Re: REA as a universal business firstname.lastname@example.org (Fritz Lehmann)
Date: Fri, 13 Jan 95 07:20:50 CST
From: email@example.com (Fritz Lehmann)
Subject: Re: REA as a universal business model
Cc: firstname.lastname@example.org, KMCD@utrc.utc.com, email@example.com,
Bill McCarthy wrote on the Conceptual Graphs list:
Almost all of my research and teaching work has been in the area of what F.
Lehmann calls universal business models. In particular, we have been working
for more than 10 years on establishing the utility of semantic-based
descriptions of accounting or economic phenomena in business enterprises. All
of our work at present stems from a basic semantic model described in a
July 1982 paper in THE ACCOUNTING REVIEW, generally considered to be the best
and most widely available accounting journal in the world.
Good. Can you indicate to us the size and scope of
your "REA" model, with samples of some of the basic concepts? I.e.
you have "resources", but do you have "people", "obligation", "legality",
"money", "shipper", "purchase order", "accounts-receivable" or what?
. . . [O]ur work in
areas such as automated view integration facilities for business databases,
intensional reasoning with accounting or economic data structures, and basic
ontologies for economic phenomena in manufacturing environments may find an
audience in the AI forums (like Avignon), but not usually among accountants.
We have some small practitioner victories (limited use in certain IBM systems
and a separate practice unit in Price-Waterhouse consulting), but our overall
position within accounting is definitely considered to be way out on the
This is not surprising. I surmise that the intellectual fascination
of many accountants with their own field is low. Unless they see 'payoff'
they will wonder why they should listen. By the way, what is "Avignon"?
. . . [O]ur basic model is based on three primitives: economic
Resources, economic Events, and economic Agents (hence the common acronym of
REA accounting), and it seems that a lot of other people are starting to come
to grips with similar ontologies for business and non-business organizations.
For five years or more, I have felt that, as a research group, we were dumbing
down our message and spending too much time on non-basic issues in effort to
popularize matters. What we should have been doing is joining groups like the
CGers and pushing ourselves in more basic directions in other fields like
Yes, I know there are others who are going in this direction.
One fairly new field which does this is "Enterprise Integration" --
see the collection edited by Charles Petrie, MIT Press. Are you
familiar with the TOVE system at the Univ. of Toronto? It's a
business model. Also there is something called the "BIM" related
to EDI (Electronic Data Interchange); it's supposed to be a semantic
trading model. There are so many "business integration" models that
real businesses don't know which one to use. I once saw a presentation
to the St. Louis SIGART local on a system which your description
brings to mind. It was a sort of frame-based alternative to the
centuries-old double-entry approach. (It occurs to me: was that
you?) The notion of expendable resources, etc. is part of the
ARPA/Rome Planning Ontology. The new STEP standard has a rather
elaborate model of business activity with all kinds of Work Requests,
Approvals, Design Changes, etc. (This standard is supposed to be
for product descriptions, but it covers quite a bit more.)
I don't want to clutter up the list with accounting discussion unless people
think that it is a good example of issues that many CGers would be interested
in, so I would encourage replies to my own email : "firstname.lastname@example.org" Even if
people are not interested in this business side of conceptual modeling, I
would appreciate pointers to research groups who might be.
Don't hesitate to discuss this in detail on these lists;
a truly semantic approach to this will involve many issues
of more general interest. I assume you have time, change, and
of course you have "event". We in CCAT have not come up with a
good definition of an event which would distinguish it from
an object in all cases (but I think we will). A different field like
accounting might provide examples which shake us up a bit;
that's what happened when Allen Brewer pointed out to us earlier
that, at one level of analysis (government contracts), a mere
passage of a time period can constitute an "event"; I wrote
a long rejoinder explaining this in terms of people, courts,
custom, clocks, and the standard atomic clock from which the
government derives "official time" from atomic pulsations.
His abstract contractual point of view turns out to be grounded in
a deep social and legal system which is in turn physically based
(like almost everything). The EDI standards are chock full of
named (but not defined) business events, their participants, and
the documentary instruments which cause or reflect the events.
Question: What concepts and relations do almost all business
models have in common? (I'm assuming approximately corresponding
concepts, i.e. A's notion of "resource" might be a little different
from B's, but they are more or less the same, and they would agree
on whether most real-world things are "resources" or not.) By the
way, what is your definition in REA of "money"?
In CCAT we very much need a formal model of a business and
of trade between businesses. The hope is to have a generally
useful one which will not be refuted by the next researcher who
comes along. It's rules and assertions should be fairly uncontro-
versial and unassailable. The accounting-related stuff will have
to co-exist with other aspects of a business, such as its physical
existence, its status as a legal person, its manufacturing/service/
engineering capabilities, etc. The CYC project is supposed to have
released it's (stripped-down) concept taxonomy by now; I will be
interested to see if their notions of a business converge with
yours or at least are compatible.
You said REA uses resources, events and agents as "primitives";
in CCAT we would want to define these interms of "more primitive"
concepts whenever it can be done reliably. Do you have careful
(formalizable?) definitions of all three? We want to cross-link
"ontologies" and ground upper-level things in primitives. For
example, some people associated with the KQML agent-language want
to come up with a definition of an abstract 'agent' which can be
a person or just a certain kind of computer program. How does
this fit with your idea of "economic agent"?
Note: I'm sorry for not answering this and other email earlier.
I have a big backlog of very interesting messages to answer.
Yours truly, Fritz Lehmann
GRANDAI Software, 4282 Sandburg Way, Irvine, CA 92715, U.S.A.
Tel:(714)-733-0566 Fax:(714)-733-0506 email@example.com