This comment on claims that what we are conscious of is linked to what we can report was provoked by discussion (mainly led by Bernard Baars) on the psyche-b email forum. The full discussion (and many other things) can be read in the February 2002 section of: http://listserv.uh.edu/archives/psyche-b.html REPLY to BAARS From Aaron Sloman Wed Feb 6 20:05:09 GMT 2002 To: PSYCHE-B AT LISTSERV.UH.EDU Subject: Why should *all* conscious phenomena be reportable? I have not been able to keep up with all this discussion. Apologies if my points below have already been made. From: Bernard Baars Date: Sun, 03 Feb 2002 13:09:59 -0500 > There is a stage in scientific development when evidence becomes clear but > theory, the deep level of analysis, is not yet able to explain the evidence. Sometimes there is a prior problem: the *concepts* used to describe the evidence, to ask the questions, and to formulate theories, are so confused that the relationships between evidence and theory are too obscure to be usefully discussed without prior deep conceptual analysis and clarification. Often conceptual analysis on its own cannot help because the whole ontology currently in use is inappropriate (earth, air, fire, water, considered as elements, or the four "humours" thought to explain biological phenomena, etc. http://www.bbc.co.uk/education/medicine/nonint/prehist/dt/prdtid2.shtml ) Sometimes an inadequate ontology can be gradually developed and clarified through the construction of new, initially speculative, theories (e.g. the development of the notion of "atom", and separating out force, momentum, velocity, acceleration, various kinds of energy). Sometimes ill-defined or undefined terms have to be introduced and used in theory construction (e.g. "gravity", "atom", "electron", "neutrino", "gene", "extended phenotype") and these then gradually become clarified through a growing web of theoretical relationships, usually not directly testable, and never provable. [[Chapter 4 of The Computer Revolution in Philosophy (now online) is a 25 year old first draft attempt to explain what conceptual analysis is: http://www.cs.bham.ac.uk/research/cogaff/crp/chap4.html ]] > I think we are (finally) in a data gathering state of science regarding > consciousness. The journals show it every week. What we do not have is > confirmed and settled explanations. Nor agreement on what *sort* of explanation would be adequate: what form it could take, how it would fit in with the rest of science, whether and how it might be testable, etc. Consider the disagreements regarding which animals have consciousness, or whether consciousness is a matter of degree or something that is either present or absent. E.g. some people think that a computer system modelling in fine detail everything we know about a typical person's mental states, events and processes would in some sense explain how minds work. Others think it would be completely irrelevant. Others think it is impossible in principle. And they object to the model for different reasons: that it uses different infrastructure from brains, that it was produced artificially rather than naturally by evolution, that it is not embodied in a similar body similarly situated, that computer models do not explain, they merely simulate, etc. etc. All this is evidence of deep conceptual confusion: not an advanced science. We still lack clear concepts with which to ask the questions. Do we really know what questions we are asking? > But causality is a function of > experimental design. In true experiments one can prove causality. Never in science: only in wishful thinking. People once thought that Newton's theory of gravity had been proved, or that the wave theory of light had been proved. Both theories were later rejected and replaced by better ones. Those may one day be rejected too. "Currently best available" does not mean "proved". > Accurate report is the standard observable index of consciousness, > which is consistent with that definition. I have never understood why so many people accept this. Certainly we can report some things. I am conscious of white lettering on a black background, changing as I type -- and I can report my experience at that high level of abstraction. But I cannot report *everything* I see and feel and hear (e.g. listening to a symphony orchestra: the composer's score does not describe the experience). What we can report is partly limited by the "bandwidth" of verbal descriptions. That is inherently very low. The same goes for other forms of external behaviour available to humans, despite many attempts to get round this in diverse art forms. Contrast the content of very rich and dynamic visual experiences: e.g. standing on a bridge looking down at constantly changing, swirling patterns of motion in a fast-flowing river. I have no reason to believe that I can ever report what I experience there except in a coarse-grained, incomplete summary. Visual experiences depend on internal information processing mechanisms driven in large part (though not entirely) by the millions of photons arriving at the retina. The information bandwidth of the retina and other relevant parts of our brains (with huge amounts of parallelism) vastly