© Roger M Tagg 2009 revised 2010
Welcome to FROLIO – a new attempt to merge philosophy and the "semantic web" . This website is under continuing development.
This essay is about the struggle to find a common structure of meaning for the field of Information Systems, to try and reduce confusion when IT analysts brought up in different methodologies have to work together, especially on large projects. A standard ontology would, researchers argue, be preferable to a Tower of Babel of different meta-models and methodologies.
It seems to be part of its nature that the subject of Information Systems Development is dominated either by fashions (e.g. Databases, Process Modelling, Object Orientation, the Unified Modelling Language 'UML', Web Services, Agile Methods etc) or gurus (e.g. James Martin, The "Three Amigos" of Booch, Rumbaugh and Jacobson, and Michael Jackson (not the recently deceased pop icon!). There are also other popular conceptual IS models, more oriented to describing business processes; two leading names are SAP and ARIS. There have also been many good ideas that have never quite made it - at least in the information systems (IS) world, e.g. LAP (Language Action Perspective), Organizational Semiotics and Activity Theory.
Academics have attempted to address this Tower of Babel by going back to a more formal, philosophical and rigorously defined ontology that shows a semantic structure of all the things one needs to talk about when developing information systems. As well as being formal and rigorous, such a basis should not have more concepts than are absolutely necessary.
The ontology most discussed at the time of writing is the Bunge-Wand-Weber (BWW) ontology, which dates from around 1990. Wand (from Canada, University of British Columbia - UBC) and Weber (from Australia) developed it on from a more philosophical ontology developed around 1977 by Mario Bunge (originally from Argentina but now living in Canada; he taught at McGill Uni (Montréal) for many years. BWW follows some of Bunge's original ideas, but not all. The idea is that it can be used to evaluate the "grammars" of the various conceptual IS modelling methods such as those mentioned above. Are they, for example missing certain essential concepts, or are some of their concepts redundant?
Some of the concepts (objects) included in BWW are listed in the table below.
| Thing Property - In General alias Attribute - In Particular alias Attribute value for an individual - Hereditary - Emergent - Intrinsic (of a single thing only, i.e. not "Mutual") Class (of objects that have one property the same) Kind (of objects that have 2 or more properties same) Acts On Coupling - Binding Mutual Property - Non-Binding Mutual Property |
State - Stable State - Unstable State Conceivable State Space State Law - Stability Condition - Corrective Action Lawful State Space |
Transformation Lawful Transformation - Stability Condition - Corrective Action Conceivable Event Space Lawful Event Space External Event Internal Event - Well-defined Event - Poorly-Defined Event History |
System Subsystem System Decomposition Level Structure System Composition System Environment System Structure |
Relationships, such as they exist in BWW, are represented either by Functions or by Mutual Properties. The basic relationship functions are:
- A Thing possesses a Property
- A Property precedes another Property (in the sense that
possessing the first is a pre-condition for possessing the second)
- An Event marks the change between* State 1 and State 2 (* - my
words, not W&W's!)
- A Composite Property is a conjunction of Property 1, Property 2
etc (e.g. Date is a conjunction of Day, Month and Year)
- A Composite Thing is an association of Thing 1, Thing 2 etc
- A Composite Event is a composition of Event 1, Event 2 etc
There are a number of "supplementary functions" which test if, for example, one of the individual objects is part of a given Composite object, or a member of a given Class or Kind. Other than these (mainly just composition and Categorization), any relationships have to be thought of as Mutual Properties.
Various authors have commented that BWW - as presented - is not intuitive for many of the people that might use it, and some researchers have proposed "meta models" of the BWW ontology itself. An example is described in a paper by Kiwelekar and Joshi; this paper is freely available on the web.
Criticisms of BWW (apart from the above-mentioned difficulties in understanding and applying it) fall into two types:
Allen and March (from the US) mainly address the first issue. They say that Bunge's ontology "has no place for human intentions, interpretations or meaning", or what Searle (a leading philosopher from Berkeley) call "institutional reality", which includes "corporations, government agencies, money, educational institutions, contracts and transactions". They are also concerned about "rules, policies and procedures".
