Title

Resources for a Christian worldview

Basden

 

A Different Way of Approaching Data Models and KR Ontologies

Andrew Basden.

This article suggests a way to employ philosophical ontology in the generation of knowledge representation ontology; it also introduces a little-known philosophy which is admirably suited to this task. A knowledge representation ontology (KRO) is a belief about what basic constructs are necessary and sufficient to offer to software designers and others who must represent knowledge to create working information systems.

The KRO of the object-oriented (OO) data model (OODM) includes among other things, objects, attributes, quantitative and qualitative values to fill those attributes, methods (procedures), messages and a limited range of types of relationship. That of the relational data model (RDM) includes attributes and values, the ability to define tuples of such attributes and a few mathematical operations on sets of these, but it leaves it to the designer to use this to represent objects, relationships and so on.

For each we may ask questions like: Why were these particular components chosen as the basis for representing knowledge of a domain and how effective is the choice in practice? Budgen [2003:341] notes first the immense popularity of OO approaches:

"Objects with a capital 'O' have occupied a position centre-stage in software development since the early 1980s. The word 'object' appears in many conference titles. The literature on most aspects of their form, development and use is voluminous, with the possible, and rather significant, exception of empirical studies. ... Indeed, object-oriented programming is widely accepted as an important and powerful way of creating software systems."
He then continues with an answer to the question:
"Despite all of this, the task of designing with objects remains a significant problem. Analysis and design methods have proliferated ... However, useful as all of these have been, designing with objects remains a complex cognitive task."

This article explores why this is so and then makes a proposal of a new kind. It first sets the issue in an unusual context, not that of the professional software developer, but that of 'the people' who might use information technology in their everyday life and work. This highlights the importance of appropriateness, with the brief observation (rather than watertight argument) that most DMs are inappropriate, including that proposed by Wand and Weber [1995]. Its philosophical underpinning is found wanting. A little-known philosophy is introduced that is explicitly geared to everyday experience, and which enables a strategic proposal to be made for how to approach the design of KROs and DMs. A practical proposal is then made for a KRO that might be more appropriate to the task of KR, especially for 'KR to the people'.

'KR to the People'

Looking back over the experience of the 1980s, the decade in which knowledge representation (KR) had become mature as a discipline, Brachman [1990] suggested a number of likely scenarios and which issues would particularly need research in the future. Many are still relevant today, especially one that stood out as relating KR to the wider context of IS development: 'KR to the people'.

"It is likely that by the millenium 'knowledge systems' will be a common commercial concept. This has important implications for the future of KR. Among other things, KR components will increasingly find themselves in the hands of non-experts, raising a novel set of issues" [Brachman, 1990,p.1090].

His vision is that ordinary people (e.g. managers) would undertake their own KR rather than hiring expert system developers. The need for 'KR to the people' is clear in knowledge management, where entry of complex knowledge (as opposed to simple data) should be undertaken by 'the people' - should be, but usually isn't.

But what makes a KRO or data model suitable for 'the people'? Basden [1993] made an initial proposal that there are four 'aspects of knowledge' - items and relationships, quantitative and qualitative values, spatial things and change (events and processes) - which, in the general application, it is necessary to represent. Many business applications involve a mix of these. He argued that these aspects of knowledge are irreducible to each other, such that, if an attempt is made to represent one aspect in terms of others, this will lead to difficulty in KR and an increase in errors. But he relied solely on intuition and on observation of errors that tended to occur, and had no foundation in philosophical ontology. So he had no basis for determining whether text, for example, should be represented by these four aspects or is a separate aspect of its own.

Wand and Weber [1995] did attempt to ground KRO in philosophical ontology. They chose the philosophical ontology of Mario Bunge [1977, 1979] because "it deals directly with concepts relevant to the information systems and computer science domains" and "Bunge's ontology is better developed and better formalized than any others we have encountered." [Wand and Weber, 1995,p.209] The following grammatical constructs of their KRO map to Bunge's ontological constructs:

  • things, properties, states (stable and unstable), events (external and internal, well- and poorly-defined), transformations, histories, couplings, systems, classes and kinds,
  • laws that pertain to these: state laws, (lawful) conceivable state spaces, (lawful) event spaces, lawful transformations,
  • and, related to system: system compositions, system environments, system structure, subsystems, system decompositions, level structures.

The Wand-Weber data model ((WWDM) has wider coverage, and goes further towards appropriateness, than other extant KROs do Wand and Weber and show how fruitful it can be in critiquing the entity-relationship data model. But it does not go far enough for 'KR to the people'.

Appropriateness

If Brachman's vision is to be realized, the data model must be 'appropriate' to the lifeworld of 'the people' - the background stock of knowledge which we share and intersubjectively understand. Basden and Klein [2008] discuss the relevance of the philosophical notion of lifeworld to data and knowledge engineering and natural language processing. Two characteristics are of particular interest here. One is that much lifeworld knowledge is taken for granted. This implies that a KRO should be usable without complex mental exertion. What is primitive to the lifeworld of 'the people' should be directly representable by a primitive construct in the data model; representation of what is not primitive in the lifeworld can be assembled from the primitive constructs.

The other is that the lifeworld exhibits a wide diversity of meaningfulness to us, with many aspects that are irreducibly distinct from each other. This implies that the KRO should incorporate sets of constructs for each aspect likely to be encountered by 'the people' in their role as system developers. But if an aspect is not available in the KRO then a number of problems can be expected, which make it difficult for 'the people' - unnaturalness, hidden pitfalls, specialised knowledge, oversimplifications, all likely to generate errors [Basden, 2008:274-5]. Such a KRO would be 'inappropriate'.

