REPUTATION BASED INTELLIGENT AGENT NEGOTIATION

FRAMEWORKS IN THE E-MARKETPLACE

Malamati Louta

Technological Educational Institute of Western Macedonia, Koila, Kozani, Greece

Ioanna Roussaki

School of Electrical and Computer Engineering, National Technical University of Athens, 9 Heroon Polytechneiou Str,

Athens, Greece

Lambros Pechlivanos

Department of International and European Economic Studies, Athens University of Economics and Business,

Athens,Greece

Keywords: Intelligent Agents, Negotiation Protocol & Model & Strategy, Ranking Mechanism, Reputation Mechanism.

Abstract: E-commerce is expected to achieve high market penetration if coupled with the appropriate technologies.

Mobile Agent Technology (MAT) may enhance the intelligence and improve the efficiency of systems in the

e-marketplace. Such a highly competitive and extremely dynamic market should encompass mechanisms for

enabling users (Buyers) to find the most appropriate service providers (Sellers), i.e., those offering adequate

quality services at a certain time period in a cost efficient manner. In this study, the Buyers’ decision on the

“best” Seller is based on a weighted combination of the evaluation of the quality of the Sellers’ offer

(performance related factor) and of their reputation rating (reliability related factor). Efficient negotiation

frameworks are enhanced with a Sellers’ collaborative reputation mechanism, which helps estimating their

trustworthiness and predicting their future behaviour, taking into account the Sellers’ past performance in

satisfying the Buyers’ expectations. In essence, Sellers are rated with respect to whether they honoured or

not the agreements they have established with the Buyers, thus introducing the concept of trust among the

negotiators. The reputation mechanism considers both first-hand information (acquired from the Buyer’s

past experiences with the Sellers) and second-hand information (disseminated from other Buyers’ based on

their own past experiences with the Sellers), while spurious reputation ratings are taken into account.

1 INTRODUCTION

In the liberalised and deregulated e-marketplace

some key factors for service providers’ success are

the following. First, the efficiency with which

services will be developed. Second, the quality level,

in relation with the corresponding cost, of new

services. Third, the efficiency with which the

services will be operated (controlled, maintained,

administered, etc.). The aim of this paper is, in

accordance with efficient service operation

objectives, to propose enhancements to the

sophistication of the negotiation functionality that

can be offered by e-commerce systems in open

competitive communications environments. This

study is based upon the notion of interacting

intelligent agents which participate in trading

activities on behalf of their owners, while exhibiting

properties such as autonomy, reactiveness, and

proactiveness, in order to achieve particular

objectives and accomplish their goals (He, 2003).

Automated negotiation is a very broad and

encompassing field. Thus, it is vital to understand

the dimensions and range of options available. When

building autonomous agents capable of sophisticated

and flexible negotiation, three broad areas need to be

considered (Faratin, 1998): (i) what negotiation

protocol and model will be adopted, (ii) what are the

issues over which negotiation will take place, and

Louta M., Roussaki I. and Pechlivanos L. (2006).

REPUTATION BASED INTELLIGENT AGENT NEGOTIATION FRAMEWORKS IN THE E-MARKETPLACE.

In Proceedings of the International Conference on e-Business, pages 5-12

DOI: 10.5220/0001426600050012

 SciTePress

(iii) what negotiation strategies will the agents

employ. The negotiation protocol defines the “rules

of encounter” between the agents (Rosenschein,

1994). Then, depending on the goals set for the

agents and the negotiation protocol, the negotiation

strategies are determined (Roussaki, 2003).

In the highly competitive and dynamic e-

marketplace users (Buyers) should be provided with

mechanisms that enable them to find the most

appropriate service providers (Sellers), i.e., those

offering the desirable quality of service at a certain

time period in a cost efficient manner. In this study

we present such mechanisms. As a first step, a

negotiation protocol to be employed in an automatic

multi-lateral, multi-issue negotiation model is

proposed and efficient negotiation strategies for

Business-to-Consumer e-commerce are presented. In

this framework, the roles of the negotiating agents

may be classified into two main categories that, in

principle, are in conflict. These two categories are:

the Buyer Agents (BAs) and the Seller Agents (SAs)

that are both considered to be rational and self-

interested, while aiming to maximise their owners’

profit.

