Walking Up and Down the Abstraction Ladder: Reflections about the TAMI Project Carlos Delgado Kloos

Walking Up and Down the Abstraction Ladder: Reflections about the TAMI ProjectCarlos Delgado Kloos<[email protected]><[email protected]>

Logic and Law

The first logicians were lawyers... there are famous mathematicians

that were also lawyers (Leibniz)... but Legal Systems are far

from being Logic Systems.

Why is it taking so long?

Two Essential Points

1) The concepts Defining the right concepts Evolving them over time

2) The reasoning Finding the right logic

Scenario 3

JOHN DOEFlies LGA-ORD

(LaGuardia – O’Hare)June 2004

AIRLINECreates a PNR

(“Passenger Name Record”)

COMMERCIAL DATA VENDOR

Provides match data

Terrorism ScreeningCenter (TSC)

Provides “no fly” list

National CounterterrorismCenter (NCTC)

Provides “stripped” version of aggregated dataset -

known & suspected terrorists

Transportation SecurityAdministration (TSA)

Finds a “possible” matchwhile testing

Secure Flight program

Joint Terrorism Task Force (JTTF)

Seeks evidence of other clear legal violation

NY StateDept. of State

Website search revealsJohn Doe is

“deadbeat dad”

JOHN DOEIs arrested

Scenario 5

Scenario 3






Provides match data













“deadbeat dad”

JOHN DOEIs arrested

Scenario 4

TAMI Scenarios

Scenario 3






Provides match data













“deadbeat dad”

JOHN DOEIs arrested

Scenario 3

Scenario 3:pnr-1

a :Transfer

:flight-test-search-1a :Search

:transfer-1a :Transfer

:open-investigation-1a :OpeningCase

:assignment-1a :Assignment

:assignment-2a :Assignment

:open-source-search-2a :Search

j:Arrest

Architecture

cwm

Schema

Rules

Data Data’

Rules# if data transferred to SOR, purposes intersect with purposes of appl.

SORN{ X a ProofNode. G a TamiData. G log:includes { X ts:purpose P; ts:recipient R }. G log:includes { R ts:notice S }. G log:includes { S a ts:SORN; ts:purpose P }.} => { X sorn S; purpose P; a JustifiedNode; recipient R }.

# if data transferred out of SOR, intersect with purposes of routine use { X a ProofNode. G a TamiData. G log:includes { X ts:purpose P; ts:source R; ts:antecedent Y }. G log:includes { R ts:notice S }. G log:includes { S a ts:SORN; ts:routineUse [ ts:purpose P ] }.} => { X purpose P; a JustifiedNode; source R }.

# if transaction has no purpose, it retains the purpose of its antecedent{ X a JustifiedNode; purpose P. G a TamiData. G log:includes { Y ts:antecedent X }. G log:includes { Y ts:purpose tb:any-purpose }.} => { Y purpose P; a JustifiedNode. }.

An Ontology for TAMI Scenario 3

Classes defined in RDF/Notation3: Person

PersonName, Location, Organization

DataRecord Database, Flight, File

Event Search, Transfer, Assignment, CaseOpening,

WarrantIssue

Person PersonName

Subject

Official

name

LocationhomeAddress

Organization

work

xsd:DatebirthDate

Person

Employment

position

EmploymentPosition

employer

Location

address

Victim Suspect

domain range

subClassOf

PersonName

u:personSurName

u:personMiddleName

u:personGivenName

Literal

Location

u:locationStateName

u:locationCityName

u:locationAddress

Organization organizationName

u:locationPostalCodeIDAddress

Person (cont.)

PassengerNameRecord

Databasesource

Personpassenger

Flightflight

Flight

xsd:Datedate

Literalnumber

air:Iataorigin

air:Iatadestination

DataRecord

DataRecordxsd:Datedate

Organization

owner

Event

xsd:Datedate

Literalname

Search Transfer Assignment CaseOpening WarrantIssue

Reasontrigger

Event

OrganizationlogOwner

John Doe lives in New York

Lifting up the strings

Web


Person homeAddress Location

name u:locationStateName

John Doe lives in New YorkWeb

SW


PersonName homeAddress Location

u:personGivenName u:locationStateName

Person

name

u:personLastName

John Doe lives in New YorkWeb

SW

SWSSW

Questions

Which concepts (classes and properties) to define? How are they related?

How many levels of abstraction to use? Is there a natural number of them? Is there such a thing as the semantics?

How do the defined concepts relate to concepts defined elsewhere? How to define categories of concepts?

How useful for future scenarios are the concepts defined?

Conceptualization

Ontologies are used in many areas with success

has an commonsense ontology with 300,000 terms

Conceptualization: Difficulties "Concepts are the glue that

holds our mental world together" Necessity of category fuzziness,

typicality of items:is a telephone furniture,is an olive a fruit?

Absence of transivityfor category membership: a chair is furniture,a car seat is a chair,but is a car seat furniture?

– Gregory L. Murphy:The Big Book of Concepts, MIT Press 2004

Example 1: Concept ‘Book’

What is a book? It is a clear concept!