These problems may be illustrated by considering the following application. A species of predator lives in woodland but at night ventures out a certain distance into the surrounding land, where it robs the nests of a species of ground-nesting birds. We wish to find out the region in which nests are in danger. To do this, we need to represent the boundary of each piece of woodland, such as that shown in Figure 1a, and then extend this by the predator distance (Figure 1b). To our everyday, lifeworld spatial intuition, this would seem quite straightforward.

Figure 1. (a) Complex shape (b) Extended

Typically we would represent the points along the boundaries as lists of (x,y) coordinates - as tuples in RDM, as objects with a couple of attributes in OODM, and as things with properties in WWDM. This is reasonably straightforward, at least in cultures that are used to coordinates, and is germane to the application, but all that follows is, arguably, extra complexity which 'the people' should not have to handle.

To associate each point with its boundary requires a second table (RDM) or class (OODM) for the boundaries as such, and the addition of primary and foreign key attributes, and in WWDM requires the use of couplings. Moreover, we must ensure that order among the coordinates is preserved and managed (e.g. when points are inserted or removed), which involves the addition and management of an artificial order attribute when using RDM and the creation of some kind of linked list in OODM and WWDM. Already 'the people' begin to find difficulty which is not germane to the application. It gets worse because of hidden pitfalls, such as that the last point must link back to the first.

To expand the boundary requires complex trigonometric calculations which 'the people' cannot be expected to know. From these calculations a new set of points must be generated, but the new points do not correspond directly with the original ones, since new points must be created while other points must be obliterated. Then 'the people' must find and implement the complex algorithm for deciding whether any given position (e.g. of a nest) is inside or outside this new boundary.

It gets worse, because what has been described so far is an oversimplification of shape-handling of which most people - including experienced system developers - might not be aware. Some shapes, when expanded enough, can generate shapes with holes; see Figure 2. The basic strategy of representing a boundary by a list of points must be rethought, to allow for an indeterminate number of such lists, one per hole, along with algorithms to manage the distribution of points to the correct lists.

Figure 2. Expanded further - shape with hole

Oversimplifying assumptions of this kind (a boundary can be represented as a single list) are not uncommon, and can remain undiscovered for a long time, even after the system has entered routine use. When such oversimplifications are discovered, either an ad-hoc, error-prone modification must be made which usually proves to be unsatisfactory, or the entire system must be rewritten from the ground up, which is not only expensive and disruptive but can introduce new errors into a working system.

'The people' should not be exposed to any of these problems. They are not germane to the basic lifeworld understanding but arise from trying to use an inappropriate KRO. To our spatial intuition, such things as complex shapes, expansions and holes are not advanced features but are relatively basic and primitive. So such things should be supplied as well-designed primitive constructs in the KRO right from the start. These kinds of problem (complexity, specialist knowledge, oversimplification and error-proneness) are not confined to spatially-intensive applications but are experienced in many types, including text-handling, resource planning, physical simulation, computer music, legal applications and so on.

What is needed is not just to ground a KRO in philosophical ontology but to find a philosophical ontology that is adequate to everyday lifeworld. Bunge's philosophical ontology is not.

A Critique of Bunge

Bunge attempted a very logical general ontology, but his attempt may be questioned, both in itself and as a foundation for a KR ontology. Bunge's ontology is split into 'Ontology 1: The Furniture of the World' [1977] and 'Ontology 2: A World of Systems' [1979]. 'Furniture' refers to those conceptual constructs with which we analyse the world, and Bunge suggests this includes substance, form, thing, possibility, change and spacetime. In [1979] he discusses the nature of systems and emergence as a prelude to introducing five 'systems genera': physical, chemical, biological, social, technical.

It is Bunge's analytical furniture (with the exception of spacetime) on which the WWDM is built, though translated into furniture for representation and modelling [Wyssusek, 2006]. (Wand and Weber also ignored Bunge's systems.) We may however question why Bunge chose the furniture he did. The system of thought that Bunge develops is one in which certain ontic questions "make hardly any sense" [1977:2], including "What is being?" and "What is there?" Not only do most of the questions he does allow depend on, and presuppose, at least these two, but, by ignoring the question "What is being?", he uncritically assumes a certain notion of being as thing or "autonomous entity" [1977:9]. This is problematic practically because it precludes non-entitary reality such as spatial extension, physical fields and 'stuff', and nuance and harmony in aesthetics. It is problematic philosophically in that substance and form echo the presuppositions of ancient Greece and the inclusion of spacetime alongside thing and possibility has echoes of Kantian a prioris. That both of these philosophic roots have been fundamentally criticised, for example by Dooyeweerd [1984], indicates that we should not take Bunge's choice of furniture without question.

Bunge's view of the five systems genera also has problems. While physical-chemical-biological form a series, as defined by emergence, social and technical do not, but are placed in parallel, and yet Bunge does not account for this difference in relationship between genera. It seems that something more than emergence is needed than Bunge offers to account for which levels/genera there are and for the relationship between them. Whereas most thinkers who recognise distinct levels include a psychological level (e.g. Hartmann [1935], Poli [2008]), Bunge refuses to do so, because [1979,p.247]:

"We might have distinguished a system genus between biosystems and sociosystems, namely psychosystems. We have refrained from doing so from fear of encouraging the myth of disembodied minds."

Popper and Eccles [1977] suggest not only a psychological world but also a world composed of concepts that have emerged and entered the public domain; why does Bunge not include this? 'Fear' seems an unusual reason for a philosopher to employ!

