstracts at http://www.aaai.org/Press/Proceedings/AAAI/1998/Abstracts/A345.html http://www.aaai.org/Press/Proceedings/AAAI/1998/Abstracts/A162.html http://www.aaai.org/Press/Proceedings/AAAI/1998/Abstracts/A505.html There will also be a workshop on emotions at the AAAI Fall symposium in Florida this year. See http://arti.vub.ac.be/~lola/ei-fs98.html Alas I can't go! The latest ACM (Association of Computing Machinery) SIGART Bulletin (vol 9 no 1) has an article on emotions in synthetic agents by Clark Elliot (at DePaul University). A group led by Joe Bates, previously at Carnegie Mellon University has been developing emotional synthetic agents for entertaintment purposes. http://www.cs.cmu.edu/afs/cs/project/oz/web/staff/scott.html http://www.cs.cmu.edu/afs/cs/project/oz/web They have now left to form a company, I believe. I expect a lot of this sort of work will be going on for Disney, Sony (who employ one of my former students) and others concerned with entertainment and computer games. Cyberlife produced the "Creatures" system which includes norns, whose biologically inspired behaviour evokes amazingly strong emotional reactions in humans! See http://www.cyberlife.co.uk http://www.cyberlife.co.uk/cyberlife/pr/pr_29JUN98_milprod.html In 1997 MIT press plublished a book, Affective Computing, by Rosalind Picard. My draft review can be found (in postscript) at ftp://ftp.cs.bham.ac.uk/pub/groups/cog_affect/Sloman.picard.review.ps (She will have a chance to write a reply.) Is the design of such computer models relevant to studies of mind and brain? Yes if it helps to clarify some of the questions to be asked about brains, and provides a richer conceptual framework for building theories about the information processing architectures that may be found in brains. (COMPARE: birds and (some) aeroplanes are subject to the same principles of aerodynamics. A bird's wing combines the functions of a plane's wing and its propellor. A propellor is just a wing that rotates instead of oscillating. A brain is just a computer that ..... Of course, many people have too narrow a conception of "computer" as a sequential bit manipulator, or blind slavish instruction follower, or, ....) Of course, a lot of the work on computer models of emotion will be naive, partly because so few people have degrees in five or more disciplines required to write effectively about mind, brain, cognition, emotions, development, evolution, etc. Moreover, the pressure to publish, to get PhDs, to get jobs, etc. makes it hard to find the time to straddle disciplines even for people who have the ability and the motivation. One of the common symptoms of naivety is blurring boundaries between different concepts (e.g. motivation, emotion, mood, attitude, etc.) Another is not relating such distinctions to the information processing architectures which underpin them (and the more subtle distinctions which will eventually replace these ill-defined common sense notions, as implied in the quote from Minsky above.) Yet another symptom is excessive concentration on a small set of cases, e.g. emotions in adult humans, or in rats with their brains opened up. (This is also true of many studies of consciousness: they focus only on adult humans, usually academics contemplating themselves in their offices, or performing laboratory experiments, as opposed to climbing a sheer rock face, or emerging from the womb, or suffering Altzheimers disease, to say nothing of bonobos and bats and angry ants.) A theory of emotions, or any other aspect of mind should consider architectural differences between different animals, or between humans at different stages of development or with different forms of brain damage or degeneration. Such differences in information processing architectures entail a need for different affective and cognitive ontologies in describing their capabilities. (Can an ant be infatuated, or humiliated? If not, why not?) (The pains, pleasures, fears and obsessions of a ten year old child are mostly impossible for a newborn infant with a very different architecture, or for a rat... A mature adult human is different again. Culture may also affect architecture. Doug Watt points out that major forms of brain development continue during the first few years after birth. Additional architectural development may be more like "software" changes. Then there's physical degeneration of various kinds in old age. We need to understand all these architectural changes.) For all these reasons I strongly agree with the main point Lis makes about evolution: > Notable in this body of work is a concern to mount > evolutionarily plausible arguments, to explain emotional mechanisms in a > way that makes sense with regard to what we know (or can reasonably > conjecture) about the nature of the adaptive problems which emotional > processes were designed to address. I.e. comparative studies should also take account of evolutionary trajectories which can give insights into existing architectures. However, for reasons which I'll elaborate another time it may be a mistake to assume that all the main types emotional processes are adaptive in themselves. It could be that many of them, especially sophisticated emotions in adult humans, are essentially emergent side-effects of architectures which developed to support some of the control processes described by Simon, which are not all intrinsically emotional. This is a key mistake in Damasio's and Picard's books I think. Compare addictions. I talked about the relationship between different sorts of information processing architectures and different types of emotions at the Conference at Elsinore last August and again at Tucson III, though probably not clearly enough. (I got the impression a lot of people were baffled both times.) The ideas are still evolving, and I now want to distinguish three major categories of emotions, i.e. those called primary and secondary emotions by Damasio and a third category of "tertiary emotions" (previously called "perturbances" because they involve partial loss of control of thought processes). The last group are characteristically human (though possibly present also in some other primates -- I don't know). Primary emotions relate to evolutionarily very old reactive architectures shared with many other animals, even insects. All they require is a reactive architecture with a global "alarm" system. (Brain stem, limbic system?) Secondary emotions (including apprehension, relief, excited anticipation(?)) are possible only in organisms which also have deliberative mechanisms, which developed much later and are probably far less widespread (e.g. probably not in insects, and maybe not in molluscs or fish???) (Simon has much to say about this.) Tertiary emotions (e.g. infatuation, humiliation, guilt, debilitating grief) require a third architectural layer, which evolved even later and is probabably very much rarer, and I suspect doesn't exist in new-born humans. We've called this a "meta-management" layer, which provides mechanisms for monitoring, evaluating, and controlling *internal* processes, e.g. controlling attention, controlling thinking style. Tertiary emotions involve (partial, dispositional) loss of that control. [Further architecture-based sub-divisions among types of emotions are possible.] The different architectural layers operate in parallel and more or less independently: there's no simple control hierarchy. The idea of meta-management clearly overlaps with Baars' global workspace notion, and other similar notions, though there are differences of detail. (Is a grieving dog, separated from its owner, like a grieving human? I suspect there are important differences because of the differences in architectures.) A lot more needs to be said, but that's enough for now. More later. Incidentally I'll be talking about some of this at the SMC98 conference in SanDiego in October, in a symposium organised by Stan Franklin. ( http://www.engr.rutgers.edu/~smc98 ) If any of you are going to be there I hope we can meet and talk. Aaron === Aaron Sloman, ( http://www.cs.bham.ac.uk/~axs/ ) School of Computer Science, The University of Birmingham, B15 2TT, UK EMAIL A.Sloman@cs.bham.ac.uk Phone: +44-121-414-4775 (Sec 3711) Fax: +44-121-414-4281