Challenges in the New E-ra: the Pandora's Box for AI?
This document emerged after a series of two talks given by Yannis
Kalfoglou and Daniela
Carbogim to the SSP Research
Group at the Division
of Informatics, on the 8th
and the 15th
of March of 2000, respectively.
As a result of the aforementioned talks, we have produced this
document in order to support timely dissemination of knowledge and
alert scholars on issues that might be interesting to them. We would
very much appreciate comments, ideas and contributions you may
have, so feel free to send us an email.
Contents
The New E-ra
Nowadays, we are witnessing a new electronic era (in short, e-ra) and
the development of a global information society.
The Internet has some remarkable characteristics which have given rise
to the Information Revolution:
- interactivity / communication
- distributed / non structured
- cooperation
- large quantity / heterogeneous quality of information
But "revolution" is about radical change (whereas "evolution" is about
gradual change), so what are the changes in this new Information
Society? Here are some candidates
- INFORMATION
- COOPERATION
- easier in the world of bits
- catalyses and facilitates cooperative construction of
information
- among geographically and culturally distant parties
These changes create needs and expectations in the Information
Society - in the new e-ra - which result for instance in new
forms of work organisation (Domenico
De Masi argues that in the post-industrial society we won't be
able to differentiate work from leisure, neither from learning), new
communities, new types of legal and juridical problems. The new e-ra
creates cooperation-driven demands, business-driven demands,
technology-driven demands...
AI has provided a number of methods and techniques for representing
and structuring knowledge and reason about it. This is the main reason
why we feel that AI has an important role to play in the new
Information Society.
| Links: The Information Society |
|
The New E-ra Demands &
Expectations
It is a widely acknowledged fact that the Internet alters the way
we do business, we communicate, we socialise, in short, changes our
lives. This has injected unparallel interest from all sorts of
businesses in an area traditionally studied by computer scientists;
and poses such demands as:
- short time-to-market schedules;
- exploration and exploitation of state-of-the-art technologies;
- reliable, robust, agile software;
- integration of existing processes and transformation of today's
way of doing business to the new e-ra way.
On the other hand, recent advances of the new e-ra technologies drive
demands as:
- exploration of new domains and applications;
- experimentation with and establishment of the underlying
infrastructure for the new e-ra;
- evaluation and assessment of potential technologies;
- transformation or discontinuation of legacy systems.
which fuels our expectations:
- business and technology driven demands are complement to each other;
- as such we expect to smoothly fulfil these demands;
- which is not that straightforward: for example, most of research
effort is focussed on how to create new applications for the
new e-ra;
- however, there is a need from businesses to find ways on how to
convert existing processes and connect working
systems;
- at the end, we expect to see a more valued role for the so called
KT
(Knowledge Technology) which could, eventually, leapfrog the
omnipresent IT.
AI Contributions and Potential Applications
We identify the following areas of AI contributions:
- ontologies;
- knowledge representation and inference techniques;
- logics;
- argumentation;
- mediators;
- ...others under scrutiny?
which could be deployed to application areas such as:
- e-commerce, in particular B2B and B2C areas;
- knowledge management;
- organisational memories;
- "intelligent" brokering services;
- adaptable knowledge components;
- ...many more undefined at present?
Ontologies
As this document is not meant to be an ontologies recourse, we
limit our links and references to collections of information
related to the field. These are collections accessible online and not
references to the literature, like special issues of journals.
Computational Logic and Logic-Based Techniques
Let us look at the role of logic-based reasoning techniques can
play in this new e-ra, especially from the perspective of information
integration.
"Information integration is about getting information sources to
talk to each other"
This definition is given in a roadmap paper produced by the
computational logic community entitled Information
Integration and Computational Logic.
Some of the work carried out by the SSP Group has been
related to logic-based information integration, such as:
In the Computational Logic
community, most of the work on information integration has been
about using logic as an integration framework for defining
mediators. Mediators are components that link different
information sources to the user and applications.