This is not the only place where Bunge falls short of his own standard of strict logicality; he also dismisses Heidegger with a sneer [1977:3]. This could be just a matter of style which should be overlooked, were it not for the fact that if we want a KRO that is appropriate to everyday experience, then we can reject, a priori, neither the possibility of minds without reference to bodies, nor the idea that being is constituted in its relation to world, as Heidegger proposed. What else, one might wonder, does Bunge prevent us from thinking about because of his prejudices? The reason for this lapse is Bunge's pre-logical adherence to dialectical materialism [Wyssusek, 2006].

Even more problematic might be the split between furniture and world itself. It implies that analysis stands above the world in a privileged position and is not seen as part of the world. This split has three consequences. A practical one is that it precludes analysing analysis itself. A second, more fundamental, consequence is that it privileges the analytical way of seeing the world over others. Adam [1998] criticises this and singles out artificial intelligence for ignoring non-logical ways of knowing, such as intuition, feeling and embodied knowing, which are important in everyday experience. The third consequence is that there are several candidates for furniture in addition to Bunge's and no sound way to choose between them from within Bunge's approach. In the RDM, logic programming, functional programming and the OODM, for example, the 'furniture' is, respectively, mathematical tuples, predicates and propositions, functions and lists, and active objects which are instances of classes. Why should Bunge's furniture have favour over these?

And yet there is something valid about the Bungean distinction between furniture and world. In analysing a forest, for example, though we might stand among the trees, the trees are not our thoughts. In representing a forest, we use words about trees, not trees themselves. We need to understand more clearly the relationship between furniture and world, as well as find a rather sounder approach to recognising systems than Bunge offers.

A New Foundation

The mid-twentieth Dutch thinker, Herman Dooyeweerd (1894-1977) was unusual in his approach to philosophy. Instead of first propounding a philosophical theory (as for example Bunge did) he sought to "clear away that which keeps us from seeing the structure of reality" [Skillen, 2007]. Dooyeweerd [1984] made a deep critique of theoretical, analytical, scientific, philosophical thought, but he showed how this need not be rejected as the anti-rationalists do, but can be related to everyday experience in fruitful and philosophically sound ways. A very brief summary of the relevant parts of his thought is presented here; for a longer introduction see chapters II and III in Basden [2008].

Bunge, like most thinkers, presupposed the possibility of taking a theoretical attitude, by which to understand the world. By contrast, Dooyeweerd made the theoretical attitude itself a philosophical problem, beginning instead with the everyday (naïve, pre-theoretical) attitude and reflecting that everyday experience presents us with distinct aspects among which there is a curious coherence. For example, of things like a factory, a book or a journey to Mars, there is a physical, a logical and a juridical aspect which cannot be explained in terms of each other, and yet physical causality is strongly echoed in the logical antecedent-consequence relation and in the juridical relation between crime and punishment. Dooyeweerd actually identified at least 15 such aspects (see Table 1, which gives his suite of aspects along with the types of things they make meaningful).

Table 1. Dooyeweerd's suite of aspects, with their kernel meanings
Aspect Kernel meaning
Quantitative Discrete amount
Spatial Continuous extension
Kinematic Smooth movement
Physical Energy, mass, etc.
Organic (Biotic) Life functions, integrity of organism
Psychic Sensing, feeling, emotion
Analytic Making distinctions, conceptualizing
Formative Formative power: shaping, achieving and creativity resulting in deliberate structure and process, in history and technology
Lingual Symbolic signification, representation
Social Social interaction, institutions, respect
Economic Frugality, managing limited resources
Aesthetic Harmony, enjoyment, fun
Juridical "To each their due"
Ethical Self-giving love
Faith Vision, faith, committing

On the basis of a lengthy critique of 2,500 years of theoretical thought (which does not concern us here but includes ancient Greek and modern Kantian thought especially) he argued that the logical aspect (also called analytical) has no privileged position 'above' the other aspects but is just one among many. It is no more nor no less important than others. The historical importance of the logical aspect in science and philosophy is acknowledged not by elevating it above the others, but rather by differentiating the roles it can play in our functioning as human beings.

This is explained most clearly in Clouser [2005]. In the everyday attitude of thought we are open to all aspects; for example, we might function logically as well as socially and ethically in any situation, such as building or working in a factory, writing a book, or arranging or undertaking a journey to Mars. In reflective types of activity, the analytical-logical aspect comes to the fore, but is still integrated with all the other aspects; for example, we might analyse (reflect on) the social aspect of the situation, or the physical or legal, etc. But in theoretical thought, the analytical aspect 'stands over against' an aspect of the situation (Dooyeweerd called this a Gegenstand relationship). In analytical Gegenstand we effectively isolate the target aspect, abstracting it from the situation and, indeed, all situations; for example, we might study sociality or physicality or legality as such. Dooyeweerd argued that analytical Gegenstand is inescapable in theoretical-scientific thought, which is what Bunge was doing when he compiled a 'scientific ontology'.

To Dooyeweerd it was important that the analytical aspect is seen as one among the aspects of the world rather than set apart from them. It is rather by our forming of a Gegenstand relationship that we set the analytical aspect over against other aspects of the world when undertaking theoretical thinking. Dooyeweerd's approach overcomes the problems mentioned above. It is possible to analyse analysis - the analytic aspect can be Gegenstand to itself. It allows for many ways of seeing the world beyond the theoretical (see Basden [2008:92-97]). And it provides a basis for deciding which type of furniture to use in preference to others. In theoretical thought it is the analytical which stands over against others, but in other human activity it could be another aspect that does this, for example, the lingual aspect when undertaking representation. The type of 'furniture' we use depends on which aspect stands over against others. As Wyssusek [2006] points out, Bunge's furniture was that of analysis, the paraphernalia we use to think about the world, while Wand and Weber's 'furniture' was that of modelling, of representation, the paraphernalia we use to represent the world. Whereas Bunge was involved in analytical Gegenstand, we are involved in lingual Gegenstand.

