Knowledge Management Model for the Real (Irrational) Organisation
Research Proposal (Revision 2.0)
This proposal has been prepared in support of an application for a Part-time PhD. It is based on some existing unstructured research conducted in a K-Blog (or Knowledge Web Log) currently hosted at Psybertron. More background to this proposal can be found there.
Information and communications technologies are becoming ubiquitous and convergent at many levels. This creates an illusion that communication itself is somehow easier or commoditised, whether concerned with informal “browsing” or structured business transactions.
There has been considerable focus on the business models (or lack of them) needed to exploit opportunities arising from this, but relatively little on the information models needed. Paradoxically the more ubiquitous the technology, the more inadequate traditional models become, and the reasons are concerned with human organisational behaviour rather than the technology.
Computer applications have been (reasonably) successful in the past, because their explicit logical models and deterministic, even statistical, behaviour were adequate models for their implementation. Automating operations, or their decision support, between operations in a single organisation have relied on the fact that the humans in the business cycle, whether involved in the business operation itself, or the analysis and design of the system, shared some context. To a large degree, the semantics of any communication involved, relies on the context of the participant, rather than the content of the communication, even between integrated systems.
The more extended become enterprise business models, and the more commoditised the communication components, the more business opportunities there seem to be, but the less the participants can share common context, and the greater the risk of mis-communication, and mis-management of the relevant knowledge.
The thesis is that information models that do not rely solely on objective information, but also take account of context and subject are likely to prove more successful. This is not a new idea, and in fact it lies behind concepts of artificial intelligence, natural language processing and pattern recognition, where, whilst there are examples of successful implementations, in general these have failed to deliver business benefit.
The objective is to propose and test alternative knowledge models, which are pragmatically implementable, but which exploit appropriate non-classical logic. Several avenues are currently under investigation, starting with fundamental philosophical views of knowledge, non-classical logic, fuzzy-logic, chaos, complex systems and quantum computing to name a few, but linking these to topics previously explored in organisational behaviour and culture, around decision rationality / action irrationality in managing change.
The Hypothesis :
Whether concerned with structured communication between organisational units or individuals involved in business, or concerned with less structured information searching or exchanges between individuals, there is a ubiquity and convergence of communication technologies – www for short. The nuts and bolts of communication are getting simpler and more freely available.
Not only that, but protocols for defining and representing information being communicated also appear to be converging into very similar mark-up language schema and document object meta-models, whatever the domain, and whatever the level of structured granularity and formal management in the communication.
Together these create an illusion that communication is itself simplified and commoditised, so that the focus of business is to find and exploit applications of the technology. However, as information models become more flexible, generic and remote from specific application domains the more context, or its significance, needs to be captured (explicitly defined or inferred implicitly), if the semantics of the communication are to be preserved.
In less structured communication contexts, there is already recognition of the value of artificial intelligence, inference engine, pattern recognition, learning strategies in increasing the reliability of extracting maximum meaning from communications.
There is an arrogance in more structured information applications that tends to overlook that the explicit information model used is after all only a context specific approximation of the real information involved, once the model is itself established. Paradoxically therefore, as information sets are shared more widely and exploited for more flexible re-use, beyond the original context, the greater the risk that small errors in the definition of either content or context, will cause mis-information. This is potentially catastrophic if the information is being used to support a critical business or operational decision, whether the action is automated or human mediated.
Whichever the case, a single decision based on a single potential misunderstanding, is unlikely to be the basis of a “mission critical” application, and in practice, human users are often far more creative in their use of a system, and in interpretation of its outcomes, than system design explicitly allows for. The net result is inefficiency and ineffectiveness in communication between applications, and an overhead in avoiding the risk of catastrophic mis-communication, rather than catastrophe itself. Paradoxically the same inefficiency is also a source of damping against precipitous mis-guided action.
Given the above, the ultimate objective of the research is to make progress towards, information models or implementation strategies, which are ; (a) tolerant to imperfections in source data and schema or effectively infer their own mappings between contexts, and/or (b) recognise the role of human interpretation in the information communicated and the model used to describe it.
Where is the Value? :
In general terms the value of “better” information modelling is clear.
More specifically, in several industries close to the proposer’s own experience there are clear business cases for exploiting generic and extensible information models in support of the operation of large complex capital intensive engineered assets. (Everything from aircraft, shipping, processing plants and production platforms, and the whole vertical industries supporting engineering, supply and lifecycle support for field operations and maintenance of such assets. Equivalent initiatives and business cases exist across the broadly parallel civil and transport infrastructure sector. The dot.com / B2B / e-Business cycle has raised awareness in these industries that most of the information modelling issues exist across any large distributed enterprise in any industry sector involving semantically rich communications.)
Being such a vast subject area, there is no hope of reducing the problem to a single “issue” and no future in seeking a silver bullet to solve it even if it were. Not the least significant argument against seeking a single high value solution is the scale of the consequences rather than any inherent complexity. There are too many businesses with too much to gain and too much to lose from a panacea for all electronic communication, to give the idea a second thought.
The value in succeeding in defining a “better” generic information model, is going to be in applying it to a particular business case in a particular industrial domain. Unfortunately most high value business cases in the proposer’s experience are concerned with quick-win solution implementations, with no allowance for the idea of an R&D phase prior to business analysis and solution design. It will be necessary to have some tangible results from the R&D investment before the concept can be applied directly to such business opportunities.
There is no foreseeable shortage of such opportunities however.
