Re: [ontolog-forum] Constructs, primitives, terms

To:	"'[ontolog-forum] '" <ontolog-forum@xxxxxxxxxxxxxxxx>
From:	"Rich Cooper" <rich@xxxxxxxxxxxxxxxxxxxxxx>
Date:	Fri, 2 Mar 2012 14:19:21 -0800
Message-id:	<7DD0C52B00EF4A488F6867EF2E948906@Gateway>

Dear Pat,

I like your approach to considering how the errors are related to word usage. Among the documents on your web site at micra.com I found a reference to Doug Lenat's article in AI Magazine, which I traced to:

https://www.aaai.org/ojs/index.php/aimagazine/article/download/2106/1996

The quote most relevant to our current discussion is this one:

The Smurfs is a cartoon series in which a race of tiny blue elves employ the word smurf here and there, to stand in for almost any word—most often a verb—mostly for cuteness. For example, “Papa, smurf me up some supper!” In AI, we often become just as enchanted with some word we’ve made up, and we keep reusing that word to mean different things. Some examples: frames, actors, scripts, slots, facets, rules, agents, explanation, semantics (this one is particularly ironic), ontology (this would be the winner if not for semantics), and AI (okay, perhaps this is actually the winner). In some cases, we commit the converse mistake: inventing new words for old concepts. And even worse, both of these errors intermix: for example, some meanings of actors are equivalent to some meanings of agents.

It seems that Lenat is also concerned with the variability of how we each use words.

-Rich

Sincerely,

Rich Cooper

EnglishLogicKernel.com

Rich AT EnglishLogicKernel DOT com

9 4 9 \ 5 2 5 - 5 7 1 2

-----Original Message-----
From: ontolog-forum-bounces@xxxxxxxxxxxxxxxx [mailto:ontolog-forum-bounces@xxxxxxxxxxxxxxxx] On Behalf Of Patrick Cassidy
Sent: Friday, March 02, 2012 12:18 PM
To: '[ontolog-forum] '
Subject: Re: [ontolog-forum] Constructs, primitives, terms

John,

I think we agree that local usage of data *may* affect the local logical

interpretation, and therefore its intended meaning, but where that occurs, I

would say that the local conditions should be logically specified and made

part of the ontology that is used in common (i. e., a merger of local

ontologies) for interoperability.

(I try to avoid the word 'definition' so as not to imply that only

necessary and sufficient conditions are intended, but in my previous post I

did add qualifiers which, for brevity, I hoped would be understood as also

applying. In the below when I say 'define' or 'definition' it means only

the kind of logical specification (usually with necessary conditions) that

is in an ontology, but which may include some necessary and sufficient

specifications for some classes.)

[PC]

>> Does this refer to cases where two communicating systems define (or

>> otherwise logically specify the intended meaning) of the *same* term

>> (e.g. 'widget'), but use different definitions?

> That depends on what you mean by "definition". You can say that the

> official definition of a term is the statement designated by its URI.

> But each bank uses each term in its own context. Each use in each bank

> may be consistent with the common definition, but other relationships

> in each context may be different.

> As a result, a customer who switches to a different bank may get

> unpleasant surprises about the fees or constraints that one bank

> imposes that the other one didn't. Humans can adjust to such

> "surprises", but computer systems usually cannot.

> Are you going to call that additional context-dependent information

> part of the definition? . . . .

Not necessarily, though it might be in some cases. Take the case of a

check for which sufficient funds are not available: for some accounts it

might merely generate an "overdraft" fee, for others, the check might be

returned. I think people and intelligent machines can be aware of what it

means to be a "check for which there are insufficient funds" (and thereby

have a common understanding) without having a clue as to the policy of any

given bank for any given account in such circumstances. It would be a

mistake of a system to assume that one bank's policy is the same as

another's without explicit information. This would be a case where, even

with a common perfect understanding of the terms used, one still cannot

predict the actions of agents. PIFO-based Interoperability can provide a

common language and allow all systems to understand the common logical

specifications of terms, but cannot predict the actions of any given agent

in a given set of circumstances unless the policies governing those actions

are also coded in the same language and shared.

So for the case of how an, e.g. overdraft, is handled I would say those

policies are *not* part of the logical specification of the term

'insufficient funds in a checking account', or loosely using the term

'definition', not part of the definition. They are part of the logical

specification of bank policies, which must be understood in order to predict

bank actions. If properly drafted, I think that an ontology that includes a

specification of 'checking account' would include provision for specifying

bank policies for specific subtypes of 'checking account', in which case

those policy specifications *would* become part of the specifications of

those account subtypes, as necessary conditions for certain contingencies.