Since the lingual aspect, like the analytical, is not to be treated as separated from, or 'above', the others, but is part of the world, this means that it is possible to model models or represent representation (which is important, for example, in meta-models as well as in representing text itself).

Towards a KR Ontology

With these ideas we may return to our attempt to find a KRO appropriate to 'the people'. The proposal made here is that to design a general KRO requires consideration both of the lingual aspect 'furniture' and of all its possible Gegenstand aspects - and this includes every aspect. That is, we should, for every aspect, provide a set of language constructs that express basic things meaningful in that aspect. This is exactly the intuitive requirement that we noted earlier, to supply primitives for each 'aspect of knowledge', but it is now philosophically grounded. (Note: 'Language construct' is not restricted to text but can include graphics, animation, sound and even gestures.)

Dooyeweerd enables us to begin to answer the question we posed of which 'aspects of knowledge' we should cater for, because he delineated fifteen aspects as shown in Table 1 (somewhat in contrast to Bunge, Dooyeweerd held his list to be in need of critique and refinement). None of these aspects can be reduced to any other.

Almost every application involves several, if not many, aspects. For example, the Elsie expert system advises quantity surveyors on what budget to set for new office developments [Brandon, et al., 1995]. Within its knowledge base is represented at least the following aspects:

  • Quantitative Aspect, as in number of storey, number of staff designed for
  • Spatial Aspect, as in shape and extent of building, site area and gross floor area
  • Physical Aspect, as in ground conditions and whether special foundations are required
  • Organic (Biotic) Aspect, as in designing for health of occupants, e.g. air flow
  • Psychic Aspect, as in quality of fittings and furnishings
  • Formative Aspect, as in structure of building, its flexibility for future modification for different purposes, functional quality, suspended ceilings and raised floors for installation of cables, and much more.
  • Social Aspect, as in circulation space,
  • Economic Aspect, as in space limitations, cost of materials, total estimated cost of building, contingency cost
  • Aesthetic Aspect, as in aesthetic quality level
  • Juridical Aspect, as in adhering to building regulations, supplying lifts for disabled people.
  • Faith Aspect, as in whether a prestige building such as a head office of large corporation.

The remaining aspects are exhibited more in the way Elsie works and related to its user:

  • Analytic Aspect, as in clear presentation of results, and clarification to user of what each question means
  • Lingual Aspect, as in the wording of questions and results put to the user
  • Ethical Aspect, as in a 'generous' program which gives to the user more facilities than were strictly necessary

For a fuller discussion of aspects of Elsie, see Basden [2008].

All these, and more, had to be represented in a single knowledge base, such that all worked together in a coherent manner. The challenge to the KRO used for building systems like Elsie is to provide good means of representing knowledge of each aspect.

Comparison

In the light of this, both Basden's [1993] set of aspects of knowledge and Bunge's [1979] world of systems appear as rather ill-formed subsets, though not without some merit.

Basden's 'items and relationships' is now seen to be analytical aspect (items) and formative (relationships), quantitative values are of the quantitative aspect, spatiality is the spatial aspect, and his suggestion of a fifth aspect of text is the lingual. His aspect of change does not correspond with any one Dooyeweerdian aspect because in every aspect there is change. Change is properly addressed under Dooyeweerd's theory of time (which is not considered here).

With Bunge's systems genera the correspondence deserves closer inspection. As a mode of being each Dooyeweerdian aspect is able to define a distinct type of system. Concerning the correspondences between genera and aspects:

  • The physical, biological, technical and social genera are almost identical with Dooyeweerd's physical, biotic, formative and social aspects.
  • A minor difference is that Bunge's physical and chemical are combined in Dooyeweerd.
  • Bunge jumps straight from the biological to the technical-social genera; in Dooyeweerd two aspects lie between the biotic and the technical (formative): the psychic and analytic aspects.
  • By including the psychic aspect, Dooyeweerd allows for psychosystems, which Bunge refuses to do. How Dooyeweerd overcomes the possibility of disembodied minds is explained below.
  • By including the analytic aspect among all the others, Dooyeweerd refuses to give any privileged place to rationality; analysis is not intrinsically 'above' the world but is part of it.
  • In Dooyeweerd there is a lingual aspect between the formative and the social, while Bunge does not recognise its equivalent system, the symbolic. This enables Dooyeweerd to recognise the unique properties of symbol systems.
  • Dooyeweerd has a number of post-social aspects, which are completely absent from Bunge.
  • Bunge included space-time in his furniture (following Kant). In so doing he confused foundational aspects of the world with furniture for thinking about the world.

Concerning inter-aspect/genus relationships:

  • In Bunge the social and technical genera are in parallel, without justification or explanation, whereas in Dooyeweerd the formative aspect precedes the social on the grounds that social functioning cannot operate without formation (of, for example, social institutions or of speech).
  • Whereas Bunge accounted for difference between types of system by a doctrine of emergence which, we saw earlier, is insufficient, Dooyeweerd held that types of thing are always aligned to what he called qualifying aspects, though all other aspects are also involved in the thing. Dooyeweerd recognises fifteen radical types of system to Bunge's five, and can also distinguish genotypes and phenotypes [1984, Vol III].
  • Dooyeweerd was able to recognise the psychic aspect (psychosystems), without incurring the possibility of disembodied minds, because the psychic aspect depends foundationally on the biotic and physical even though it is irreducible to them. Dooyeweerd held a notion of inter-aspect dependence which does not contradict inter-aspect irreducibility. But Bunge conflated dependency and reducibility, perhaps because of his doctrine of emergence, and so any recognition of a psychic aspect that is irreducible to the physical would imply independence from the physical and hence disembodied minds.