Susana Herrera and her Argentinian co-authors have pointed out that Bunge did (in 1993) also publish a Social ontology. Obviously that came after the BWW originally appeared. They have proposed extensions to BWW to take this on board, under the name IOMIS.
Michael Rosemann (Queensland Univ of Tech) has been involved in a number of papers relating to BWW. With Peter Green (Uni of Qld) he looked at the need for BWW to cater for multiple perspectives in IS conceptual modelling, particularly the Process-centric perspective. They tested their ideas on the Activity Based Costing component of SAP, where concepts such as direct and indirect costs, cost pools, cost allocation base and activity were involved.
With Boris Wyssusek (also QUT), Rosemann suggested going back to Bunge's original ideas and adding back in Bunge's concept of "hierarchies of systems". At the bottom of the hierarchy is the Physical system, at the top is the Socio-Technical system. This may seem an odd use of the word "hierarchy" - to me it's more like an interconnected network of different spheres.
In 2006, a debate was published in the Scandinavian Journal of Information Systems. Boris Wyssusek of Queensland University of Technology (who had worked with Rosemann at QUT) summarized a number of criticisms of the BWW Ontology. Comment papers were contributed by Guarino & Guizzardi, Krogstie, Lyytinen, Miller & Kazmierczak, Opdahl and Wand & Weber themselves. The complete set of papers is available on the web via the DBLP website at the University of Trier, Germany.
A number of other projects that I have looked at have taken a less dismissive view of relationships, and have given them more attention in their proposals.
Veda Storey (ex colleague of Wand at UBC) and colleagues at Georgia State Uni have published a paper titled "An Ontology for Classifying the Semantics of Relationships in Database Design" (web link - needs a SpringerLink account).
Herre, Heller and colleagues in the Onto-Med group in Leipzig, Germany have been developing an ontology system for application in medical informatics. They started with a proposal GOL intended for informatics and data dictionaries, which has evolved into GFO - General Foundational Ontology and GFO-Bio. It recognizes a whole range of relationships types, some of them with 3 rather than the usual 2 roles. In some cases the third role is Context (which is a standard role in Frolio).
Another group looking at ontologies for IS development is led by Guizzardi in Brazil (he was a former collaborator with the Onto-Med group above, together with Wagner from Cottbus, Germany). They have proposed a family of ontologies named UFO - Unified Foundational Ontology. UFO-A covers similar ground to BWW, while UFO-B addresses temporal relationships and UFO-C social relationships. Guizzardi was co-author (with Guarino) of one of the comment papers in the Wyssusek debate (see above).
Finally, Amit Sheth, formerly leader of a research group at the Univ of Georgia (not the State Univ), now leads a group named kno.e.sis at Wright State University in Dayton, Ohio. In 2007 he and colleagues published a number of presentations (e.g. this one) and papers, in which one of their research directions would be to recognize the so-called "Relationship Web" - an idea that overlaps a lot with FROLIO. The priority at kno.e.sis seems to be automatic extraction of relationships from material on the web. However their relationships are generally binary, i.e. between two things only; they are not at present including roles such as Authorship, Context, Theory, Tool and Process.
FROLIO is really just a partial ontology, concentrating on relationships as a major concept. Its justification arises from the view that objects by themselves do not carry much meaning; this comes only when they are involved in relationships. FROLIO however is not very complete on the "things" aspects of ontology - I have tended to assume that FROLIO would follow available ontologies that already exist, like SUMO, Cyc etc. Possibly some of the projects above would also provide a good infrastructure for FROLIO.
However in 1999 Wand and Weber published a paper describing how their ontology addresses the concept of relationships. It does not regard them as "first class concepts", on the grounds that they are too imprecise and vague in their semantics. Instead, they rely on the concept of "mutual properties".
The BWW ontology, while oriented to information systems, clearly doesn't address some of the "soft" (i.e. less formal and human-oriented) areas of information systems. These areas, however, are ones that FROLIO is specifically trying to address more fully.
Index to more of these diatribes
Some of these links may be under construction – or re-construction.
This version updated on 9th February 2010
If you have constructive suggestions or comments, please contact the author rogertag@tpg.com.au .