(i.e. '*Each* checking account must have *some* policy for acting when a

check exceeds remaining funds').

We can't expect systems with perfect interoperability to be able to

predict events that a person could not predict with the same information.

The problem in the above scenario isn't the inadequacy of the common

ontological language, it is the inadequacy of the knowledge about bank

policy, whether it is a person or an automaton that is trying to predict the

result.

This is a topic of great interest to me, and I would be interested to see

any kind of detailed example of how an attempt at interoperability failed in

spite of accurate common terminology, because of local usage that changed

the meanings of terms. That can be useful in learning what needs to be

included to adequately specify the meanings of certain terms.

Pat

Patrick Cassidy

MICRA Inc.

cassidy@xxxxxxxxx

908-561-3416

-----Original Message-----

From: ontolog-forum-bounces@xxxxxxxxxxxxxxxx

[mailto:ontolog-forum-bounces@xxxxxxxxxxxxxxxx] On Behalf Of John F. Sowa

Sent: Friday, March 02, 2012 11:08 AM

To: ontolog-forum@xxxxxxxxxxxxxxxx

Subject: Re: [ontolog-forum] Constructs, primitives, terms

Pat and Amanda,

> I am curious to know what kinds of inconsistency you are referring to:

JFS

>> So far, so good. But a problem arises when systems A and B

>> exchange messages that use only the "common" vocabulary defined

>> by the shared URIs. They may use the same terms, but their

>> specialized "senses" of the terms may be inconsistent.

Let me restate my previous discussion in more concrete terms:

1. Suppose A and B are both banks. They use many common terms that

are defined precisely, but without much detail -- for example,

a checking account with a globally unique routing number for

electronic funds transfer.

2. But every bank adds more detail about the options for their

accounts, and they differ from the options at other banks. Each

word for describing those options can also be precisely defined

with its own URI, but each bank might combine those terms in

patterns that are unique to that bank.

3. As a result, banks can "interoperate" only on a narrowly defined

set of EFT message types. Messages that use the common terms in

any combination that has not been standardized may be interpreted

in inconsistent ways by bank A and bank B.

> Does this refer to cases where two communicating systems define (or

> otherwise logically specify the intended meaning) of the *same* term

> (e.g. 'widget'), but use different definitions?

That depends on what you mean by "definition". You can say that the

official definition of a term is the statement designated by its URI.

But each bank uses each term in its own context. Each use in each bank

may be consistent with the common definition, but other relationships

in each context may be different.

As a result, a customer who switches to a different bank may get

unpleasant surprises about the fees or constraints that one bank

imposes that the other one didn't. Humans can adjust to such

"surprises", but computer systems usually cannot.

Are you going to call that additional context-dependent information

part of the definition? Most people don't say that explicitly, but

it is just as much a part of the broader "meaning" as any definition.

Unfortunately, that contextual information may be difficult or

impossible to state in a "closed form" definition. It may be

scattered in database formats, procedures, or implicit conventions.

> Systems A and B may use the same "term" where "term" = lexical item,

> to express different concepts. That's not a logical inconsistency.

> It's an ordinary difference in language (natural or artificial),

> and can be handled well by mapping those lexical expressions to

> different concepts.

Yes, but those terms have a common subset of meaning that can be

exchanged with the common URIs for many operations, such as EFT.

But when that additional contextual "meaning" is significant, you

need further conventions. One option is to rename every term X that

was supposed to be "common" to both banks by a qualified name such

as X.A or X.B. But then you need to determine when you can use the

generic name X for some operations or the more specific names such

as X.A and X.B for others.

Furthermore, each bank has many business policies, systems analysts

who interpret those policies, and programmers who implement them.

Each change to any program can change some of that contextual meaning.

Therefore, we need to have version numbers for all our systems.

That means we need unique URIs of the form X.A.v1.0, X.B.v2.7, etc.

Then we need to know when it's possible to interchange data according

to name X and what to do with data that uses names like X.A or worse.

Any belief that URIs and closed form definitions can magically solve

all these problems is an illusion.

John

_________________________________________________________________

Message Archives: http://ontolog.cim3.net/forum/ontolog-forum/

Config Subscr: http://ontolog.cim3.net/mailman/listinfo/ontolog-forum/

Unsubscribe: mailto:ontolog-forum-leave@xxxxxxxxxxxxxxxx

Shared Files: http://ontolog.cim3.net/file/

Community Wiki: http://ontolog.cim3.net/wiki/

To join: http://ontolog.cim3.net/cgi-bin/wiki.pl?WikiHomePage#nid1J