A multi-round negotiation framework is

exploited, which demonstrates inherent

computational and communication advantages over

single step mechanisms in such complex

frameworks (Conitzer, 2003). In essence, the agents

hold private information, which may be revealed

incrementally, only on an as-needed basis. The

framework considered covers multi-issue contracts

and multi-party situations, while being a highly

dynamic one, in the sense that its variables,

attributes and objectives may change over time. The

designed negotiation strategies assume the case

where the negotiators face strict deadlines, and assist

agents to reach to a satisfactory agreement within

the specified time-limits.

E-marketplace is commonly perceived as an

environment offering both opportunities and threats.

Buyers’ or Sellers’ misbehaviour due to selfish or

malicious reasons can significantly degrade the

performance of the e-market. To cope with

misbehaviour the negotiators should be able to

automatically adapt their strategies to different

levels of cooperation and trust. Reputation

Mechanisms provide means of obtaining a reliability

rating of participants in e-marketplace environments

exploiting learning from experience concept and

serve as an incentive for good behaviour to avoid the

negative consequences of a bad reputation spreading

in the market.

In the context of this study, as a second step, the

proposed framework is enhanced by a Sellers’

collaborative reputation mechanism, which takes

into account the Sellers’ past performance in

consistently satisfying Buyers’ expectations. To be

more specific, the reputation mechanism rates the

Sellers with respect to whether they honoured or not

the agreements established with the Buyers, thus

introducing the concept of trust among the

negotiating parties. Most reputation based systems in

related research literature aim to enable parties to

make decisions on which parties to

negotiate/cooperate with or exclude, after they have

been informed about the reputation ratings of the

parties of interest. The authors in this study do not

directly exclude / isolate the Sellers that are deemed

misbehaving, but instead base the Buyers’ decision

on the most appropriate Seller on a weighted

combination of the evaluation of the quality of the

Sellers’ offer (performance related factor) and of

their reputation rating (reliability related factor).

The reputation mechanism considers both first-hand

information (acquired from the BA’s past

experiences with the SAs) and second-hand

information (disseminated from other BAs), while

spurious reputation ratings are taken into account.

The rest of the paper is structured as follows.

Section 2, presents the negotiation framework

adopted in detail. Different contract ranking

mechanisms are employed instead of the usual

alternating sequential offers pattern, while the

concept of decision issues is introduced. In Section

3, a collaborative reputation mechanism is presented

aiming to offer an efficient way of building the

necessary level of trust in the e-market. Finally, in

Section 4, conclusions are drawn and directions for

future plans are presented.

2 THE PROPOSED

NEGOTIATION FRAMEWORK

In order to create a successful negotiation

framework, the design of an appropriate protocol

that will govern the interactions between the

negotiation participants is necessary. Depending on

the specific negotiation problem that needs to be

solved, a protocol is the set of rules that

correspondingly constrain the proposals that the

negotiation parties are able to make. In this section,

we initially describe the adopted negotiation

protocol that is based on a ranking mechanism on

the Buyer’s side. Subsequently, an efficient dynamic

negotiation model is presented, based on the multi-

issue value scoring system introduced by Raiffa

(Raiffa, 1982), in the context of bilateral

negotiations. Based on the designed negotiation

protocol, the proposed multi-party, multi-issue,

dynamic model is exploited by the SA in its contract

generation process, and by the BA during the

ICE-B 2006 - INTERNATIONAL CONFERENCE ON E-BUSINESS

contract evaluation phase. Our focus is laid on the

rationale of the SA, while simplifying assumptions

are made regarding the BA’s logic. We consider that

a negotiation is successful, if a mutually acceptable

contract is reached within reasonable time. Since an

exhaustive exploration of the possible contract space

may form a computationally intensive task, the SAs

are provided with a mechanism enabling them to

find good (near optimal) solutions in reasonable

time, by means of computationally efficient

algorithms. The rest of this section is structured as

follows. In subsection 2.1, the designed negotiation

protocol and model are presented, while in

subsection 2.2 the basic elements of the negotiation

problem and the designed negotiation strategies are

provided.