It has an author a title ...

Right? Wrong!

FRBR: 4 Concepts for ‘Book’

Functional Requirementsfor Bibliographic Records 1) Work: distinct intellectual or artistic creation 2) Expression: specific intellectual or artistic

form that a work takes each time it is 'realized' 3) Manifestation: physical embodiment of an

expression of a work 4) Item: a single exemplar of a manifestation

Book Hierarchy

Work Expression Manifestation Item

EN

ES

DE

Book in RDF

Work

realized through

Expression

embodied in

Manifestation

exemplified by

Item

Don Quijote

Don Quijote

Don Quijote

Don Quijote

realized through

embodied in

exemplified by

written byM. Cervantes

John Ormsby

8420467286

Shelf A.1

translated by

ISBN

location

Equivalences forWorks and Expressions

Example 2: Carlos I am Carlos No, this is my name in Spanish

(in English it would be Charles) No, this is the textual representation of my

name in Spanish (the sound representation is )

No, this is a particular image of … (another is Carlos)

No, this is what you see of the image… …

Carlos Hierarchy

Equivalence

Notion of equivalence Going up one level, implies putting a

number of objects in the same equivalence class

We might characterize by means of a representative

Semantics is abstraction

Carlos in RDF

Person nameES Word

Carlos CarlosnameES

text String

Carlostext

…

…

semantics syntax

Eliminating levels

Is the left lane turning left,or the cars on it?

Do you get an e-mailor an e-mail message?

What is “Carlos”?(a Person, a Name,a Text, an Image, …)

What do you mean by “book”?

Implicit Abstraction Levels

It is fundamental to recognize

that normal text will make

some abstraction levels implicit

Implicit Abstraction Levels

At any moment, it is possible to introduce additional abstraction levels

When I say this, I mean that

Lifting to complex data structures with no explicit symbol [1,2,3] + [4,5,6] = [5,7,9]

Example 3:Through the Looking Glass `The name of the song is called "Haddocks' Eyes".' `Oh, that's the name of the song, is it?' Alice said, trying

to feel interested. `No, you don't understand,' the Knight said, looking a little

vexed. `That's what the name is called. The name really is "The Aged Aged Man".'

`Then I ought to have said "That's what the song is called"?' Alice corrected herself.

`No, you oughtn't: that's quite another thing! The song is called "Ways and Means": but that's only what it's called, you know!'

`Well, what is the song, then?' said Alice, who was by this time completely bewildered.

`I was coming to that,' the Knight said. `The song really is "A-sitting On a Gate": and the tune's my own invention.'

Through the Looking Glassin RDF

Song named Name

calledLiteral

A-sittingOn a Gate

The AgedAged Man

Haddocks' Eyes

named

called

calledLiteral

Ways andMeans

called

domaindomaindomain

range

range range

Classes

Instances

“The” TAMI Ontology?

Some decisions: We’ll simplify this for the moment… We might want to change this in the next scenario… Better to take this big existing ontology for that…

The ontology will necessarily need to evolve continuously Might need to introduce more abstraction levels Might need to make other kind of changes

Better be prepared for that How and how well is RDF prepared for it?

US Constitution

establishes

Legislative Executive Judiciary

Territories

States

enacts

Statutes

President

issues

HomelandSecurity

PresidentialDirectives

National Security

PresidentialDirectives

ExecutiveOrders

Agencies

issue

Regulations

Resultingin

Policies

Practices

issues

Decisions

CommonLaw

Evolution

Finding the final solution at the beginning is impossible, because you don’t know where you are going

Intermediate decisions are necessary, but they bind and have implications (backward compatibility) Radio: Mono → Stereo Television: B/W → Color Credit card processing: Manual → Computerized Phone: Pulse → Tone

Refactoring

How to Represent Evolution in RDF?

Possibilities: More classes (code relation into the name) Use the subclass mechanism Relate classes outside of RDF rather than

inside the language

Use syntax to mean semantics

Context

What is my (place of) residence? Where I live presently Where I have lived on a permanent basis

(until next change) For the Immigration Authorities For Tax Purposes (Federal or State?)

The importance of the context “Yo soy yo y mi circunstancia”

– José Ortega y Gasset

Evolution

Context Concurrent systems: model the system and

the environment (what is importamt about the rest of the world)

Evolution Notion of refinement to relate different

versions

Conclusions: Conceptualization

Knowledge Representation more complex than what I thought Psychological, philosophical questions as well

as Software Engineering considerations No predefined number of right abstraction

levels Lift up the strings Beware of/introduce additional levels

Conclusions: Evolution

Careful balance between finding the final solution and intermediate ones The importance of taking evolution into

account from the very beginning It would be good if the representation

language (RDF) took care of this, but it is not necessary

Modelling in the small vs modelling in the large

Law and Logic

The art of finding the right abstraction levels

http://creativecommons.org/licenses/by-nc-nd/2.5/

Documents

Walking Up and Down the Abstraction Ladder: Reflections about the TAMI Project Carlos Delgado Kloos