|
The Core of LeadershipReprinted from an AOK Knowledge Network Dialogue, November 14, 2005 From: William P. (Bill) Hall, PhD Visiting Faculty Associate Subject: Reply to Marshall, Snowden & Leppik: personal vs. objective knowledge [Caution, this is a considered and long essay.) John, your simple request to define terms used in knowledge management is completely reasonable, but my experience from participating in several KM forums is that attempts to define "knowledge" more often than not degenerate into flame wars rather than any reasonable consensus! I think this is because defining knowledge and what it means to know plumbs the depths of philosophy and the biological, cognitive and social sciences - extending from science into genuinely metaphysical realms that cannot be rationally decided. Definitions of knowledge are also central to what I hope will be a fruitful (rather than flaming) discussion with Alistair, Peter, Dave and hopefully others on the possibilities for genuine artificial intelligence and knowledge. In this discussion I will argue that the major barrier to developing "knowledgeable" artificial intelligences is that to now we haven't had enough of a scientific understanding of the nature of knowledge to give sensible guidance to AI developers. However, I think at least a few people, such as Georg von Krogh, Rodrigo Magalhaes, Joe Firestone, Mark McElroy, Bill McKelvey, Haridimos Tsoukas, Luis M. Rocha and Pamela Lyon are well down the path to developing a genuinely scientific understanding of knowledge that will point AI developers in the right direction to build systems that can genuinely become knowledgeable. To me, understanding the nature of knowledge is a genuine foundation problem for the discipline of knowledge management that I first encountered 5 years ago when I started trying to make sense out of the academic and practitioner literature in knowledge management for a book I was writing on the coevolution of human cognition and our cognitive tools, Application Holy Wars or a New Reformation? A Fugue on the Theory of Knowledge (I'm happy to share the draft). Before I could finish the book, I had to go back to university and start a research program to understand the different definitions for knowledge and other knowledge-related terms. (I'm an evolutionary biologist by training but have been working in industry as a documentation manager and knowledge management systems analyst for the last 20+ years. I also spent two years of a postdoctoral fellowship studying epistemology and the history and philosophy of science. I now have three PhD students who have just completed or are completing their theses. I also have several articles on the topic published or in press on the topic. I'll begin my response to the Peter Marshall - Dave Snowden AI thread by giving what I think is a fruitful definition of knowledge. Defining Knowledge: Background The most fundamental issues in defining knowledge are the natures of what can be "known" and "truth" - which are metaphysical. I am a realist, in that I believe that there is a law-like world that exists independently from any cognitive activity and that includes cognition in its possibilities. Others may take a view that there is no such thing as an external reality, and the only thing that counts is that which is internally constructed by consciousness. I don't propose to debate these alternatives, as the many stances along the dimension of realism vs. anti-realism have been very ably set out and reviewed by Ilkka Niiniluoto in his 1999 book, Critical Scientific Realism, Oxford. Suffice it to say that I personally view the world as a critical scientific realist. Another dimension to the framework for defining knowledge is the nature of truth in the range of "-isms" from positivism to postmodernism. Bill McKelvey has published a number of papers seeking an appropriate construction of truth along these dimensions as a foundation for the organizational and social sciences (see "Postmodernism vs. Truth in Management Theory," in E. Locke, (ed.). Post-Modernism & Management: Pros, Cons, and Alternatives. Amsterdam, NL: Elsevier, 2003, 28 pages and other papers available via his web site. On this dimension, I am a Popperian constructivist (explained below). A third dimension of debate is whether knowledge is something that can only be held in a living mind known as the subject or person, or whether knowledge is something that can exist independently from the person. Two poles are represented by Michael Polanyi (e.g., Polanyi, M., 1958, Personal Knowledge: Towards a Post-Critical Philosophy, University of Chicago Press, Chicago; and 1966, The Tacit Dimension, Routledge & Kegan Paul) and Karl Popper (Popper, K.R. 1972. Objective Knowledge: An Evolutionary Approach. London, Oxford Univ. Press, 380 pp. and Popper, K.R. 1978. Three Worlds. The Tanner Lecture on Human Values (pdf). Delivered at The University of Michigan April 7, 1978 - Again, as does Niiniluoto, I follow Popper's construction. Some definitions of knowledge from the KM literature are given by Stenmark, D. (2001). The Relationship between Information and Knowledge. In Proceedings of IRIS 24 (pdf), Ulvik, Norway, August 11-14. A common definition used by many practitioners is that knowledge is "justified true belief". This assumes that it is possible to know the truth and justify that what you claim to know is true, and that both "truth" and "to know" are well defined concepts. The conclusion I draw here is that any definition of knowledge needs to be firmly grounded in a theoretical framework. Both Bill McKelvey and Haridimos Tsoukas (2005 Complex Knowledge: Studies in Organizational Epistemology, Oxford) are attempting to build a scientific basis in organizational studies for defining knowledge in an organizational environment. My approach begins in Popperian epistemology and biology, but ends up with similar conclusions. Popper eventually defined knowledge as solutions to the problems of life, and argued that knowledge grows through time as the product of selection. The biological origins of knowledge Here I will give only a brief abstract of an epistemological framework that I have developed more extensively in three research papers:
My framework is most fundamentally grounded in Popper's epistemology of three worlds as presented most fully in his 1972 book, Objective Knowledge. According to Popper, existence involves three ontological worlds: world 1 ("W1") - the physical reality that encompasses everything (i.e., the "reality" of critical scientific realism); world 2 ("W2") - the world of cognition and subjective perceptions of the world held in the dispositions of cognising entities; and world 3 ("W3"); knowledge that has been codified into persistent forms able to exist independently from cognitive entities and that has intersubjective meaning for such entities. As an aside, it should be noted here that Michael Polanyi's personal and tacit knowledge, frequently referenced in the KM world, are fully encompassed in Popper's W2. The lack of any attempt by these two authors to coordinate their respective ideas was most likely due to personal issues of mutual contempt and metaphysical issues relating to faith vs. reason, rather than any fundamental incompatibility in their epistemologies. Watkins's 1997 obituary of Popper describes the event that most likely engendered their personal conflict: "When in 1951 Findlay, [Popper's] old friend and admirer from New Zealand days, arrived at King's College London, across the Strand from LSE, he found Popper... surrounded by a court of admirers, his conduct of his seminar magisterial, and his views of people moralistic and prejudiced. Michael Polanyi was gravely offended by the treatment that Popper, as chairman, meted out to him when he read a paper (on 'The Stability of Beliefs', 6 March 1952) to the Philosophy of Science Group." The relationship between Popper and Polanyi seems to have many similarities with the much better documented antipathy between Popper and Ludwig Wittgenstein, one of the other great philosophers of the 20th Century, as documented by Edmonds & Eidinow's 2001 book, "Wittgenstein's Poker", who also comment extensively on Popper's abrasiveness and intellectual arrogance. I amalgamate Popper's ideas with ideas about the nature of living things following from Humberto Maturana and Francisco Varela's works on autopoiesis and cognition - first introduced into the organizational theory and KM domain by von Krogh and Roos and Magalhaes (see my 2005 papers for citations). Autopoiesis describes the inherent properties of metabolism, complexity, self-recognised boundedness, self-regulation, self-maintenance and self-production that collectively sets living things apart from the rest of existence. Aspects of the autopoietic entity's structure that enact these properties can be said to be a primitive form of W2 knowledge - an argument that is ably developed by Pamela Lyon in her 2004 paper, Autopoiesis and knowing: reflections on Maturana's biogenic explanation of cognition (see Hall et al. 2005 for citation). Maturana and Varela's work contributed
substantially to the radical constructivist view that all knowledge
is constructed internally and can never be proven to be a true
reflection of an independent reality (see von Glaserfeld references
cited in Hall et al 2005). Although I have not seen this recognised
in the literature, Popper (1972 and However, Popper differs from the constructivists in that he accepts (1) that external reality exists, (2) theories about the world are needed in order to cope with the problems of life (3) that it is possible (though unprovable) that a claim or theory about the world may be true through its correspondence with that external reality, and (4) that the truthlikeness or verisimilitude of claims to know can grow closer to that truth by making many claims and criticising them to selectively eliminate erroneous claims (i.e., selectively retaining claims that have proved to be comparatively workable). In shorthand, Popper called the iterated process described in (4) the "tetradic schema", consisting of (a) problem, (b) tentative theories to solve it, (c) criticism through error elimination and ultimate selection of an action responding to the problem - which changes the world in such a way that new problems (d) facing the entity are not quite the same as that which was successfully solved by the entity. In his most elaborate discussion of the concept, he called the process of knowledge growth his "general theory of natural selection". In several places, Popper briefly asserted that W2 knowledge is an emergent property of life. Popper accepts that codified knowledge in the form of DNA and in computer memories, as well as the products of human expression such as manuscripts and books qualify as forms of W3 knowledge. Without citing Popper's ideas, Howard Pattee and Luis M. Rocha (see my 2005 papers for citations), focusing mainly on "codified" knowledge, also argue at substantial length, from the standpoint of fundamental physical theory, that knowledge can only exist as an emergent property of life and can achieve codification only through evolutionary processes. Speaking as an evolutionary biologist once particularly interested in the origins and evolution of genetic systems, it is clear that processes to codify and retrieve knowledge in the form of RNA and DNA were achieved early in the evolution of autopoietic organisms (but note that there is nothing in autopoietic theory claiming that codification is a necessary component for the origin of autopoiesis). Only with the origins of linguistic expression in early humans do we have the development of a major alternative means to codify the cultural knowledge of our species. The bulk of Popper's 1972 book focuses on the existence and nature of objective knowledge in W3. He also shows the importance of intersubjective criticism in the growth of knowledge that can be expressed objectively. Erroneous knowledge that is expressed only in W2 through the dispositions of the individuals carrying the knowledge is criticised mainly through the failure/death of individuals expressing errors. When a claim to know can be expressed linguistically, it can be criticised intersubjectively and eliminated as a bad idea before it has to be expressed in action. Knowledge and the capacity As a one-time journeyman technical writer, whose profession was to collect and distil data, information and the personal knowledge of other people into objective documentation able to transfer that knowledge in objective forms to other people who needed it, I can never agree with the many in the KM discipline who claim that knowledge can only ever be held by the mind of a person. (Note that I do not in any way denigrate the importance of personal knowledge in organizational knowledge management - see Nousala, S., Miles, A., Kilpatrick, B., Hall, W.P. 2005. Building Knowledge Sharing Communities Using Team Expertise Access Maps (TEAM) (Word download). KMAP05 Knowledge Management in Asia Pacific Wellington, N.Z. 28-29 November 2005. Given the understanding of knowledge growth and evolution deriving from Popperian autopoiesis, I see no mystical aspects to the nature of knowledge that prevents knowledge from being transferred to and processed by an electronic system. Also, given that evolution by natural (or human) selection is a happenstance process that depends on satisficing rather than optimising solutions, I see nothing to prevent us from building "artificial" intelligences based on more optimum processing architectures than the wetware we have in our heads. Silicon logic gates have already been shrunk to the size of single neurones, and assuming no insurmountable catastrophes, two more decades of shrinkage combined with new technologies (e.g., see 2003 Jurvetson, Steve T. 2003. Transcending Moore's Law with molecular electronics and nanotechnology will give us tools of a cognitive capacity that are unimaginable today. It is not even necessary for a knowing AI to be autopoietic. Given that we are beginning to understand cognition and the growth of knowledge, there is no reason to suppose that we cannot build machines able to think independently, and to be able to do at faster and better than we do with our wetware. In my parents' lifetimes calculators were women using pencils and paper (or perhaps slide rules). The same calculations are performed today in less than a billionth of the time it took human calculators to perform them. In the 1930s all calculations had to be processed via a human brain. Today, in many cases even the numbers are collected automatically, and the only thing a human sees is the result. Thus, even though many of our knowledge management problems remain human oriented, I think there is no doubt that our profession will be impacted by increasingly powerful cognitive tools that have increasingly greater impacts on what we still consider human knowledge. My own life is an example of the kinds of changes we can expect. When I started tertiary education in 1957, my personal KM tools were a slide rule, a manual typewriter and a library card that gave me access to the books and magazines in a university library. Since then I have progressed through using all generations of computer technology from an early Burroughs system whose memory consisted of a magnetic drum able to hold 1024 32 bit words that could be communicated with only via machine language encoded on a paper tape - to today's super computers. I acquired my first personal computer in late 1981 that had a 64 KB and two 8" floppy diskettes each offering about 90 KB. I considered it portable because I put it on a wheeled frame. I am currently writing this on a Pentium 4 desktop machine that cost ~$1,000 four years ago that is $1,000 Pentium 4 computer that has more than 3.8 billion times more raw processing power than was available to one of the nation's largest life insurance companies 50 years ago from a million dollar room full of electronics. Another interesting datum: For what a single book cost a scholar 500 years ago, today the scholar can purchase a book-sized laptop computer able to access a significant fraction the entire corpus of humanity's knowledge recorded in World 3 via the Internet. When we started assembling maintenance routines for ANZAC Frigates 14 years ago, they were drafted by people as standalone paper documents where the maintenance activities on board the ships were scheduled and resourced by human maintenance managers. Much of the metadata governing aspects of the maintenance activities was delivered in completely separate and discrete databases. Today, we are writing modular information in the S1000D international standard for interactive electronic technical manuals within an integrated product data management environment that collects all engineering and logistic support information and knowledge relevant to particular ship components in a database we deliver to our client's on-board maintenance management system. The maintenance management system determines what maintenance is needed when, and then assembles and prints out maintenance instructions as required for the particular maintenance to be performed on the day. We have also developed condition monitoring, auditing and feedback system able to analyse operational details on spares usage and downtimes to give us feedback on the performance of our logistic support deliverables and maintenance routines, allowing these to be optimised and corrected with new knowledge deliverables to the ships. These feedback processes are also being increasingly automated to the point that our ship + people systems will be largely self-maintaining and self-regulating, where knowledge growth is effected by the system itself - for the development of a genuinely intelligent ship. Thus, given my life and hands-on experience with theory, technology and hands-on practice, I have no doubt that within 20-25 years we will have artificial intelligences that are faster and by far more accurate than are unaided humans. Whether such AI's will also have the creative ability to discover new things remains to be determined, but as long ago as 1983 when I was confronting the epistemological issues in my research in evolutionary biology I believed that it would not be too difficult to write the appropriate algorithms (see Fig 1 and associated text in Hall, W.P. 1983. Modes of speciation and evolution in the sceloporine iguanid lizards. I. Epistemology of the comparative approach and introduction to the problem. (in) A.G.J. Rhodin and K. Miyata, eds. Advances in Herpetology and Evolutionary Biology - Essays in Honor of Ernest E Williams. Museum of Comparative Zoology, Cambridge Mass. pp.643-679. I look forward to further discussion, and promise any replies will be much briefer! William P. (Bill) Hall, PhD Visiting Faculty Associate |