2.1 Designed Negotiation Protocol &

Model

In the negotiation research literature, the interactions

among the parties follow mostly the rules of an

alternating sequential protocol in which the agents

take turns to make offers and counter offers

(Rubinstein, 1982). This model however necessitates

an advanced reasoning component on behalf of the

BA as well as the SA. In this study, we initially

tackle a simpler case where BA does not give a

counter offer to the SA, but ranks the SA’s offers

instead. This ranking is then provided to the SA,

which generates a new offer hopefully closer to a

mutually acceptable contract. This process continues

until an agreement is reached, or one of the parties

withdraws. This protocol is very efficient in case the

BA is not able to express the user

requirements/preferences in a completely quantified

way, while being capable of selecting, classifying or

rating the contract(s) proposed.

The protocol adopted can be described as

follows. Once the agents have determined the set of

issues over which they will negotiate, the

negotiation process consists of an alternate

succession of contract proposals on behalf of the SA

and subsequent ranking of them by the BA according

to its preferences and current conditions. Thus, at

each round, the SA sends to the BA

N contracts

(i.e.,

N packets consisting on n -plets of values of

the n contract issues), which are subsequently

evaluated by the BA and a rank vector is returned to

the SA. This process continues until a contract

proposed by the SA is accepted by the BA or one of

the agents terminates the negotiation (e.g., if the

time deadline is reached without an agreement being

in place). Even though negotiation can be initiated

by SAs or BAs, only the SAs propose concrete

contracts, as there is no counter offer generation

mechanism for the BAs. We hereafter consider the

case where the negotiation process is initiated by the

BA who sends to the SA an initial Request for

Proposal (RFP) specifying the types and nature of

the contract issues and the values of all non

negotiable parameters. The main issue is assumed to

be the price of the good/service under negotiation,

while various other issues may be considered as

well.

Subsequently, we propose a dynamic model for

agent negotiation that can be exploited by strategies

in order to construct contracts acceptable to the

opponent parties but which, nevertheless, maximise

the agent’s own utility function. The notation used

by this negotiation model is as follows. The agents

that represent Sellers are denoted by

{}

,...,

SSS =

and the ones that represent potential Buyers are

denoted by

{

}

,...,

BBB

. We introduce the notion

of decision issues (DIs), issues that even though

their values are not under negotiation and they are

not included in the contract parameters, they affect

the evaluation of the values of the contract issues.

Without being exhaustive, such issues may consist

of: the number of competitor companies, the number

of substitute or complementary products/services,

the quantity of product in stock, the number of

current potential buyers, the time until the

negotiation deadline expires, the resources

availability and restrictions, etc. The values of the

DIs may change overtime, depending on the e-

marketplace conditions and on the Seller’s and

Buyer’s state. The values of the DIs are denoted by

fj ,...,1

. We may now introduce the utility

function of the proposed framework as follows. Let

[

]

[

]

1,0,: →

MmU

denote the utility that agent

BSa ∪

∈

assigns to a value of contract issue

the range of its acceptable values. In order for the

utility function of any contract issue

for any

negotiator to lie within the range

[]

1,0

, the value of

issue

i must lie within the range of its acceptable

values. To ensure this, we introduce the notion of

value constraints, that is expressed as follows:

Mcm ≤≤

. In case the value constraints hold for

all contract issues, the utility function can be used to

measure the satisfaction of a negotiator as far as the

proposed contract is concerned. Nevertheless, often,

the value constraints are not met for some contract

issues, thus constituting the contract completely

unacceptable, regardless of the utility level. In this

case, there is not much value in using the above

specified utility function to measure the satisfaction

degree of this negotiator, as the contract is

completely unacceptable. Thus, we may introduce a

value constraint validity vector:

[

]

VCVVCV =

REPUTATION BASED INTELLIGENT AGENT NEGOTIATION FRAMEWORKS IN THE E-MARKETPLACE

ni ,...,1=

, where

{}

1,0∈

VCV

, depending on whether

the value constraint for negotiating party

is met

for contract issue

(i.e.,

VCV

) or not (i.e.,

VCV

). The requirement of mere presence or

absence of a particular feature can be reduced to

value constraints and thus will not be further

analysed.

Let

be the importance of issue

for agent

where

∑

. Using the above notation, the

agent’s

utility function for a contract

{}

knkk

ccC ,...,

can be expressed by the following

equation:

()

∑

jki

dcUwCU

, where

mj ,...,1=

is the value of decision issue

at the

time

, when contract

is proposed. It should be

mentioned that the utility function

()

jki

dcU

may

be of any form (e.g., linear, polynomial, exponential,

quasilinear, etc.), as nonlinear formulations of the

overall utility function do not affect the basic ideas

of the model.

