Dear Thomas,
I look forward much to see you again in Bratislava. I am to be a participant of the 2 days training to be held in my this city. I am pretty sure that my participation will be very valuable as I already e3xperienced, among others, your excelent lectures in Leiden (arranged by Gerard Freriks). I hope that in course of your stay we might meet also informally aiming to exchange ideas related to issues of mutual concern. Maybe we could spend also some pleasing time in some nice pace.
The propsed training agenda seems to be interesting and promissing. Particularly due to my certitude that standardisation is a key iissue for both sides of the coin called knowledge. By these coin sides I mean declarative knowledge and its procedural komplements operating on and with the descriptive items e.g. archetypes (frame or object like bits and pieces). My experience and skills accumulated in course of decades lasting developments of knowledge based (expert) systems and their successful real implementation in practical healhcare services makes express my discomfort with regard to concern given to descriptive-procedural knowledge relations.
Let me accentuate the phrase ‘procedural knowledge.’ If extensive sets of correct and perfectly designed archetypes and templates would be available without complementing proceduralities, their computerized knowledgeable applications would be extremely limited. This assertion is based predominantly on extensive experience with expert system applications not only in the medical domain. This supports various theoretical assumptions e.g. the ones stated by Kenneth Craik [already in 1943]. More convincingly and influentially by Phil Johnson-Laird as well as Paul Thagard [1996)]. These scientists clearly showed that excercising rational human models of reasoning might be emulated by computers. Experience gained from development and application of knowledge based systems in health care services proved that implentation of the revealed principles are feasible. At least in case of rame-like knowledge representation.
In this sense it may of value to find out your position regarding the two sided knowledge ‘coin’, or in other words to knowledge base situated on ‘both indicated legs’. The actual issue is wheter the procedural component should not deserve appropriate attention in context of archetype systems. From this perspective it might be of significance to assess intensely understood existing well elaborated and efficient inference (procedural) mechanisms known from frame based knowledge systems. My knowledge of the ADL is extremely limited, therefore I do not know whether it provides adequate non trivial means for design inevitable scope for diverse functionalities, i.e. efficient operations on and with archetypes. It would be worth to clarify this matter.
May be the following sample issues might illustrate my concern.
· Semantic interoperability belongs to very important issues being currently in center of attention. The role of standard archetypes as formal representation of concepts (mental entities) has paramount significance. Their core components - the attribute-value pairs – and diverse archetypes (concepts) relations represent and convey meaning. They might be domain and contextually dependent, even influenced by situation under which are considered. An archetype might be referenced or searched for in several ways, mostly by its name (a term). However also by presence of attribute names (terms) as archetype components or even by a specific attribute values. Evidently, search and matching processes are procedural phenomena either simple or sophistikated (’intelligent’) ones. The archetype components and relations are parameters for respective processes for example to identify that diverse natural language expressions convey the same meaning, that irespective of dissimilar terms might serve the same purpose. Procedures are inevitable for making this happen.
· Corectness, completeness, and consistency check of accumulated clinical data is another important issue which in general might be closely related to the previous issue. No doubt, procedurality is inevitable.
· Records harmonisation (‘federation’ seems a better term) belongs to the same category.
· Contextual assessment of attained data and their interpretation based on medical (cognitive) reasoning - actually the most complex clinical issue to be computerized – which also belongs to the same family characterized by the most complex inference processes, including also employment of qualitative evaluation methods to assess certainity of the inferred outcomes.
These samples ought to certify that the procedural knowledge complement is inevitable. Our experience proved that without it one can hardly expect that a bit experienced physicians would accept systems based only on one ‘leg’, particularly the declarative one. Designig the other ‘leg’ should avoid simplification. This means that that respective procedural software means should be based on understanding reasoning scripts of proficiant medical doctors providing health care services. This would yield such concepts as syllogism, association, gating, complementing, deduction, focusing attention, non monotonic forward and backward chaning, i.e. backtracking reasoning, preferences/priorities, task decompositions, partial outcome syntheses etc., as it is natural in fluidness of mind.
It would be very much welcome to learn whether such cognitive techniques can be implemented in frame of - for me not yet known
· Clinical Knowledge Manager (CKM) – online collaborative management space for archetypes, templates and other enterprise artefacts for concepts representation,
· version of ADL/AOM, the accepted Clinical Information Modelling Initiative (CIMI) as the formalism of choice for internationally shareable health information models
· any other mechanism.
If as yet means of such mechanisms are not available, are they at all considered?
Dear Thomas,
I hope you will pardon me for the risen queries. I hope also that you would in your talks at least pay som short attention also to the considerered declarative-procedural complex.
Kindest regards,
Mike Popper