Thus, I submit, Dooyeweerd offers a sounder philosophical foundation than does Bunge (and indeed Basden [1993]) for understanding the nature of analysis, 'furniture', world, diversity and their relationship to everyday experience.

A Proposal for Aspectual KRO

We may employ a study of aspects like that undertaken by Dooyeweerd, along with a notion of lingual Gegenstand, to formulate a KRO. Representation involves both sides of the Gegenstand between the lingual aspect and each of its target aspects. For each aspect, the lingual 'furniture' would include symbols to express anything that can be meaningful within the aspect. Unlike Bunge, who discussed furniture first, we look first at the 'worlds' (as Bunge called them) or spheres (as Dooyeweerd also called aspects) that need to be represented.

The Target Spheres of Meaning to be Represented

To Dooyeweerd, aspects are spheres of meaning and law, which enable all reality to Be and Occur. The kind of content we might expect for each aspect includes all that aspects enable. As explained in chapter III of Basden [2008], they enable, and thus provide a foundation for, at least the following:

  • The spheres are distinct modes of being (thing, stuff).
  • The spheres are distinct basic kinds of property.
  • The spheres are distinct types of rationality or inference.
  • The spheres define distinct ways of relating.
  • The spheres are distinct types of law or meaningful constraint.
  • The spheres define distinct ways of functioning.

The world demands appropriate representation, so each aspect deserves equal respect in a KRO. The following concrete proposal, discussed in greater detail in Basden [2008], goes some way to achieving this. It is by no means complete, but rather illustrative of what this approach would suggest needs to be included as primitives in a KRO that covers all aspects.

Quantitative Aspect (discrete amount)

