APPLYING KNOWLEDGE BEADS FOR AUTOMATED
NEGOTIATION
Zhuang Yan, Francis Yan and Simon Fong
Faculty of Science and Technology, University of Macau
Shi Meilin
Department of Computer Science, Tsinghua University, Beijing, China
Keywords: Automated Negotiation, Agents, Knowledge.
Abstract: A methodology called Knowledge Beads (KB) that was proposed in our previous paper features object-
oriented formatting and expressive power in defining a complex bill-of-material. In this paper, we show
how this knowledge representation methodology could be used in the design of automated negotiation
systems and how KB helps in giving a unified approach for representing the data throughout the process that
includes evaluation, negotiation, and post-negotiation. We identify two types of knowledge namely General
Knowledge and Negotiation knowledge, in the negotiation process. A methodology is proposed on how
these two types of knowledge can be used in automated negotiation.
1 INTRODUCTION
Previous research (Reeves et al, 2000) attempted
fusing knowledge into agent communication
languages and negotiation functions. They mostly
based on rule-based and/or logic-based approaches.
However, these techniques show their advantages in
individual applications. Knowledge management is
important in many scenarios where agent negotiation
is performed based on knowledge instead of rules
and logics alone. Furthermore those agents do not
have self-learning ability if they cannot interpret
knowledge.
Therefore, we opt for a model which can capture
the key concepts and elements involved in multi-
bilateral multi-attribute e-Procurement negotiation
traces. Especially, these include the relationships
among multiple negotiation parties, negotiation
strategies for trade-off on multiple procurement
attributes, and decision-action rules that drive an
automated negotiation system (Jennings et al.,
1996).
In essence, to automate the agent negotiation
process, it has been widely accepted that two
important technical tasks must be done: Firstly
formulate the negotiation process; and secondly
incorporate necessary negotiation knowledge.
Formalization of the negotiation process enables the
software to automate the process by following some
pre-defined algorithms. Incorporation of expert
negotiation knowledge enables a negotiation system
to conduct automated negotiations effectively and
intelligently on behalf of its users. The knowledge of
the human negotiation experts can be captured in
some form of requirements, constraints, events,
strategic rules, and preference scoring and
aggregation methods. However, due to the diverse
and complex nature of negotiation, the lack of
knowledge interoperability and knowledge-reuse has
posed certain difficulties to automated negotiation.
Through this research work, we anticipate more
efficient e-Commerce could be achieved with the
support of the knowledge in automated negotiation.
This paper sheds some light on knowledge
empowered automated negotiation in the multi-
bilateral multi-attribute e-Procurement environment.
The research aims are to formulate a complete
negotiation life cycle with a knowledge framework
for constructing both the negotiation context and the
negotiation expertise; and to develop a model of
automated agent negotiation based on the knowledge
framework.
90
Yan Z., Yan F., Fong S. and Meilin S. (2007).
APPLYING KNOWLEDGE BEADS FOR AUTOMATED NEGOTIATION.
In Proceedings of the Second International Conference on e-Business, pages 90-93
DOI: 10.5220/0002114500900093
Copyright
c
SciTePress
2 KNOWLEDGE IN
AUTOMATED NEGOTIATION
In a negotiation session, there is plenty of data
which can be collected, manipulated and utilized.
Besides, domain knowledge and negotiation
expertise are crucial for defining negotiation
strategies, plans of actions, and preference scoring
and aggregation methods. Both negotiating parties
are responsible for collectively maintaining their
own data, rules and knowledge repository.
2.1 Knowledge Taxonomy
The goal of a corporate taxonomy is to provide a list
of authorized terms in knowledge management and
information seeking (Conway 2002), as well as the
mapping between concepts to connect negotiation
parties with the right knowledge at the right time.
Mainly two categories of knowledge are addressed
in the proposed framework: (1) general knowledge
and negotiation knowledge. General knowledge
provides the specification of different categories of
objects in the e-Commerce domain, which are the
fundamental knowledge. An object can be a RFQ, a
trader, a deal, or any object that associated with
manipulation methods. (2) Negotiation knowledge,
or negotiation expertise comprises knowledge of, or
skill in observation of experience gained through
negotiation in e-Procurement process. The concept
of experience generally refers to know-how or
procedural knowledge, which is the knowledge of
how to perform certain tasks.