As already mentioned, the BA ranks the contracts

proposed by the SA. For the simplest ranking

function, the ranks that may be assigned to any

contract proposed are boolean variables, i.e., one

instance of the set

{}

rejectaccept,

. A second ranking

scheme may entail the identification of the contract

best suiting the Buyer’s needs without any further

classification of the contracts proposed, while in a

more sophisticated approach, the ranks lie within a

range

[]

Mm ,

, where any contract rated with less

than

is not acceptable by the BA, while, when a

contract is rated with

, then the proposed by the

SA contract is accepted by the BA. In order to signal

the case where at least one value constraint is not

met for the BA for a certain contract, we introduce

another parameter called contract value constraints

validity that will be denoted by

CVCV

for contract

and is given by the following equation:

∏

VCVCVCV

. Based on the previous analysis,

in case all value constraints are met for contract

it stands that

CVCV

. On the other hand, in case

at least one value constraint is not valid for contract

, it stands that

CVCV

, and then the particular

contract is definitely rejected.

Furthermore, the vector of the

1≥N contracts

proposed by the Seller agent

S to the Buyer agent

at time t is denoted by

{

}

CCP ,...,

= , the

vector of the

n contract issues values proposed by

S to

at time

for the k -contract of this proposal

(

Nk ,...,1

) is represented by

{

}

ccC ,...,

= , while

the value of issue

proposed by S to

at time t

for the

k -contract of this proposal is denoted by

(

ni ,...,1

). Let now

{

}

rrR ,...,

= be the vector of

ranking values that

assigns at time t to the

previous contracts proposal made by

S , and

(

Nk ,...,1

) be the rank that

assigns at time

the

k -contract of this proposal. The range of values

acceptable to agent

{}

BSa ,

∈

for issue i is

represented by the interval

[

]

Mm , .

A contract package proposal is accepted by

when at least one contract is rated with

M , while

the negotiation terminates in case the agent(s)

deadline is reached or when a boolean variable

expressing the wish of the agents to quit the

negotiation is set to true. If an agreement is reached,

then we call the negotiation successful, while in case

one of the negotiators quits it is called unsuccessful.

In any other case, we say that the negotiation thread

is active. A detailed presentation of the negotiation

protocol and model adopted can be found in

(Roussaki, 2004a).

2.2 Negotiation Problem and the

Designed Strategies

The objective of the negotiation problem on the

Seller’s side is to find a contract

final

},...,,{

21 nfinalfinalfinal

ccc

that maximises his/her

overall utility function

)(

final

, i.e., the

satisfaction stemming from the proposed contract,

within the negotiation deadlines for both the BA and

the SA. Nevertheless, there are constraints on the

acceptable value ranges that should apply for both

negotiating parties, while their individual utilities

should be above a minimum acceptable threshold

(i.e.,

Accfinal

UCU

min

)( ≥

and

Accfinal

UCU

min

)( ≥

). Based

on the selected protocol and the proposed model,

designing a negotiation strategy can be reduced to a

decision problem on finding the contract package

proposal

{

}

111

,...,

+++

lll

CCP of the N contracts

{

}

111

,...,

+++

lll

ccC ( Nk ,...,1

) that should be

proposed by the SA to the BA in the next round

l ,

given the vector

{

}

lll

CCP ,...,

= proposed by the SA

to the BA during the previous round

l , the vector

{

}

lll

rrR ,...,

= of the ranking values

r (

Nk ,...,1=

)

that the BA assigns to the previously made by the SA

contract proposal at the negotiation round

l and the

value constraint validity vector

{

}

VCVVCV =

(

ni ,...,1

) for at least one of the contracts proposed

subject to the SA’s related constraints and to the

ICE-B 2006 - INTERNATIONAL CONFERENCE ON E-BUSINESS

existent resource and computational limitations.

The complexity of the negotiation problem is

increased with regards to the number of the contract

issues involved and the range of their acceptable

values. In this respect, the design of computationally

efficient algorithms that may provide good (near-

optimal) solutions in reasonable time is required.

A detailed presentation of the proposed

negotiation strategies can be found in (Louta,

2004a), (Louta, 2004b), (Roussaki, 2004b). The

general idea is that all contracts

(

nk ,...,1=

) of a

negotiation round

are generated by the same

“source” contract that will be hereafter denoted as

. All contracts of the same round are generated so

that they correspond to equal utilities for the Seller.

