Re: [ontolog-forum] Ontological Means for Systems Engineering

To:	"'[ontolog-forum] '" <ontolog-forum@xxxxxxxxxxxxxxxx>
From:	"Chris Partridge" <mail@xxxxxxxxxxxxxxxxxx>
Date:	Fri, 23 Jan 2009 15:37:15 -0000
Message-id:	<00b901c97d70$782ec530$688c4f90$@net>

Matthew, Ian,

I think buried in here is something quite important, and I am sure that we have discussed before.

One can roughly categorise two ways of analysing: speculative and empirical.

Exaggerating slightly. The speculative approach is an armchair approach that thinks we can get to the essence of things without leaving it. The empirical approach looks at the evidence (in my case, this is often in legacy systems).

In the speculative approach, analysts often fixate on a name – e.g. system – and try and find its essence. They assume an almost Adamic approach, assuming that some divine power has handed out the names of the objects from our ontology. Or, more prosaically, that since we use the names they (that is the type not the token) must consistently refer to something.

In the empirical approach, one looks at what the names are used to name. As Ian pointed out, it is certainly true that people use different tokens of a name in ways that turn out to give very different results – and Ian has described how these can be disambiguated. One starts from a descriptive perspective (within a top level framework), but may find that some revision is required to get an accurate result. Once enough evidence has emerged (and revision been done), then one can start trying to generalise these a level. At this stage, a general pattern that covers all uses of the term may emerge (and debate about what to call it), or it may become apparent that there are a Wittgensteinian family of senses. What is key, is that our trust of the higher level pattern is based upon the evidence from earlier use or more specific patterns – and these play an important role in fixing the sense of the general pattern. Without the earlier use, there is little or no evidence for higher level pattern. This is why the evidence gathering is so important.

My suspicion is that a lot of use of terms like system have not been submitted to sufficient empirical analysis to establish what they refer to accurately. In the IDEAS workshop there was quite a lot of evidence to support this view.

Regards,

Chris

From: ontolog-forum-bounces@xxxxxxxxxxxxxxxx [mailto:ontolog-forum-bounces@xxxxxxxxxxxxxxxx] On Behalf Of Matthew West
Sent: 23 January 2009 14:59
To: ian@xxxxxxxxxxxxxxxx; '[ontolog-forum] '
Subject: Re: [ontolog-forum] Ontological Means for Systems Engineering

Dear Ian,

Then we can add a third class, call it “System” and put that name in a NameType intentionally constructed by Matthew. The point was that the name doesn’t matter as long as we can identify the extent of the class. For example, I might decide to call the class you identified “widget”. Extensionally, it’s the same class, I just give it a different name. Because IDEAS/BORO has a sophisticated(ish) naming pattern, I know that when Canadians say “system” they are referring to a different class to the one you call “system”.

We didn’t have the option to impose a new definition of “system” on any of the parties.

[MW] That’s fine. My approach is usually to try to find the most general sense, and then qualify it to produce the more restricted senses. This has the advantage of giving you a model that shows the relationships between them.

They had their own definitions, and a tiny ontology project was going to stop the combined systems engineering departments of some the largest defence agencies in the world from doing things their own way. The same goes inside companies, I should add. This is why corporate taxonomies are all given a stiff ignoring by the users, and one of the reasons why all those “corporate data model” projects of the 1990s ended up producing something no-one ever used. Either the users don’t agree with the terminology (in which case, they dismiss it off hand), or they see a word they think they recognise and use the model/taxonomy in an unexpected manner.

[MW] Which is why with the DDM we (including you) tried hard to preserve the meaning the business gave to terms, but place them in the ontology such that they were correctly defined by the relationships they had to other entity types (especially subtype/supertype).

Regards

Matthew West

Information Junction

Tel: +44 560 302 3685

Mobile: +44 750 3385279

matthew.west@xxxxxxxxxxxxxxxxxxxxxxxxx

http://www.matthew-west.org.uk/

This email originates from Information Junction Ltd. Registered in England and Wales No. 6632177.

Registered office: 2 Brookside, Meadow Way, Letchworth Garden City, Hertfordshire, SG6 3JE.

Cheers

Ian Bailey

www.modelfutures.com

From: ontolog-forum-bounces@xxxxxxxxxxxxxxxx [mailto:ontolog-forum-bounces@xxxxxxxxxxxxxxxx] On Behalf Of Matthew West
Sent: 23 January 2009 13:26
To: ian@xxxxxxxxxxxxxxxx; '[ontolog-forum] '
Subject: Re: [ontolog-forum] Ontological Means for Systems Engineering

Dear Ian,

<snip>

The big challenge in ontology is to figure out what’s different about a system to any other physical item.

[MW] I agree

Why is a car considered a system, but a rock isn’t, for example. We had a lot of debates around this in IDEAS meetings. You can easily ascertain the extents of aircraft, armoured vehicles, etc., but the extent of the classes each nation were calling system differed. The UK framework (MODAF) saw a system as any kind of man-made object that had functionality (i.e. cars=yes, rocks and humans = no). The Canadians were very specific that a physical object only became a system when it was manned, maintained and ready to function – a narrower set than the UK one. The US (DoDAF) description was similar to the UK but also included humans/animals. So, we had three overlapping classes, each of which was called “system” by different parties. By extensional analysis, we worked out that what MODAF calls “CapabilityConfiguration” was an exact match for the Canadian “System”.

[MW] Well I don’t think any of those are wrong, but perhaps a little restrictive. The two things that characterise a system for me are:

1. It has a function/capability/purpose.

2. Has parts that can be replaced by functionally equivalent parts.