  • Being: integers, ratios, fractions, proportions, etc.; also types that anticipate later aspects such as 'real numbers' for the spatial aspect
  • Properties: accuracy, approximation
  • Inferences: arithmetic
  • Relatings: greater and less than, sets, etc.
  • Constraints: e.g. a given quantity remains that quantity until changed
  • Actions: incrementing, scaling, statistical functions, etc.
Spatial Aspect (Continuous extension)
  • Being: space itself, shapes, lines (straight or curved), areas, regions, dimensional axes, etc.
  • Properties: size, orientation, distance, side (in, out, left, right), etc.
  • Inferences: those of geometry and topology
  • Relatings: spatial alignments and arrangements, touching, crossing, overlapping, surrounding, topology, etc.
  • Constraints: e.g. boundaries should not have gaps
  • Actions: join, split, stretch, deform, rotate, overlap, expand, etc.
Kinematic Aspect (smooth movement)
  • Being: movement, path, flow, centre of rotation, etc.
  • Properties: velocity, speed, direction, divergence, curl, duration of movement
  • Inferences: e.g. s = v times t -- those often found in the field of mechanics
  • Relatings: faster/slower, forward/back, travel together, etc.
  • Constraints: the Hare does beat the Tortoise
  • Actions: start, stop, rotate, follow a path, etc.
Physical Aspect (Energy, mass, etc.)
  • Being: waves, particles, forces, fields, causality, impacts; also mechanical Being, chemicals, solutions, liquids, gases, crystals, materials, etc.
  • Properties: mass, energy, charge, frequency, force, field strength, Newton-power, etc.
  • Inferences: various energy functions, etc.
  • Relatings: causes, attracts/repels, etc.
  • Constraints: conservation of mass / energy / momentum, laws of thermodynamics, etc.
  • Actions: physical interaction, expanding a field by inverse square law, dissolving, chemical reacting, etc.
Organic (Biotic) Aspect (Integrity of organism)
  • Being: organism, organ, system boundary, tissue, food, life, population, environment, dysfunctions; checksums, etc.
  • Properties: health, stamina, age, etc. (c.f. the 'stats' in role playing games)
  • Inferences: e.g. parent implies child
  • Relatings: parent/child/mate, food chains, symbiosis, system-environment, etc.
  • Constraints: need for sustenance and benign environment, etc.
  • Actions: regulate, grow, ingest, excrete, reproduce, repair, die, etc.
Psychic Aspect (Sensing, feeling)
  • Being: signals (sounds, sights, etc.), channels, states (esp. emotional), memories, motor actions, etc.
  • Properties: colour (hue, saturation, value), pitch, volume, etc.; angry, happy, etc.
  • Inferences:
  • Relatings: e.g. stimulus-response
  • Constraints: sensitivity ranges of sense organs, etc.
  • Actions: respond, remember, forget, feel, push, etc.
Analytic Aspect (Distinction)
  • Being: distinct concepts, objects, labels to identify things, etc.
  • Properties: truth values, difference and sameness, etc.
  • Inferences: those of logic, etc.
  • Relatings: contradiction, logical entailment, identity, etc.
  • Constraints: e.g. principle of non-contradiction, entity integrity (as in relational databases)
  • Actions: e.g. distinguish, deduce
Formative Aspect (Formative power)
  • Being: structure, relationships, modifications, plans, means and ends, goals, intentions, power, etc.
  • Properties: feasibility, efficacy, version, strength (as of a relationship), etc.
  • Inferences: graph searching, synthesis activity, etc.
  • Relatings: means and ends, the purpose of something, sequence of operations (history), part-whole, etc.
  • Constraints: e.g. referential integrity
  • Actions: form, compose, relate, revise, undo, seek, effect a meaningful change (change a state), plan, etc.
Lingual Aspect (symbolic signification)
  • Being: nouns, verbs, etc.; words, clauses, sentences, etc.; bullet lists, headings, cross references, quotations, etc.; word roots, languages
  • Properties: tense, case, emphasis, cultural connotation, etc.
  • Inferences: those of syntax, semantics, etc.
  • Relatings: synonyms, antonyms, opposites, cross references, rhymes, thesaurus relationships, etc.
  • Constraints: spelling, grammar, pragmatic context, etc.
  • Actions: write, draw, understand, send message, text search, find equivalent meaning, translate, etc.
Social Aspect (social interaction, institutions, keeping company)
  • Being: person, group, role, institution, title, name, nickname, etc.
  • Properties: status, leadership, formality and informality, address (postal, phone, email), etc.
  • Inferences: e.g. how to address someone
  • Relatings: friendship, acquaintance, respect for, membership, organizational structure, hierarchies, etc.
  • Constraints:
  • Actions: communicate, befriend, adopt a role, give respect, etc.
Economic Aspect (Frugality, limited resources, managing)
  • Being: resource, limit (complex), supplier, consumer, exchange, market, human resources, etc.
  • Properties: limits, prices (values), etc.
  • Inferences: e.g. management forecasting
  • Relatings: supplier-consumer, relationship with resource limits, inter-currency, etc.
  • Constraints: e.g. no net loss of resources except via defined inputs and outputs (hence double-entry book-keeping)
  • Actions: distribute resources, allocate price, etc.
Aesthetic Aspect (Harmony, enjoyment)
  • Being: nuances, harmonies, surprises, humour, fun, leisure, sport, etc. plus all the beings found in the various arts
  • Properties: situatedness, harmony, surprisingness, paradox, interesting/boring, etc.
  • Inferences:
  • Relatings: nuance, echoing, counterpoint/complementarity, etc.
  • Constraints: "Less is more in art" [C.S. Lewis]
  • Actions: harmonize e.g. music, play with, etc.
Juridical Aspect ("to each their due")
  • Being: dues, responsibilities, rights, coded laws, policies, contracts, security measures, owners, policies, (in)justice, etc.
  • Properties: security ratings, equity, proportionality, appropriateness, etc.
  • Inferences: e.g. consider evidence
  • Relatings: retribution, ownership, etc. inter-clause links
  • Constraints: laws of land, idea of what is due to each type of thing, ensure consistency, etc.
  • Actions: make contract, decide the essence of a case, judge, make retribution or recompense, etc.
Ethical Aspect (Self-giving love)
  • Being: attitudes, gifts, sacrifices, etc.
  • Properties: generosity, etc.
  • Inferences:
  • Relatings: Buber's I-Thou relationship, marriage/troth, etc.
  • Constraints: self-giving must be genuine, not for gain, etc.
  • Actions: give (without expectation of reward), forgive, etc.
Faith Aspect (vision, faith, committing)
  • Being: commitments, beliefs, trust, creeds, rituals, etc.
  • Properties: degree of certainty, trustworthiness, etc.
  • Inferences:
  • Relatings: committed-to, believe-in, trust, etc.
  • Constraints: commitments should be kept, etc.
  • Actions: make a commitment (after, maybe, weighing up the evidence), trust, worship, etc.

With a few gaps still to be filled, that provides a host of reasonably primitive elements for each sphere of meaning of everyday experience, each of which can serve as a Gegenstand to the lingual aspect in representation. However, the above list should not be treated as definitive, because it requires considerable debate and development.

The Lingual Furniture

What about the style of lingual 'furniture' in which these are to be expressed? Whereas most KROs assume a text-based 'language' in which to express knowledge, there is, according to Dooyeweerd, no need to be restricted to text because symbolic signification can occur via any medium, including graphics, animation, sound, music, the haptic channel, gestures and more. In adopting this approach we are merging issues of KR language with those of human-computer interaction.

Dooyeweerd discussed analogical relationships between aspects [Basden, 2008:72,287], such as that between physical, analytic and juridical aspects mentioned earlier (causality, logical consequence and retribution). This may be what Gibson [1977] had discovered when he formed his notion of affordance, which Greeno [1994] employed in designing human-computer interaction. In Table 2 a few examples of input and output furniture (human-computer interactions) are given which may be appropriate for each aspect.

Table 2. Lingual 'furniture' for each aspect
Aspect Output Input
Quantitative Digits, Bar length, Contour lines Hit keys, Drag to lengthen a bar,
Spatial Shapes, spatial arrangements Drag to draw or modify shape, position or size,
Kinematic Animation Joystick/keys to give direction, speed,
Physical 3D ray-traced perspective view Haptic devices,
Organic (Biotic) Fractal 3D views 'Soft' haptic device,
Psychic Colour, sound Linear sliders (e.g. HSV for colour),
Analytic Icons, Menus, Tick boxes Click to select,
Formative Box-and-arrows graph; Buttons Drag boxes, arrows; Click to activate,
Lingual Text (written or spoken), diagrams Alpha-numeric characters (keyboard),
Social Organisation charts, etc. As analytical+lingual,
Economic e.g. Tables of figures As analytic+quantitative,
Aesthetic Decoration + accompanying music As psychic with fine control, e.g. Colour circle device,
Juridical Text with cross references As lingual,
Ethical As lingual+aesthetic As lingual+aesthetic,
Faith As lingual+aesthetic, As lingual+aesthetic, but also the action itself by which the user makes a commitment,

Both these lists are to be seen as very tentative initial suggestions.