Specifically,

contracts are proposed at each

negotiation round

, which yield the same utility

concession quantity

with respect to the source

contract

. Thus, the utility of the contracts

proposed is equal to

(

)

(

)

dCUdCU

Θ−=

while

(

)

(

)

dCUdCU

−

nk ,...,1=∀

. Utility

concession quantity

has been considered to be

constant and equal to

Acc

)(

min

max

−

for each

negotiation round, where

is the number of

negotiation rounds that could take place before the

SA’s negotiation deadline is reached. It has been

assumed that the values of all decision issues are

invariable for the entire negotiation procedure. It is

noted that in case an agreement between BA and SA

is feasible (that is there exist at least one contract

C for which it stands:

(

)

Acc

UCU

min

≥ and

(

)

Acc

UCU

min

≥ ), our approach will succeed in

identifying a mutually acceptable contract due to the

fact that as its deadline approaches, the SA concedes

up to its reservation value

Acc

min

Based on the RFP sent by the BA, the SA

proposes an initial contract

{

}

000

,...,

ccC =

to the BA

tt =

, setting all contract issues at the values that

maximise the Seller’s utility (i.e., if

(

)

[

]

>∂∂

cdCU

, then the SA sets

Mc =

, while in

case

(

)

[

]

<∂∂

cdCU

, then the SA sets

mc =

The utility of the initial contract

for the SA is

denoted by:

(

)

max

UdCU =

, as

max

is the

maximum utility that can be achieved for the Seller,

given the values of the decision issues

{

}

dd =

time

tt =

. With respect to this initial contract

two distinct cases may be identified. First, no value

constraint violation exists and the Seller aims to find

a contract satisfying the Buyer’s utility constraint.

Second, value constraint violation occurs, in which

case the BA also provides its value constraint

validity vector

VCV

, while the SA, initially tries to

generate a contract that satisfies the BA’s value

constraints. Until a non value constraint violating

contract

is acquired, at each negotiation round

1>l

the source contract

is generated based on the

contract

−l

by distributing the utility concession

amongst the contract issues, whose values are

not acceptable to the BA. This process continues

until a non value constraint violating contract

produced, in which case the SA’s strategy is

modified in order to generate a mutually acceptable

contract within reasonable time.

3 REPUTATION MECHANISM

The establishment of trust is of outmost importance

in the highly dynamic e-marketplace, where small

players emerge and vanish, anyone can choose to be

anonymous, while users may participate in only a

few transactions that may be of relatively low value

and potential contracts may cross jurisdictional

boundaries, raising the difficulty of legal contract

enforcement.

Traditional models aiming to avoid strategic

misbehaviour involve Trusted Third Parties (TTPs)

or intermediaries (Atif, 2002) that monitor every

transaction, which is very costly and sometimes

impossible to apply due to the complexity and the

heterogeneity of the environment. Misbehaviour

means deviation from regular functionality. In the

most general case, it may be unintentional (due to

faults) or intentional in order for selfish parties to

take advantage of certain situations. Reputation

mechanisms are claimed to provide a “softer” notion

of security considered to be sufficient for many

multi-agent applications (Zacharia, 2000). In

essence, they discourage the parties involved from

misbehaving, since the gains expected by future

potential contracts establishment due to a higher

reputation rating can offset the loss incurred by

honouring the transaction terms. Dissemination of

reputation related information to a large number of

negotiating participants may multiply the expected

future gains of honest negotiation parties.

Our study is related to previous pertinent work in

the literature, since reputation based mechanisms is

a quite popular research field, attracting researchers

working in various different areas (Buchegger,

2005)

. In most cases, a reputation based mechanism

is used in order to automatically isolate a

misbehaving party. Thus, the goal of a reputation

system is to enable parties to make decisions on

which parties to negotiate / cooperate with or

exclude, after they have been informed about the

REPUTATION BASED INTELLIGENT AGENT NEGOTIATION FRAMEWORKS IN THE E-MARKETPLACE

reputation ratings of the parties of interest. Feedback

received from negotiating participants related to an

agent’s past behaviour may be formulated as a

reputation measure exploiting learning from

experience concepts. The reputation related

information obtained may be used by the parties in

order to adjust their decisions and behaviour. In this

study, Sellers that are deemed misbehaving are not

directly ostracised, but instead the Buyers’ decision

on the most appropriate Seller is based on a

weighted combination of the evaluation of the

Sellers’ offer quality (performance related factor)

and of their reputation rating (reliability related

factor). The agents may only use first-hand

information, based on their own experiences or they

may additionally exploit second-hand information

disseminated from other parties, which enables them

to identify misbehaving participants early enough.