Significant prior research:
The proposer has been involved for over five years in the application of generic information modelling concepts to asset lifecycle management of project assets in the process plants industries. The application of generic pan-industry models relies on collaborative efforts and standardisation agreements, and the proposer has also been active in many cross-industry and ISO industrial data standardisation initiatives.
Most early attempts in this sphere adopted traditional entity / attribute data models with ontologies and taxonomies deemed to exist in particular industrial application domains. Of necessity such an approach “freezes” a model of a given industry in a form, which is simple to implement, but inflexible in use. Later attempts have recognised the benefits of flexibility and extensibility by adopting more generic object / relational models, to allow for business process re-engineering opportunities that are facilitated by the very data integration possibilities created by such models. These models are notoriously more difficult to implement in scalable applications, and these later attempts cannot really be claimed to be successful in much beyond proof-of-concept implementations, but already many domains, which started with explicit models, are starting a migration to the generic approach.
Even these generic models require a basic meta-model of the fundamental entity types that exist in the world, irrespective of the particular industrial application. For this reason there has been a learning curve, which has taken active participants into areas of philosophy and meta-physics, which would have seemed a million miles from the pragmatic industrial and business scenarios being addressed. Unfortunately this learning curve has been applied inconsistently across several different work initiatives in progress, and the level of philosophical thought being applied is patchy to say the least, ranging from “barrack-room philosophy” on the one-hand to somewhat random selection of ideas from important, but disparate schools of philosophy.
Not only have such models been built on inconsistent or philosophically suspect premises, there are a number of other possible shortcomings being overlooked. The first possible shortcoming is that the focus has been the physical world and a model for what actually exists. In my view, for business or other human organisational needs, a better focus might be activities and intent. Secondly, we would do better to focus on a model of what is known and communicated about these entities, rather than the entities themselves, bearing in mind that however fundamental our view of the actual world, our information is likely to be imperfect or incomplete at any given time.
Given these views, the proposer has been researching a number of avenues.
Following the general convergence of hierarchical and “grove” type document object models and similar in the w3c and Oasis web standards area. (REF)
Researching basic epistemological theories, starting from earliest objectivism / positivism of Wittgenstein to later Wittgenstein and Russell, and gravitating towards most recent interpretivism after Walsham et al. (REF) Taking in Artificial Intelligence & SP Theory.
Investigating concepts of complexity and chaos, in relation to imperfect information and context definition. (and related “many worlds” / probability based views) (REF)
Reviewing the ISO SC4 WG10 Data Architecture projects and in particular the ISO-18876 proposal known as IIDEAS (Integration of Industrial Data for Exchange, Access and Sharing) (REF)
Following activities under the “KnoW” (Knowledge on the Web) initiative, and proceedings of various semantic web / knowledge technologies conferences. (REF)
Personal Motivation and Qualification for Research:
The proposer’s research to date has been informal based on following interesting leads as a spare time activity. This has so far led to a broadening of areas of relevance and a risk of losing focus on what remain the core issues of interest to the proposer, namely :
Information models which recognise imperfect knowledge
Information models which focus on communication intent
The proposer is already active in a number of initiatives aimed at developing, implementing and gaining agreement on standard generic information models. Of necessity these involve broad cooperation and a great deal of international “committee” working. This is a long and slow process, full of political and pragmatic compromise and recycle, for all involved. Agreement is ultimately a democratic exercise, whether by voting, or by de-facto popular acceptance of solutions made available commercially.
There is no off-line space in this self-motivated but cooperative environment, or in the business of a normal day job, for thorough objective research, until such theses are sufficiently developed to support specific recommendations. For this reason I see the core issues as valid subjects for independent research by an individual.
As well as being personally committed over a number of years to the subject, and having formed strong, but thoughtful, views on the significance of the issues to the success of generic modelling, I am personally very motivated to undertake the necessary research.
Not only that, my working style is more naturally analytical and suited to research, than implementation of any model I believe to be fundamentally flawed. This is in large measure actually a part of the motivation itself – ie if I don’t research and bottom out the issues, I will have problems making progress with any future implementations. Also, given the subject areas, essentially engineering and IT applications on the face of it, I also find it unusual amongst my peers to have such a strong interest in the human / behavioural (soft / harder) aspects of the subject, that the more obvious concrete (hard / easy) aspects.
I believe this commitment and fit of style and motivation with the proposed subjects is borne out by the experience of my previous MBA thesis amongst other things. The subject was cultural and attitude aspects of change management in the engineering contracting organisation of my then employer. My project and thesis was awarded the highest marks for my year on the course. (REF)
The subjects discussed above clearly skim across a wide range of information and communication subjects. A major part of the research methodology must involve a thorough search and review of published material. Early research shows a wealth of relevant material easily accessible via the web – both formal publications via academic institutions and journals, and also many special interest group sites and individuals.
I intend prior to too detailed analysis of academic materials (old and new) to capture and crystallise my own impressions of the importance of what I have identified as key issues, based on the anecdotal evidence and analysis of my own experience.
I happen to hold the opinion that this subject will actually yield a great deal of learning simply from analysis of published evidence (again I need first to analyse why I believe this). Each time I find references taking me into an apparently new area, eg from engineering data to info-modelling to forms of language to AI I find parallel issues recurring. It is difficult to identify, without assistance of others, any immediately testable hypotheses. One immediate objective of the initial research should be to develop such test methodologies. This initial period could easily be a full year.