EXYSTENCE NoE Seminar on 12 Nov 2004, Helsinki Dr. Theodor Barth, SINTEF Industrial Management, Norway. Generative Analysis of Knowledge in Disordered Systems - Complexity and the Art of being Specific in Love and Business. 'Love', according to Plato is an intermediate stage between possessing and not possessing. For human beings, always striving, always becoming and never satisfied, this 'love' is a true description. What I am going to concern myself with here is the co-evolution of human beings and machines to see if we can find meaningful ways of looking at complex goal-seeking behaviour. However to introduce myself, I am a social anthropologist. Social anthropologists are people who do field work, write detailed descriptions and think about them afterwards and I have been working a great deal with (?) Barth who is my namesake, but not my father, though he is the father of the approach called generative analysis in which I am interested. I am curious as to whether or not there is a confluence of the methodologies of complexity science with those of generative analysis. First off, I want to ask why Emmanuel Wallestein (?), a French social scientist, is conspicuously absent from the complexity science reference list, because he has published in this area and in my view written some interesting things. He made a very broad statement for complexity research in social science in the Gulbenkin (?) report of 1996 called Open the Social Sciences and the team that worked under his leadership was concerned with issues that opened the social sciences to the natural sciences, technology and also to the humanities. These views are based on a book written by Wallestein (?) in 1991 called Unthinking the Nineteenth Century Social Science. His take on complexity theory is a broad contribution within social science, particularly emphasising the development of a multi-disciplined approach and he proposed that a complexity framework might better enable us to understand our world in a more active way. He argues for a research programme in social science to understand how ideas become realities and we should note that he insists on 'realities' and not 'reality'. This is the process by which ideas coalesce to become complex. In other words previously unrelated realities start to manifest collective or systemic behaviour. I want to enquire into such dynamics to see if we can combine our methods and get our views to coalesce. I think it is important that this takes place in Finland not only because of the work on innovation, but because Finland, despite its linguistic uniqueness or perhaps because of it, has grasped contemporary European ideas in a very short time. I think it has succeeded in attracting an intellectual spark from the Greek Alexandria which to me was the epitome of European culture. I think the fact that we are here, at the extreme of the continent is a good omen. First I want to ask how our interaction with technology conditions how we understand and communicate. I must admit that I have had some trouble in applying the concept of emergence to my work and one of the things I wish to explore today is how unordinary things become ordinary. Or to put it another way, how fresh knowledge becomes tacit. So I will be going into the tacit knowledge discussion from an alternative end; not discussing how we might apply it from a knowledge management point of view, but how we might build better IT systems. There could be a number of approaches resulting from our understandings of that technology. Now there may be many critical voices as to whether it is meaningful to codify tacit knowledge because when you do it then ceases to be meaningful in the way that it was. The first criticism is that when we learn to relate to the codified knowledge, our ways of working change and one of the problems with that is change itself. As David Snowden loves to say, best practice very quickly becomes past practice. Businesses consultants come in and try to straighten out systems by making add-ons in so-called maintenance and feed ever more doubtful practice into the system. The second reason why we might be sceptical of codifying tacit knowledge is that what we see as the tasks we have today are a result of a burgeoning or seminal post-industrial way of working. We are challenged with jobs which ask us to do things based on prefigured descriptions of work processes. The third reason we might be sceptical about codifying tacit knowledge, and I use the term 'tacit', but maybe we just mean any knowledge, lies in our understanding of what expert knowledge is. The company that I am working with in Norway has a knowledge support system of experts called 'the supreme court'. It is in many ways a very intelligent system as far as it goes, but how do we define the intelligence of such systems and how do we understand the interaction between human beings and such systems? What makes the Supreme Court people experts is that they know about short cuts that others don't and more importantly when they apply. But in as much as they are one step removed from the actual situation what they know are more 'rules of thumb'. We could of course imagine, รก la Turing, that we could have some form of artificial intelligence that would correspond to an all-wise human being, though we're a long way from that. The kind of agent technology we see applied these days usually makes suggestions a bit like the light advertising in big cities. It flashes at you all the time until you no longer take notice of it. To summarise. What I am trying to say is that a lot of valuable knowledge is tacit and practice and reflection warrants