In Section 3.1 the fundamental concepts of our

proposed collaborative reputation mechanism are

given, while Section 3.2 provides the mathematical

description of the reputation ratings and of the

Buyers’ decision.

3.1 Reputation Rating and Buyer

Decision Fundamentals

Assuming the presence of M SAs negotiating with a

BA for the terms and conditions of the provision of a

product / service, the BA can decide on the most

appropriate SA based on the evaluation of the SA’s

offer quality combined with an estimation of the

SA’s expected behaviour. In our approach this

estimation comprises the reliability related factor,

which is introduced in order to reflect whether the

Seller finally provides to the Buyers the product /

service that corresponds to the established contract

terms or not. The SA’s reliability is reduced

whenever the SA does not honour the agreement

contract terms reached via the negotiation process.

The SAs’ performance evaluation factor is based on

the fact that there may be different levels of

satisfaction with respect to the various SAs’ offers.

In this respect, there may be SAs that, in principle,

do not satisfy the BA with their offer.

The proposed reputation mechanism is

collaborative in the sense that it considers both first-

hand information (acquired from the Buyer’s past

experiences with the Sellers) and second-hand

information (disseminated from other Buyers). To be

more specific, each BA keeps a record of the

reputation ratings of the SAs it has negotiated with.

Additionally, a centralised component called

Reputation Manager (RM), maintains a collective

record of the SAs’ reputation ratings based on the

feedback given by the BAs on their experiences in

the e-market.

True feedback cannot be automatically assumed.

Second-hand information can be spurious (e.g.,

parties may choose to misreport their experience due

to jealousy or in order to discredit trustworthy

Sellers). In general, a mechanism for eliciting true

feedback in the absence of TTPs is necessitated.

According to the simplest possible approach that

may be adopted in order to account for possible

inaccuracies to the feedback provided to the RM by

the BAs (both intentional and unintentional), the BA

can mostly rely on its own experiences rather on the

SAs’ reputation ratings provided by the RM. To this

respect, SAs’ reputation ratings provided by the RM

may be attributed with a relatively low significance

factor. In the context of this study, we consider that

each BA is associated with a predetermined trust

level, which reflects whether the BA reports to the

RM its experiences with the SAs truthfully. To be

more specific, an honesty probability is attributed to

each BA, i.e., a measure of the likelihood that a BA

gives feedback compliant to the real picture

concerning service provisioning. Second-hand

information obtained from trustworthy BAs

(associated with a high honesty probability), are

given a higher significance factor, whereas reports

(positive or negative) coming from untrustworthy

sources have a small impact on the formation of the

SAs’ reputation ratings kept by the RM.

The BA uses the reputation mechanism to decide

on the most appropriate SA, especially in cases

where the BA doubts the accuracy of the information

provided by the SA. A learning period is required in

order for the RM and the BA to obtain fundamental

information for the SAs. In case reputation specific

information is not available to the BA (both through

its own experiences and through the RM) the

reliability related factor is not considered for the

Seller selection. At this point it should be noted that

the reputation mechanism comes at the cost of

keeping reputation ratings related information and

updating it after service provision has taken place.

3.2 Formulation of the Sellers

Reputation Rating System

Each Seller S may be rated in accordance with the

following formula:

)])([)(()()()( SrrESrrRlkSRRSRR

rprepost

−⋅

⋅

(1),

where

post

RR and

pre

RR are the Seller’s S reliability

based rating after and before the updating procedure.

It has been assumed that

post

RR and

pre

RR lie within

the

]1,0[ range, where a value close to 0 indicates a

misbehaving Seller.