exceeds anything we can do with words -- or even words and song and dance combined. So I see no reason to believe that the visually processed information should be reportable. Of course, it follows that it cannot be self-reportable either, if reporting to oneself involves internal descriptions in a language something like external languages with syntax involving sequences of discrete tokens, using a hierarchical syntax. J.L.Austin wrote somewhere: "Fact is richer than diction" He could have added "Experience is richer than diction". Some things can be self-reported in that sense (e.g. noticing that you are confused, or upset, or that you have not yet solved an equation). But why assume that ALL internal mental phenomena forming the contents of consciousness can be reported? I can easily design a computer program which once launched goes through a rich and diverse collection of internal processes which could not be fully and accurately reported through the available output channels (even if, unlike humans, it has a high resolution video display). (Anyone who has tried to trace the printing routines in a computing system will have had direct experience of the limitations of reportability.) Moreover I could, though with extra difficulty, build a system that not only does more than can be externally reported or displayed, on that system, but which does some internal self-monitoring, and monitors more than can be displayed externally through the available output channels. The monitoring may even process more details than can be stored in the system's own internal memory: i.e. many of the internal records (of many forms) are transitory and local. (There are different sorts of reasons for long term memory limitations: memory capacity limits, bandwith limitations linking storage areas, whether a usable connection exists between all local buffers and long term stores, the lack of any formalism for storing the information in a usable form, difficulties of access once stored, etc. etc.) I expect most people would agree that the processing in human (and some other) brains is also too rich to be expressed through available output mechanisms. So why assume that the particular subset of which we are conscious *has* to be different? If reportability were regarded as a requirement what would that say about consciousness in other animals? Are a chimp's conscious states, processes and events all reportable? What about two month old humans? If we can design working models of systems whose internal processing, including internal non-symbolic monitoring of internal processing, is too rich to be mapped onto the input-output capabilities of those systems, as we surely can, then why should we assume that for humans, or other animals, the available output channels have to have sufficient capacity to report contents of consciousness? If that is not an absolute necessity, then perhaps it is not necessary for biological organisms. A possible counter argument: we cannot report everything that's going on simultaneously, but everything that's going on can be selected for reporting (i.e. a tiny portion at a time). What reason is there to believe that? Does it apply to the rich and rapidly changing river experience mentioned above? Even the local details may change too fast and in ways that are too complex to be mapped onto any available form of external expression. Trivial answer: I can name the experience Rivex, and then say "I am now experiencing Rivex". Is that a satisfactory notion of reportability? How do I know whether what I have later is also Rivex or not? And how will you know what I mean by Rivex? An implication of all this is that we need to move to a far richer notion of explanatory theory that might be relevant to the study of mind than psychologists and philosophers are normally willing to consider. And we may need some new sets of concepts to describe what needs to be explained (not just what can be measured and observed and reported). That also means abandoning notions of proof and disproof that turn out to be will 'o the wisps. (http://www.mysteriousbritain.co.uk/folklore/will_o_the_wisp.html) Maybe more psychology students could then be taught to think and design and test and debug and analyse complex explanatory models instead of being taught only to measure and do statistics on the assumption that science is a search for correlations, because they mistakenly think that's the only thing physicists do. A tiny start on a web of theories concerning mind, consciousness and biological evolution, can be found in sketchy outline in talk 9 here (Varieties of consciousness -- in postscript and PDF): http://www.cs.bham.ac.uk/~axs/misc/talks/ The other talks add more fragments. But there is a huge amount of work still to be done. Cheers. Aaron == Aaron Sloman, ( http://www.cs.bham.ac.uk/~axs/ ) (philosopher in a computer science department) School of Computer Science, The University of Birmingham, B15 2TT, UK EMAIL A.Sloman AT cs.bham.ac.uk PAPERS: http://www.cs.bham.ac.uk/research/cogaff/ FREE TOOLS: http://www.cs.bham.ac.uk/research/poplog/freepoplog.html TALKS: http://www.cs.bham.ac.uk/~axs/misc/talks/ FREE BOOK: http://www.cs.bham.ac.uk/research/cogaff/crp/