We have been hearing more and more about KMS as an evolution of DBS and DBMS (database and database management systems). This relates to how we program agents with “knowledge”. My question is:
What is knowledge? How do we characterize tacit (versus explicit) knowledge?
I assume we are not returning to positivism and remain moderns so we are using “knowledge” as a short hand for “knowledge claim” and we prefer to associate specific claims with some theoretical approach.
Consider tacit knowledge such as how to ride a bike - it is not knowledge in the modern sense of information explicitly tied to theory and subject to refutation yet it is subject to refutation - riding a bike on a spinning asteroid might pose all sorts of unforeseen problems.
A parallel in a KMS might be knowledge that underlies behavior of all agents in an ABM versus knowledge that develops and is continuously refined by specific agents and specific types of agents. If we envisage a GIS that is modified both by agent behaviors and external processes we could imagine that general (built in to all agents) knowledge might itself become maladaptive and either have to be compensated by acquired knowledge or might even need reprogramming if the GIS changed significantly.
Parallels for a KMS dealing directly with the real world might substitute received dogma (from education or peers) for tacit knowledge and “explicitly theorized hypotheses open to refutation” (including counter-intuitive hypotheses) for the alternative. Feed back between implicit and explicit knowledge claims happens but might be better incorporated outside of a management system which by definition may have to be reworked.
To be worth caracterizing as an explicit knowledge claim it should be related to a theoretic area and describe a linkage between an input and an output. For this to be different from a database full of purported facts the theoretical areas need to be defined, even if only in abstract terms, and the specific body of theory to which knowledge claims belong should be apparent. Consider three knowledge claims sketched out in this way: (A)human nutrition/vitamin therapies; vitamin C may have value in facilitating the healing of skin abrasions, (B) urban poverty/racism in America; low employment levels of African American males in US cities is not entirely explainable by residual racism, © government/governance; power corrupts and absolute power corrupts absolutely.
AS hypotheses, the last is obviously the weakest because its claims are so strong as to be implausible, the first is sufficiently narrow as to be plausibly disprovable and the second is vague enough to be not particularly useful. Do we want to develop a way to categorize “knowledge claims” - Muslims have a well developed system for categorizing claims about the sayings and actions of Muhammed that, though it uses an idiosyncratic method (reconstruction of a chain of transmission and evaluation of the strength of each link), provides a good example of the importance of such categorization. Popper would have wanted knowledge claims to be clearly disprovable to count as such but many (including Feyerabend) have found this requirement less than persuasive. Epistemology is a deep topic and perhaps too deep for a discussion of KMS but this does not mean a KMS has to be totally naive. It probably should use some categorization (e.g. core hypothesis in many fields, foundational in a particular field, not broadly implicated in general theory) if it wishes to manage “knowledge claims” rather than “data.”
Imagine a claim such as “more than 60% of the insects on trees in Area 23 are unknown outside this area” this might be a “fact claim” or it could be turned into something of theoretical value in the ecology literature. If this theoretical significance is explicitly linked to the “fact” then it becomes something that is better off in an ecology KMS but until that linkage is made it might just as well be stored in a simple database. In short, if a KMS is to be different from a simple DB it should not just be a DB: it should, in some way, be epistemologically savy.
Would you say that agents and specifically humans in general rely on knowledge claims, or knowledge? You raise some excellent points and this is at the heart of the challenge of modeling social-ecological systems using data versus information (and more importantly what agents perceive they know)…