2.2 General Knowledge
General knowledge in the negotiation life cycle
contributes to the formation of the fundamental
knowledge framework of the current negotiation
context. It mainly includes buyer’s RFQ, supplier’s
quotes, negotiators’ profiles, and negotiation traces.
Table 1: General knowledge in a negotiation context.
A negotiation trace is a log recording all the
messages exchanged between two negotiation
partners in a negotiation process. For successful
negotiation in which an agreement is produced in the
end, the negotiation trace contains useful knowledge
describing the nature and progress of the
negotiation.
2.3 Negotiation Knowledge
While general knowledge describes the whole
negotiation context, negotiation knowledge provides
the necessary knowledge used to carry out the
negotiation in an automated way. It comprises
business intelligence for negotiation including a
variety of negotiation strategies and business
conventions, as depicted in the following Table 2.
Table 2: Negotiation knowledge in negotiation context.
Negotiation knowledge are formulated and
specified as procurement rules and constraints by
negotiation experts. There are two types of
constraints. The fundamental type of constraint is to
define a valid range for an individual attribute, or an
inter-attribute relation for multiple attributes within
the same knowledge item. Another type of constraint
is defined between a particular knowledge item and
other relevant ones. This kind of constraint usually
exists in a Bill-of-Material (BOM) consisting of
multiple RFQs defined for different product items
respectively. Constraints are usually used in quotes
evaluation and ranking phase for screening qualified
quotes. They are also used in negotiation to screen
for attributes that fall within predefined ranges.
Besides constraints, rules are widely used to
describe the negotiation knowledge about
relationships.
2.4 Knowledge Representation
Knowledge Bead (KB) as an object-oriented
knowledge representation scheme was defined in
(Fong and Zhuang, 2002), as an encapsulation of
definition, behavior, and data: KB = Definition +
Behavior + Data.
APPLYING KNOWLEDGE BEADS FOR AUTOMATED NEGOTIATION
91
A KB can be a composite object, or a simple,
atomic part object in most cases; each has their own
methods and data. Definition means a static unique
description; this can be a UPC (Universal Product
Code) or a unique index implemented at the
ontology databases for referencing this KB.
Behavior is described by a set of possible methods
and rules manipulating KB’s and their attributes.
Some typical ones include KB formation,
duplication, attribute alteration, pruning and linking
to other KB’s. They are analogous to class functions
in object-oriented programming, and can be
inherited from base classes. Data consists of a group
of attributes defined for the KB. Associated with
each attribute, a weight is given as a relative priority
indicating how important this attribute is in the
current KB.
The use of KB in representing general
knowledge about the negotiation context is shown in
Figure 1. KB is created first by the user through
some user interface. The data on the submitted web-
page (form) is extracted into the construction of a
KB object that resides on the server. The KB object
is then used in the quote evaluation and negotiation
processes. Note that KB is a general data
representation format that can define the general
knowledge items, and that can be implemented in
object-oriented languages such as Java.
Figure 1: Representation of General Knowledge in KB.
Every general knowledge item can be
represented in a certain template. The main
categories of domain correspond to the types of
general knowledge items including RFQs, quotes,
proposals, agreements, profiles, and traces as listed
in Table 1.
The product space is represented as a labeled,
directed graph with two types of nodes: a leaf node
and a category node, as depicted in Figure. 2.. Every
leaf node in the product space is represented in a KB
template developed by the system. It inherits
attributes and behaviors from its parent category
node, along with new features and operations added.
Each KB template has an identity number composed
of the category name and the name of the leaf node.
The category name provides the basic domain
information about a general knowledge item which
makes use of the KB template. It is represented as a
sequence of labels corresponding to the edges in the
path, e.g:
/ProductCategories/Electroics&Computers/Came
ras&Photo/DigitalCameras
Product Categories
Electronics & ComputersHome
Cameras & Photo
Digital Cameras
Electronics
Software
Lamps & Lighting
Bedding
Clothing & Assessories
Men
Women
Uniforms
RFQ 1
Figure 2: Part of the product space of KB templates.