)(Srr is a (reward) function

reflecting whether the service quality is compliant

with the picture established during the negotiation

ICE-B 2006 - INTERNATIONAL CONFERENCE ON E-BUSINESS

phase and

)]([ SrrE is the mean (expected) value of

the

)(Srr

variable. In general the larger the

)(Srr

value, the better the Seller behaves with respect to

the agreed terms and conditions of the established

contract, and therefore the more positive the

influence on the rating of the Seller. Factor

(

]1,0(∈

k ) determines the relative significance of the

new outcome with respect to the old one. In essence,

this value determines the memory of the system.

Small

values mean that the memory of the

system is large. However, good behaviour will

gradually improve the Seller’s S reputation ratings.

)(Rl

is a function of the Seller’s reputation rating

pre

RR and is introduced in order to keep the Seller’s

rating within the range

]1,0[ . In the current version

of this study,

)]1exp(1[

)( R

Rl −−⋅

−

, for which it

stands

1)(

→

→R

Rl and 0)(

→

→R

Rl .

It should be noted that Seller’s misbehaviour (or

at least deterioration of its previous behaviour) leads

to a decreased post rating value, since the

)])([)(( SrrESrr − quantity is negative. The )(Srr

function may be implemented in several ways. In the

context of this study, it was assumed without loss of

generality that the

)(Srr values vary from 0.1 to 1.

The reliability rating value of the Seller S is

updated after the user finally accesses the service.

This rating requires in some cases (e.g., when

consumption of network or computational resources

are entailed in the service provision process) a

mechanism for evaluating whether the service

quality was compliant with the picture promised

during the negotiation phase.

The Seller’s S reputation rating may be

calculated by the following formula:

)()()( SRRwSRRwSRR

RMRMBABA

⋅+⋅=

(2),

where

and

are the Seller’s S reputation

information concerning BA experiences and its

collective rating stored by the RM, respectively.

RR is calculated based on equation (1), while

RR is obtained through the following formula:

)])([)(()()()()( SrrESrrBTRlkSRRSRR

rprepost

−⋅

⋅

⋅+=

(3),

where

)(BT is the trust level attributed to the BA. It

stands

]1,0[)( ∈BT with level 1 denoting a fully

trusted BA.

Weights

and

provide the relative value

of the reputation rating of the Seller S as experienced

by BA and the reputation rating of the Seller S as

maintained in the RM component. It has been

assumed that weights

w and

w are normalized

to add up to 1 (i.e.,

1=+

RMBA

), while

BARM

ww <

giving thus a higher significance value to the BA’s

own experiences.

According to the presented approach, the value

w could be close to the value of

w since

potential erroneous decisions (based on fake and

misleading feedbacks) are avoided by incorporating

to the formation of the

values the

trustworthiness of each BA. This way, the limitations

of the simplified approach (e.g., underestimation of

all BAs’ reports, even those reflecting the real

picture) are overcome. At this point it should be

noted that we have assumed that the trustworthiness

of each BA is known and is not modified in the

course of time.

Finally, the BA decides on the most appropriate

Seller S (i.e., the Seller best serving its current

service / product request) and selects the Seller that

maximizes the value of the following formula:

)()( SRRwCUwA

rfinal

pPR

⋅+⋅=

(4)

As you may observe,

is an objective

function that models the performance and the

reliability of the Seller S. Among the terms of this

function there can be the overall anticipated user

satisfaction stemming from the final contract

reached within the negotiation phase, which is

expressed by the function

)(

final

CU with respect to

the contract proposed to the BA and the reputation

rating of the Seller S. Of course, one of the two

factors (anticipated user satisfaction or reputation

rating of the Seller S) can be omitted in certain

variants of the general problem version considered

in this paper. Weights

w and

w provide the

relative value of the anticipated user satisfaction and

the reputation related part. It is assumed that weights

w and

w are normalized to add up to 1 (i.e.,

ww ).

4 CONCLUSIONS

This paper initially presented a dynamic multi-

lateral negotiation model and efficient negotiation

strategies based on a ranking mechanism that

replaces the counter offer complicated scheme. The

proposed framework covers multi-issue contracts

and multi-party situations, while being a highly

dynamic one in the sense that its variables, attributes

and objectives may change over time. The agents

hold private information which may be revealed

incrementally, only on an as-needed basis. The

designed strategies assume that the negotiators face

strict deadlines, which mostly is considered to be

private information. Since e-marketplace is

commonly perceived as an environment offering

REPUTATION BASED INTELLIGENT AGENT NEGOTIATION FRAMEWORKS IN THE E-MARKETPLACE

both opportunities and threats, in order to cope with

negotiating parties’ misbehaviour, as a second step,

the proposed framework is enhanced with a Sellers’

collaborative reputation mechanism, which helps

estimating their trustworthiness and predicting their

future behaviour, taking into account the Sellers’

past performance in consistently satisfying Buyers’

expectations. The reputation mechanism considers

both first-hand information (acquired from the

Buyer’s past experiences with the Sellers) and

second-hand information (disseminated from other

Buyers’ past experiences with the Sellers), while

spurious reputation ratings are taken into account.