all kinds of codification which brings a number of problems. I'm trying to highlight the difficulties of describing or codifying and I think some systems could be quite good at describing the first loop of learning which is doing things right as procedures or following specifications. Whereas a second loop is doing the right things in a different way and this requires a different type of add on if you want to support it with technology. That's my main argument. What I want to do is to look at emergence and ask how holistic properties that are related to the interaction of elements in the system become folded into the system. How is it that human systems have such holistic manifestations and how are they in turn internalised to become part of a repertoire of professional identities. I am also concerned with the fact that such holistic properties, as Eve pointed out, are multidimensional and that there can be many holistic properties in the system at the same time. I suppose that's why we talk about complex systems being somehow textured, in that the properties that I'm talking about can be of many different types and sometimes contradictory. They are not necessarily cohesive and if they are, they may be cohesive at a different level. The production of cohesiveness or coalescent traits that are multi-dimensional and sometimes contradictory is something we can study. We should also consider the fact that systems can go from a complex dynamic situation to a simpler one. I want to give you an example of a company that went from a very dynamic and complex organisation where the management skill and knowledge was distributive and mobile, to a situation where they became bottom-line focussed and changed into a more rigid production company. This kind of development is tragic for our gospel but it's something that we should try to understand. I want to present a case history and give you some of the rationale for the development. After that I will return to some of the theoretical backgrounds that I use in general complexity and then give you a model from my own field which is generative analysis of 'discordant systems'. This is a concept of generative analysis and it is close to other concepts which we find in complexity theory though this connection has not been made explicit in academic publications. There were two companies that merged, one of them I will call Simplicitas Inc. and the other Theodorakis. Theodorakis is a company selling a modelling tool and it merged with Simplicitas to gain more experienced personnel and better technology. We have to look at the dynamics of the organisational history along with the technology. This poses a lot of difficulties because it's like trying to give a commentary on a book while at the same time describing the people that are reading it. In the first phase before transition Simplicitas had a CEO who operated in a similar way to our current visions of the future job market. The company took pride in its competent management of human and technological resources as a distributive and mobile enterprise. It incorporated what we might call 'triple loop learning', which is a concept from a book by (?) called Diversity Management. So our assessment was that we were working with a sense-making organisation; working with narratives and that we had the opportunity to bring about organisational change through our interaction. The rationale behind the Theodorakis merger was that it was to boost the power of models used in user-participant projects and that the environment would serve as a common reference for system developers and programmers. A second aim was to develop the seminal market for the Theodorakis modelling tool in the United States. However the returns on Theodorakis did not come as quickly as expected by the Board of Directors who then required a change in management. The CEO was got rid of and the company was organised into business areas one of which was devoted entirely to the Theodorakis tool. Eventually it turned out that the Theodorakis tool struck a booming market in the US public sectors and soon became the driver of Simplicitas Inc. The new CEO thus earned himself the reputation of an efficient manager, the company structure was changed and adopted the role of servicing Theodorakis users in the US public sector with some of the companies in Sweden shutting to assist the daughter company in the US. From a technological point of view Theodorakis was backstage in that it was used by the IT consultants and programmers as a reference to build systems based on user participation in AI, object orientation and interface. On the whiteboard I've put up three boxes. The top is a process description of a codified work process, below is a state report on communication and logistics and to the left, the one with knowledge resources. If you look on this as a supreme court, for example, then you would have legal occupants in the left corner with the legal process on top as a specification of all the routines and then you would have the current state of communications below at the interface where the user consults the system. The three of them are synchronised in time, allowing the user to triangulate the knowledge resources of the system. Theodorakis moved to front stage in the US and was used as a Trojan horse to market the previous technological achievements of the company. Now we have the choice of understanding the change in terms of linear evolution or a phase transition. The distinction made by Wallastein between a linear evolution and a phase transition is that a linear evolution is a transformation in time whereas a