3 METHODOLOGY OF KB’S FOR
AUTOMATED NEGOTIATION
Figure 3: Methodology of KB’s for automated negotiation.
As general knowledge items are represented by
KB’s and negotiation knowledge are specified as
ICE-B 2007 - International Conference on e-Business
92
constraints and rules that defined on attributes in
KB’s, the classification and clustering of KB’s helps
to manipulate and present the knowledge whenever
it is needed by the system. Figure 3 shows how
negotiation knowledge and general knowledge are
used respectively for assisting the user to create a
RFQ, and for the automated negotiation process. At
the end of the process, log files are generated and
added to the general knowledge database.
To make use of the knowledge contained in
KB’s, the negotiator first identifies the function that
he need to do in the negotiation process. Then it’s
the knowledge agent who provides a concrete plan
of utilizing the appropriate knowledge in the specific
function. The proposed model also provides the
negotiator the flexibility to adjust the weight of the
knowledge factors which affect the function result.
To our knowledge, most current automated
negotiation systems lack the ability of specifying the
explicit use of knowledge in a systematic way, thus
lack an efficient knowledge assisted automatic
negotiation process. For this purpose, we define
meta-KB as a meta-object for describing the
procedural knowledge necessary to perform a certain
task in the e-Procurement context. It contains the
meta-knowledge about KB’s, which is knowledge
about knowledge. The function which makes use of
the meta-KB determines its discipline. Like an
ordinary KB, a meta-KB contains attributes forming
the knowledge. The attributes are either inherited
from an existing KB or defined especially for the
specific function, depending on the meta-KB’s
discipline. For each attribute, the meta-KB specifies
how the attribute value is obtained.
Several typical functions are executed many
times in different phases or in parallel during the
multi-bilateral negotiations. These functions include
supplier credit evaluation, quote evaluation, and
negotiation result assessment.
The meta-KB for evaluation of a supplier inherits
the attributes from the KB comprising knowledge
about a supplier’s credit as shown in Table 3. It is
illustrated in the following table.
Table 3: Meta-KB for supplier evaluation.
The tag ‘Meta-KB’ denotes it a meta-KB, and the
use of the meta-KB is declared at the top of the
table. It then specifies from which KB template that
the meta-KB inherits its attributes. The value of
Base Reputation is input from a Negotiation Expert
manually. The attribute Number of Contracts Made
has a returned function value evaluated on the
negotiation log. The function is denoted by f in the
table. The attribute Average Utility also has a
returned function value evaluated on the negotiation
log. The function is denoted by g in the table. The
negotiation log is a log containing all the past
successful deals committed with the particular
supplier. Weights associated with attributes are also
inherited from the supplier credit profile, which are
not shown here. Detials of the evaluation functions f
and g can be found in (Zhang, 2006).
4 CONCLUSION
We discussed issues of applying Knowledge Beads
(KB) into automated negotiation for e-Commerce. A
methodology that is based on Knowledge Bead, an
object-oriented ontology-based building block for
knowledge representation, is proposed. Using KB
and its methodology, quote specification and
bargaining process can be streamlined, and data
resulted from negotiation can be reused as
knowledge in future negotiation. This provides a
foundation for the knowledge management life cycle
designed for coexisting with the negotiation life
cycle.
REFERENCES
Reeves, D., Grosof, B., Wellman, M., and Chan, H., 2000.
Towards a Declarative Language for Negotiating
Executable Contracts. IBM Watson Research.
Jennings, N., Faratin, M., Johnson, T., Norman, J.,
O’Brien, P., and Wiegand, E., 1996. Agent-based
Business Process Management. International Journal
of Cooperative Information Systems, 5(2&3), 105-130.
Conway, S., and Sligar, C., 2002. Unlocking Knowledge
Assets. Knowledge Management Solutions. Microsoft
Corporation.
Fong, S., and Zhuang, Y., 2002. Enabling Agent
Negotiation in e-Trading Environments Through
Knowledge Beads. IEEE International Conference on
Intelligent Engineering Systems.
Zhang, Y., 2006. Knowledge-Empowered Automated
Negotiation System for B2B e-Commerce. PhD
Dissertation. Tsignhua University. Dec 2006.
APPLYING KNOWLEDGE BEADS FOR AUTOMATED NEGOTIATION
93