Implementation

The question immediately arises, however: how should we implement this?

On Attempting OO Implementation

One common answer is to use an OO language to provide classes and methods for each aspect, perhaps assisted by the design patterns approach [Gamma et al, 1995]. Because OO assumes that everything can be broken down into active objects, however, there are problems.

Things that are naturally objects would be easy to implement while things that are not would be more difficult to implement, with a consequent temptation to omit the latter. For example, in a spatial module simple geometric shapes would be offered, but not facilities for complex shapes with holes and boundary-extension (Fignues 1, 2). It would be left to 'the people' to put such 'complex' things together, even though, from a spatial viewpoint, they are intuitive (and thus primitive) rather than complex. It is OO that has distorted our view into seeing these as complex. The only aspects where OO would not unduly distort are the analytic and formative.

Second, it would lead to inefficiencies and inelegance, and would proliferate ad-hoc facilities, the raison d'etre of which is solely to overcome them. For example, it might seen appropriate to offer the class, textObject, sporting attributes like font, size, style, etc. But, as [Gamma et al., 1995,p.195] point out, this leads to gross inefficiency when handling individual letters. In order to overcome this, they introduce a Flyweight design pattern, by which letters and indeed whole swathes of text are stripped of their normal attributes and treated simply as pure characters - ironically, returning us to the character strings offered by more primitive languages like C. The inefficiency arises, and the kluge is made necessary, because text (lingual aspect) should not be treated as a collection of objects (analytic and formative aspects).

The fundamental reason that OO should not be used to implement the aspectual KRO is because it privileges certain aspects and calls us to execute reduction of other aspects to these. The same reason could also be adduced against using any current data model, such as the relational (which privileges the quantitative aspect).

Direct Implementation

Instead, it is preferable to see implementation as forming a bridge from the various aspectual primitives to the underlying architecture of the computer which, at present, involves bit patterns, bit-pattern manipulations, addressed memory and signals. In principle, each aspectual module's primitives may be implemented by a bridge to the bit level, a different bridge for each aspect. For example, following bit-level implementation practice over the past four decades:
  • Each aspectual being could be implemented as an allocated chunk of memory.
  • Each aspectual property could be implemented as a bit pattern held within such chunks; for each aspectual property a different coding would be used, such as ASCII for lingual information, binary for quantitative and bitmaps for spatial. Funt [1980] argued for the latter within artificial intelligence.
  • Each aspectual relationship could be implemented by some means of keeping such chunks together, the main methods being contiguous memory address or address pointers. (Of course, pointers need not be made available to the KRO user.)
  • Each aspectual function could be implemented by a set of bit manipulations, such as arithmetic operations for the quantitative aspect and concatenation of bytes for the lingual.
  • Each aspectual inference and constraint could be implemented by bit manipulations and bit-level tests. Funt [1980] argued for spatial inference within artificial intelligence implemented as bitwise AND and OR between bitmaps.

Some examples of such aspectual facilities, and how they might be (and indeed have been) implemented at the bit level, are shown in Table 3.

Table 3. Some current bit-level implementation of aspectual properties
Aspect Aspectual facility Bit-level implementation
Quantitative Integer
Proportion, Probability
'Real' number
Ratio
Binary, Binary-coded decimal
Fixed-point binary (FPB)
Floating point (FP)
Two integers in contiguous locations
Spatial
(given for 2-D)
2-D field (e.g. Funt)
(x,y) coordinate
Direction, Angle
Bitmap (grid of bits)
Two contiguous FPs
Circular FPB (where 11...11 is next to 00...00)
Kinematic Route
Movement
Sequence of pointer-linked bitmaps
Pixel-colour alterations for actual movement
Physical Bulk material in 3-D
Collision detection in VR
Arrays of FP
Bitmap ANDing (Amiga)
Organic Integrity or organism (record) Checksums
Psychic Colour
Scene
Sound (waveform)
Triple FPB for RGB
2-D array of colour cells
1-D array of FPBs
Analytic Distinction
Truth value
Bit-pattern difference
Single bit value 1,0
Formative Structure
Relationship
Activity of shaping
Contiguous memory cells
Pointer linked list
Manipulation of pointers, cells
Lingual Text character
Words, sentences
Emphasis (bold, etc.)
ASCII, EBCDIC, Unicode
1-D array bytes
Style flags per character,
escape sequences (ANSI)
Social URL / email address
Names
Four FPBs
Hash tables
Economic Data compression
Double-entry bookkeeping
Zip etc. coding
Two contiguous FPBs
Aesthetic
Juridical Data protection Encryption
Ethical
Faith

Since later aspects depend foundationally on earlier ones (Dooyeweerd's notion of inter-aspect dependency is not elaborated here; see Basden [2008:71]), some later-aspect elements can be implemented at least partially using those from earlier aspects: for example, double-entry book keeping (economic aspect) can be implemented as two fixed-point numbers (quantitative aspect) in contiguous memory locations.

Conclusion

This article has made four proposals for a different systematic approach to the design of KROs or data models. The first is that we should aim for 'KR to the people', that is, KROs that offer appropriate primitives for every aspect of everyday, lifeworld experience. The problems found in extant KROs were illustrated and the philosophical foundations of the Wand-Weber data model (Bunge) were examined.

This led to the second proposal, that Dooyeweerd's [1984] philosophy offers a sounder foundation for achieving 'KR to the people', because it begins with everyday experience and can clarify the nature of the relationship between 'furniture' and world as one of Gegenstand. Dooyeweerd also usefully delineated a suite of fifteen irreducibly distinct aspects, with which we can meet the challenge of creating KROs.

The third is an orientating proposal emerging from this, about how to approach the design of KROs. First consider the aspects of the world which IS developers might encounter and devise for each a set of primitives of being, property, functioning, relating, inference and constraint. Second, consider how these primitives might be expressed via language constructs in any medium or type of human-computer interaction. Third, consider carefully the direct bridge from each aspect's primitives to the bit level, employing different coding for each aspect.

Based on this approach, the fourth is a tentative, practical proposal about what needs to be implemented as primitives for each aspect. Few of the primitives proposed are esoteric, but rather are the stuff of everyday experience, in the midst of which information technology is being increasingly applied. That extant software already embodies many of these things, suggests that such a proposal is feasible in principle.

All these proposals are in need of discussion, criticism, testing and refinement, and are offered to the KR community in order that others might join this work. Actual attempts to implement this proposal (which is discussed in more detail in chapter VII of Basden [2008]) have begun, but it must be considered a long-term project.

In the light of this proposal we can perhaps understand why extant data models are often inappropriate in many applications: each offers constructs for only one or two aspects. The RDM in its pure form is restricted to the quantitative aspect (mathematical tuples) though in its workable form supports the analytical aspect (distinction). The OODM supports the analytical and formative aspects quite well (e.g. in objects, methods, polymorphism, encapsulation) and in its class hierarchies we find echoes of the biotic aspect's notion of taxonomy. The WWDM supports analytic and formative aspects well, and is arguably the best available at doing so. But none of these DMs offer direct support for any of the other aspects.

It is common to treat the representation of the other aspects as mere application. This article makes a radically different proposal that there is something basic in every aspect that should be offered directly in a KRO, and that 'application' should refer to bringing together basic constructs selected from all aspects to meet specific needs and opportunities. It contends that IS developers should have to attend only to the 'external' complexity of the domain itself and not to the complexity 'internal' to the various spheres of meaning.

Is this merely another approach to software reuse? It does address the issue of reuse, but it is rather more than that. If all these primitives are implemented and made directly available via appropriate constructs then IS developers will be able to represent the wide diversity of domain meaning (such as in the case of Elsie above) with naturalness and relative ease, and the process of KR will be more friendly to the lifeworld. This, it is suggested, will enable us to move a little closer to Brachman's [1990] call for 'KR to the people'.

References

Adam, A. (1998). Artificial knowing: Gender and the thinking machine. London: Routledge.
Basden, A. (1993). Appropriateness. In M. A. Bramer, & A. L. Macintosh (Eds.), Research and development in Expert Systems X (pp. 315-328). Cranfield, England: BHR Group.
Basden, A. (2008). Philosophical frameworks for understanding information systems. Hershey, PA: IGI Global, 2008.
Basden, A. & Klein, H.K. (2008) New Research Directions for Data and Knowledge Engineering: A Philosophy of Language Approach. Data & Knowledge Engineering, 67(2008), p.260-283.
Brachman, R. J. (1990). The future of knowledge representation. In American Association for Artificial Intelligence, AAAI-90: Proceedings of the Eighth National Conference on Artificial Intelligence (pp. 1082-1092). Boston: AAAI.
Brandon, P. S., Basden, A., Hamilton, I., & Stockley, J. (1988). Expert Systems: Strategic planning of construction projects. London: The Royal Institution of Chartered Surveyors.
Budgen, D. (2003). Software design (2nd ed.). Harlow, U.K.: Pearson Education/Addison-Wesley.
Bunge, M. (1977). Treatise on basic philosophy, Vol. 3: Ontology 1: The furniture of the world. Boston: Reidal.
Bunge, M. (1979). Treatise on basic philosophy, Vol. 4: Ontology 2: A world of systems. Boston: Reidal.
Clouser, R. (2005). The myth of religious neutrality: An essay on the hidden role of religious belief in theories (2nd ed.). Notre Dame, IN: University of Notre Dame Press.
Dooyeweerd, H. (1984). A new critique of theoretical thought (Vols. 1-4). Jordan Station, Ontario, Canada: Paideia Press. (Original work published 1953-1958)
Funt, B. V. (1980). Problem-solving with diagrammatic representations. Artificial Intelligence, 13(3), 201-230.
Gamma, E., Helm, R., Johnson, R., & Vlissides, J. (1995). Design patterns: Elements of reusable object-oriented software. Reading, MA: Addison-Wesley.
Gibson, J. J. (1977). The theory of affordances. In R. Shaw, & J. Bransford (Eds.), Perceiving, acting and knowing (pp. 67-82). Hillsdale, NJ: Erlbaum.
Greeno, J. (1994). Gibson's affordances. Psychological Review, 101, 336-342.
Hartmann, N. (1935). Zur Grundlegung der Ontologie. Berlin: W. de Gruyter.
Poli, P. (2008). Ontology: The Categorial Stance. Volume I of this Work.
Popper K.R., Eccles J.C. (1977) The Self and its Brain New York.
Skillen, J., (2007). Jim Skillen, President of the Herman Dooyeweerd Foundation, said this at a meeting in Leeds, UK, March 2007.
Wand, Y., & Weber, R. (1995). On the deep structure of information systems. Information Systems Journal, 5, 203-224.
Wyssusek, B. (2006). On ontological foundations of conceptual modelling. Scandinavian Journal of Information Systems, 18(1), 63-80.