By modelling the contents of information sources and the relation
between the different sources and the mediator, logic-based
representation and reasoning thus provides a mechanism for
communication and "intelligent" cooperation.
- "intelligent": semantic integration of information
Note that when we talk about semantic integration of information,
a number of issues arise:
- inconsistency / conflict
- uncertainty
- partial / incomplete information
These have been at the centre of logic-based knowledge
representation research:
- default logic
- defeasible logic
- probabilistic logic
- preference and non-monotonic reasoning
- argumentation
However, if the use of logic-based information integration is to be
scaled up in the new e-ra, it should be somehow supported by web
technology.
An example is the work on integration of business rules
for e-commerce (CommonRules), which
provides an XML encoding of corteous logic programs, used to represent
business rules. Applications of business rules in multi-agent and
e-commerce scenarios include negotiation between agents.
| Links: Logic-Based Approaches for Information
Integration |
|
Argumentation
We look at argumentation as a reasoning technique for dealing with
imperfect information. Conflict is the essence of argumentation
Argumentation paradigm: reasoning by constructing and weighing up
arguments.
A shift on focus:
- Logical characterisation of argumentation: automatic generation of
arguments from a set of premises
- Computer-Supported Collaborative Argumentation (CSCA) and Mediation
Systems: representation of arguments from different sources
CSCA & Mediation Systems
Applications to design of artifacts (both in the world of bits and
in the world of atoms): supporting collaborative
processes. Argumentation-based Design Rationale is about explicitly
recording reasons why an artifact was developed in some way.
It has been argued
that argumentation-based design rationale is useful and usable, and plays
several roles in design such as:
- structuring design problems
- facilitating communication and reasoning
- supporting maintenance and reuse
- exposing assumptions and conflicts
- enabling formal incorporation of diverse types of information
The aim of mediation systems is to augment and mediate
argumentation in groups:
- mediation systems support deliberative processes between one or
more participants
- mediation systems can support structured forms of group decision
making via Internet (The Zeno System)
- goal is to reach some decision
- example: decision fora, where it is important to argue and
negotiate about different issues
Successes
and Failures of CSCA systems involve the interaction of a number
of factors such as:
- knowledge about domain and about argumentation
- training in CSCA systems
- motivation for using CSCA systems
A vision for the role of argumentation in the new e-ra:
- COLLABORATION in the new Information Society
- cultural Impact: "every culture is developed cooperatively"
- changes in university paradigm (distance learning, OU), professional training paradigm,
artifact development paradigm...
Argumentation and Negotiation
Negotiation: the problem of achieving mutually acceptable
agreements between agents.
Can argumentation provide or support intelligent interaction
between agents?
- interest in applying argumentation systems to capture
negotiation
- negotiation often involve the exchange of arguments between
agents
- various views on how negotiation is related to argumentation (see
the report of a panel discussion on Negotiation in multi-agent
systems at UKMAS98 in The Knowledge Engineering Review 14(3),
1999)
- is argumentation a type of negotiation?
- is argumentation one way to define the semantics of negotiation?
- are negotiation and argumentation distinct disciplines?
Our view is that we can ground a number of problems with
very distinctive characteristics (such as negotiation and design) into
a similar source in argumentation.
XML & XML Schemata
XML is a metalanguage for creating markup languages that describe
data:
- a method for putting structured data in a text file
- HTML x XML
- HTML: tags to describe document structure and visualisation
- XML: tags to specify parts of data
- XML is understandable text - but is not supposed to be read
DTD's (Document Type Definition) and XML Schemata
- technologies for describing the structure of XML documents
- validity and well-formedness
- XML Schemata
- more expressive than DTD's
- more complex than DTD's
- use XML syntax
StyleSheet Technologies: CSS and XSL
- how the source content should be styled on some presentation medium
- CSS x XSL: "use CSS when you can, use XSL when you must" (W3C
recommendation)
- CSS: easier to use, learn, maintain; simpler, so it has limitations
- XSL: if transformation is involved; use XML syntax
| Links: XML & XML Schemata |
|
RDF & RDF Schemata
RDF (Resource Description Framework) was among the first standards of
W3C for processing meta-data. Briefly speaking, RDF provides the means
for adding semantics to a document without making any assumptions
about the structure of the document. The strength of RDF is that it
provides meta-level facilities: you can make statements about
statement.