phase transition is a transformation of time. If we want to be naively realistic we could draw the conclusion that when people saw that Theodorakis was going to be successful in the US they got rid of the old CEO and the social democratic company values along with him. We may also think that the company didn't have lasting values after all since it was trying to sell its technology as a Trojan horse. There are all kinds of pros and cons we might bring to this view though we would probably only end up with yesterdays wisdom on capitalist enterprises. In order to understand the changes in Simplicitas Inc. as a complex system and the evolution generated by human interaction we should perhaps use some deep time analysis. I want to posit six different layers in which each constitute a world, the general meaning of which will be explained later. Each world has an emblem or a boundary object and when I talk about objects I am talking about relational objects. The objective is to learn something about knowledge management as an endowment between the public and private sphere of ownership and see whether it is some sort of currency or exchange of patterned intelligence. How is knowledge linked to value creation? When I came to Simplicitas in 2000 as a social anthropologist from the Centre for Industrial Management of Technology in Society, I became involved in projects on the use of narrative as a knowledge carrier, as a holder of reflected practice, and as a sense-making approach to the management of organisational change. This tried to focus on co-generative learning organised around large group events, but since I had a field office in Simplicitas I took the opportunity to do extensive field work following traditional methods of observation. In other words I was partaking of day to day conversations and stories told round coffee machines etc., about the development work using Simplicitas technologies and more conventional issues. I like to think of Simplicitas inc as a K.I.T.E (a Knowledge Intensive Technological Enterprise) and before the transition Simplicitas inc. was a 'synarchy'. We have talked about hierarchies and heterarchies, synarchy is a word for joint rule. It presupposes a shared state as a distributed and mobile organisation and that there are features that manifest themselves above or at another level of complexity than the interaction between agents. So Simplicitas was a truly complex organisation with the wisdom of complexity management and it meant that we were doing research with an enlightened organisation. I'd like us to consider six different developments or worlds which I shall call World 1, World 2, World 3, World 4, World 5 and World 6. Such worlds can be assessed in many different ways and we can go in and talk about them at any level. Each world is structured by four boundary markers; 'Background' and 'Purpose' are chronological boundary markers, 'Entry' and 'Exit' are spatial. Background for World 1 was that Simplicitas Inc. needed case-based learning from artificial intelligence (A.I). This was a way of problem-solving and understanding the importance of context for learning in general and for understanding how knowledge retrieved from a system can elicit competent action. Development of systems was not only concerned with representation for use by agents but also concerned with representation to its general users. The Purpose of Simplicitas Inc in World 1 was to grow the potential of human to computer interaction (H.C.I) with interfaces that optimised access, use, storage and requirement needs of customers. Entry for World I was a scenario of three computer windows to help users triangulate knowledge at all times. One window would, for example, show a display of document resources, another, process in terms of access to procedures whilst the third window would show the current status of the project in terms of the logistics and communications log. Exit was the development of a method that made the distance between the prototyping and programming as small as possible. This was achieved by user and customer participation and often required the attention of a projects manager to liaise between the customers and the developers. Background for world 2 was that Simplicities inc. moved from working electrical systems in the Norwegian process industry to developing multi-user systems for large bureaucratic organisations in the public sector. The Purpose of Simplicities inc. in World 2 was to reveal to their users in a step by step fashion, how much of their work, even innovative and creative work, was bound up in routines. By formalising these routines the quality of public services could be developed and the stumbling blocks of modern bureaucracy removed. Entry of World 2 was describing a world of process and procedure in which people delegated mental operations and precision-required operations to manual procedures on a mouse pad, a keyboard and abacus etc. Here the alternation between mental and manual operations do not flow as freely as in the previous world and step by step monitoring of the manual operation was a condition for the output to be accepted as the equivalent of the mental operation. Exit from World 2 was to experience the distance between the job that had to be done in building the actual systems and the prototypes inherited from the architect consultants working with the users. Users experienced different needs when the system is delivered than when they participated in its development. World 2 was deeper than World 1 because the distance between the virtual and the actual properties of the system was larger. Background for World 3 was as an enterprise culture in which people were expected to see challenges that they found interesting and that contributed to their professional development. Whenever the project portfolio was thick they were expected to chip in and when lean to devote time and energy to it. The portfolio was like a cherry tree; people sometimes picked from it and sometimes it picked them. World 3 Purpose for Simplicities Inc. was as a periodically loose/tight group of urban hunters and gatherers living from projects. In this world the company had 25 process owners with a total of about 170 employees, but the total R&D budget was no more than 5%. Entry in World 3 Simplicities Inc. was a market place in which people placed their bids and projects and projects placed their bids and people. This alternation is mediated by a collective environment, one in which services were available through the interface of stationary PCs connected to an intranet. Exit from World 3 was that there were ever pending tasks of uptake and maintenance for which there were no clear procedures. No mobile devices were used except for mobile phones and most transactions or agreements were reached by face to face encounters. Virtual resources consisted of sophisticated billboards available via a common interface and again it was a deeper world than the previous one because the connections between the actual and virtual properties of the system were difficult to trace. Background for world 4 was that during the 90's Simplicitas Inc. started to use the DSDM (Dynamic Systems for Development Method) approach to systems development. This was largely due to past experiences with sequential systems development where the customer's initial requirements did not match the needs revealed on delivery of the systems. Purpose in world 4 was that DSDM was a way of representing the customers and formulating their needs in a stepwise way whilst iterating model prototypes and working synchronously with a totality of models. A significant amount of effort in the delivery process of an IT system was invested in the maintenance of contract relations with the customer and the management of customer expectations was an important issue. It required a network of people who understood each other in terms of contract maintenance and customer expectations. Entry into World 4 was one in which facilities were booked by employees who turned up with groups of customers Exit from world 4 was that some of the consultants felt that they were primarily involved with people and that technology played a marginal part in what they were doing in projects. Some even grew to become techno-sceptics which was startling in a company based on a love of technology. There was an effort to create an arena for common sense-making using narratives or stories by taking the whole company to a mountain resort for a day. In world 4 the distance between the actual and virtual reached a maximum. World 5 Background was that Simplicitas Inc. had acquired the Theodorakis modelling tool through a merger in 2000 with a company that developed it through customer projects since the mid 80's. The original vision was that the two technologies could be merged and used as a common reference by consultants and developers. In World 5 the Purpose of Simplicitas Inc was to market the Theodorakis tool in the US and to service the needs of the public sector. The risks and stakes were high so the company also targeted big robust long term projects in the Norwegian public sector. Entry for World 5 was that professional interaction was expected to converge towards value creation. It had similarities with World 2 in the sense that it required a meticulous hour by hour account of tasks delegated to the knowledge engineers. Exit in World 5 was that the second CEO took the lead from the first as head of Simplicitas Inc. The relationship between actual and virtual properties of the company as a system were difficult to trace and one of the major shareholders fired all the C.E.Os of its daughter companies in order to show shareholders its willingness to act. The Background for World 6 was that the customers in the American federal organisation were developing an expertise in modelling and forming small teams working through the Theodorakis team server. The models themselves were intransigent to general ICT users in that they were mechanical and worked as a sophisticated abacus though they required the knowledge and dexterity of an expert to yield the benefits. As an abacus the model related objects that could be described and were searchable and could be grouped in very complicated pictures. The program had a capacity of about 100,000 objects. Purpose in World 6 was that the IT systems of Simplicitas inc. were to be reoriented to dual purpose as support and designed to respond to the patterns of use created by the end user and then made available to others. Entry in this World was as a possible world in which Simplicitas Inc picked up its long time interest in knowledge work and experience in adapting expert technologies to larger groups of users and starts developing the Theodorakis tool as a general application for personal users without a particular background in modelling. Exit for World 6 was a return to World 1 where object orientation was used as an approach to modelling holistic properties, though now the Theodorakis tool wasn't used for that purpose and contained a lot of technology from Simplicitas systems for which it was designed to interact. The relationship between the virtual and actual properties of the system was fairly tight in the broadening population of users.