The negotiation framework designed has been

adopted by self-interested autonomous agents and

has performed well, always converging to a

mutually acceptable contract, if any, due to the fact

that the Seller concedes to his reservation value as

his deadline approaches. Initial results indicate that

the designed strategies enhanced with the proposed

Sellers’ collaborative reputation mechanism achieve

higher social welfare levels with regards to

reputation independent frameworks, in case there are

Sellers prone to misbehaving. Future plans involve

the incorporation to our model of adaptive trust

ratings for the Buyers without predetermined values

and its extensive empirical evaluation against

existent negotiation and reputation models and

strategies and against the optimal solution that

maximizes the social welfare in multi-party e-

marketplace environments.

REFERENCES

Atif Y., 2002. Building trust in e-commerce, IEEE

Internet Computing Magazine, 6(1), pp. 18-24.

Buchegger S. and Le Boudec J.-Y., 2005. Self-policing

mobile ad-hoc networks by reputation systems, IEEE

Communications Magazine, vol. 43 no. 7, pp. 101-

107.

Conitzer V., Sandholm T., 2003. Computational criticisms

of the revelation principle, In Proc. AAMAS03,

Workshop on Agent Mediated Electronic Commerce

V, Melbourne, Australia.

Faratin P., Sierra C., Jennings N. R., 1998. Negotiation

Decision Functions for Autonomous Agents, Int.

Journal of Robotics and Autonomous Systems, vol.

24, no. 3-4, pp. 159-182.

He M., Jennings N. R., Leung H., 2003. On agent-

mediated electronic commerce. IEEE Transactions on

Knowledge and Data Engineering, vol. 15, no. 4, pp.

985-1003.

Louta M., Roussaki I. and Pechlivanos L., 2004. An

effective Negotiation Strategy for simple buyer

response in E-commerce environment, In Proc. of

IEEE International Conference 'Intelligent Systems'

(IS'04), pp. 535-540, Varna, Bulgaria.

Louta M., Roussaki I. and Pechlivanos L., 2004. An

Efficient Negotiation Strategy in E-Commerce

Context based on Simple Ranking Mechanism, In

Proc. of International Conference on E-Business and

Telecommunication Networks (ICETE 2004), pp. 18-

25, Setubal, Portugal.

Raiffa H., 1982. The Art and Science of Negotiation,

Harvard University Press, Cambridge, USA.

Rosenschein J. S., Zlotkin G., 1994. Rules of Encounter:

Designing Conventions for Automated Negotiation

among Computers, Massachusetts: The MIT Press,

Cambridge, MA, USA.

Roussaki I., Louta M., 2003, Efficient Negotiation

Framework and Strategies for the Next Generation

Electronic Marketplace, MBA Thesis, National

Technical University of Athens.

Roussaki I., Louta M. and Pechlivanos L., 2004. An

Efficient Negotiation Model for the Next Generation

Electronic Marketplace, In Proc. of the 12

IEEE

Mediterranean Electrotechnical Conference

(MELECON 2004), pp. 615-618, Dubrovnic, Croatia.

Roussaki I., Louta M. and Pechlivanos L., 2004.

Negotiation of Intelligent Agents: Dynamic Model and

Contract Ranking Strategy for Electronic Commerce

environments, In Proc. of 4

International Conference

on Intelligent Systems Design and Applications (ISDA

2004), pp. 777-782, Budapest, Hungary.

Rubinstein A., 1982. Perfect equilibrium in a bargaining

model, Econometrica, vol. 50, pp. 97-109.

Zacharia G. and Maes P., 2000. Trust management

through reputation mechanism, Applied Artificial

Intelligence Journal, vol. 14 no. 9, pp. 881-908.

ICE-B 2006 - INTERNATIONAL CONFERENCE ON E-BUSINESS