In detail, the data model of RDF provides three object types:
- resources(subjects),
- property types(predicates), and
- tatements/values(objects).
A resource is an entity that can be referred to by an address (URI), a
property defines a binary relation between resources and/or atomic
values provided by primitive datatype definitions in XML, and a
statement specifies a value for a property.
For example, the following statement:
Author(http://www.dai.ed.ac.uk/daidb/people/homes/yannisk)=Yannis
defines Yannis as the author of the mentioned web page. In the same
fashion, the statement:
Claim(Daniela)=(Author(http://www.dai.ed.ac.uk/daidb/people/homes/yannisk)=Yannis)
is a meta-statement, where we define that Daniela claims that the
author of the mentioned web page is Yannis.
RDF Schemata (in short, RDFs) provide a basic type schema which
could be "directly used to describe an ontology". They are based on
core classes, property types, and constraints. Core classes are:
Resource, Property type, Class. Core property types are: instance-of,
subclass-of. Core constraints are: range and domain.
Of particular interest, are the constraints definition
facilities. Some examples of constraints are:
- "the value of a property should be a resource of a designated
class. This is expressed by the range property. For example, a
range constraint applying to the 'author' property may express that
the value of an 'author' property must be a resource of class
'person'."
- "a property may only be used on properties of a certain class.
For example, that an 'author' property could only originate from a resource
that was an instance of class 'Book'. This is expressed using the 'domain
property'."
| Links: RDF & RDF Schemata |
|
Knowledge Management and Organisational
Memory
Knowledge Management (KM) is the: "formal management of knowledge for
facilitating creation, access, and reuse of knowledge".
Organisational memories (OMs) "provide the means for storing,
retrieving and distributing knowledge from an organisation's
repositories".
KM aims to develop and deploy knowledge whereas OMs preserve
knowledge. They both centered upon the enhancement of an
organisation's competitiveness by improving the way it manages its
knowledge.
Why do we need KM? Because of
- "environmental pressures: increasingly competitive global
market place (maybe 'marketspace'?)",
- "technological advancements: recent developments in
Internet technology".
The goal of KM is to "create valuable information: convert
individually available knowledge into group or organisationally
available knowledge".
Convert and connect processes:
- convert
- individual to group knowledge,
- data to knowledge,
- text to knowledge,
- and connect
- people to knowledge,
- knowledge to knowledge,
- people to people,
- knowledge to people.
KM, OMs and the new E-ra:
It is a fact that KM/OMs were always a concern for
organisations. There was always an interest and investment to
technologies that can effectively support and achieve efficient KM. So
what the new E-ra has to do with it?
The emergence of the new E-ra made KM/OMs a necessity. Why?
The technological infrastructure of the new E-ra facilitates the
development, deployment, distribution, and maintenance of knowledge
assets that otherwise would require expensive equipment, high level of
expertise, and labour-intensive engineering to implement. It is
exactly these knowledge assets that are now recognised by many
organisations as their most valuable assets, a situation which is
anticipated to be the rule rather than the exception in the
foreseeable future. Globalisation makes organisations to
re-think and revise their structure in order to continue to ...exist!
Whereas a well-designed KM/OMs could drive this change the new E-ra
provides the infrastructure to achieve this.
As for ontologies, this is not meant to be a definitive
resource of information for KM and OMs. We provide, however, some
interesting links.
| Links: Knowledge Management and Organisational
Memory |
|
Is The New E-ra a Pandora's Box for
AI?
After a fruitful discussion we have managed to reach a verdict! We
took into account all possible angles of viewing this intricate issue
and came up with a perspective conclusion to our issue: "is the new
e-ra the Pandora's box for AI?".
Here is the answer, in a structured way: