Modeling Processes for Managing Reputation

Information – A Petri Net Approach

Kirsten Lenz

, Andrijana Mandaric

, Andreas Oberweis

Institute of Information Systems

J.W. Goethe-University

60054 Frankfurt am Main, Germany

Institute of Applied Informatics and Formal Description Methods

University of Karlsruhe

76187 Karlsruhe, Germany

Abstract. Electronic networks bear various potentials often limited through in-

formation asymmetries and opportunism in inter-organizational (business)

processes. To reduce uncertainties when selecting trustful partners, qualitative

goods, and other objects we propose a concept for a so-called "reputation in-

formation management system" that is demand-driven and process-based. Of-

fering decision support by providing reputation information helps reducing un-

certainties in virtual environments. However, instead of focusing on technical

aspects of generating and managing reputation information we concentrate on

modeling the underlying processes. Networked processes and exchange docu-

ments (e.g., reputation information reports) are modeled by using a variant of

high-level Petri nets, so-called XML nets. XML nets allow for an integrated

modeling of the processes and the process relevant XML documents. Due to

their formal foundation, XML nets exhibit potentials for analyzing purposes

and for the direct execution by a respective workflow engine.

1 Introduction

Electronic networks realized through the Internet bear various potentials for the inter-

action or transaction between actors who can be natural/legal persons or autonomous

software agents. Besides pure technical security risks, behavior and quality uncertain-

ties in a priori choice and decision situations lead to problems, e.g., concerning pri-

vacy, authenticity, or fulfillment issues when searching for trustful interaction or

transaction partners or qualitative goods and services. Reasons for uncertainties can

be widely attributed to the lack of information but also to the physical distance be-

tween different actors. Imperfect and uneven distribution of information has the effect

that one actor lacks sufficient high-quality information, a problem which is called the

"agency dilemma" [7] in agency theory. Opportunism as the failure to comply with

implicit and explicit agreements is also a serious problem among actors. Together

with the volatility of virtual environments, these are reasons why processes with dif-

Lenz K., Mandaric A. and Oberweis A. (2004).

Modeling Processes for Managing Reputation Information – A Petri Net Approach.

In Proceedings of the 1st International Workshop on Computer Supported Activity Coordination, pages 136-148

DOI: 10.5220/0002668901360148

 SciTePress

ferent pseudonymous or anonymous actors can be disordered and even may lead to

market failure [5].

Existing approaches for reducing uncertainty range from weakly structured con-

cepts of newsgroups or evaluations in online auctions to structured and more formal-

ized mechanisms of institutions like Trusted Third Parties [15]. Concepts and imple-

mentations of information systems like recommender systems [19], reputation sys-

tems [20], or (collaborative) filtering systems [28] pursuit comparable objectives, but

do not meet individual and changing requirements of its different actors. Existing

systems for generating "reputation information" (RI) are mostly specialized in par-

ticular evaluation objects (e.g., recommender systems in products and reputation

systems in behavioral aspects of persons), information resources (e.g., subjective

evaluations of users or expert evaluations), specific contexts, and domains and thus

do not allow for a versatile use.

Motivated by the lack of integrated concepts for reducing social-economic prob-

lems of uncertainties in inter-organizational electronic networks, we propose a so-

called "reputation information management system". It is demand-driven, process-

based, and flexibly manages evaluations to generate individual "reputation informa-

tion". A requestor for reputation information can individually determine what object

he wants to be evaluated from what information sources (e.g., evaluators) and which

evaluation rules shall be used to generate the requested reputation information. Un-

derlying processes of reputation information management are described by using a

variant of Petri nets, which allow for integrated modeling of processes and the proc-

ess relevant documents and data.

The paper is organized as follows: we start with a description of evaluations, repu-

tation information, and the motivation for applying Petri nets to the management of

reputation information. Section 3 gives a critical overview of related approaches for

reducing uncertainties. Thereupon, we describe the reputation information manage-

ment system for executing individual reputation information requests and evaluations.

In section 5, a part of the evaluation mechanism process is modeled. The paper con-

cludes with a summary and a brief outlook on future work.

2 Reputation Information in Electronic Networks

We start with a short description of evaluations and reputation information. After-

wards, a brief introduction into Petri nets as the appropriate process modeling lan-

guage for the process of generating reputation information is given.

2.1 Evaluations and Reputation Information in Electronic Networks

Evaluations are equally important for commerce purposes for example within trading

communities (like Ebay [11] or Ciao [9]) and for the private use for educational pur-

poses within a knowledge sharing community (like the Reputation Research Network

[18]). In multilateral processes of electronic networks evaluations help to actively

benefit from any evaluated object in the reputation information generating process,

137

even from mouth-to-mouth propaganda [10] when made explicit. The evaluator is an

actor (human, software-agent) that evaluates an object. Central to an evaluation are

the reputation information sources and the evaluation object. The evaluation concerns

for example the behavioral aspect of a person (human, software-agent, or institution)

or the quality of goods, services, or systems.

Evaluations are the basis for the generation of reputation information by providing

(measurable) data. The reputation information management system transforms

evaluations of evaluated objects (through algorithms and/or policies) according to

individual requirements into reputation information and supports the management of

evaluations and reputation information. In accordance to the underlying transforma-

tion rules, reputation information can be either a text, a numerical value, a probabil-

ity, or a complex reputation information value.

The abstract notion of reputation is difficult to understand, but becomes more

comprehensible together with its constituent elements (i.e., information source,

evaluation, and transformation rule). Reputation with its elusive properties such as

multidimensionality, unclear authenticity, non-quantification, and subjective charac-

ter becomes explicable, understandable, and even measurable (for example in terms

of reputation information values). In general, reputation is used in a normative way to

express something positive. Within this work, however, reputation information is

used to describe either a positive or negative estimate of someone or something gen-

erated from multilateral perceptions of other actors. Consequently, reputation infor-

mation is the result of all perceived attributes imputed into any (reputation-)object

from any information source within a specific context of a domain. Reputation infor-

mation can result from experiences, from implicit knowledge (e.g., intuition) as well

as from derivative exogenous experiences and observations of third parties and insti-

tutions. Reputation information sources range from intrinsic information cues (e.g.,

properties of a product) to extrinsic information cues (e.g., product price). Sources

can be material objects like medals, immaterial objects like a brand name, or services,

like guarantees. An academic title, for instance, can serve as a source of reputation

information if brought into an appropriate context. Signaling positive reputation in-

formation reduces uncertainty [22, 23] and thus helps to prevent opportunistic behav-

ior. Within electronic networks this can represent an inherent sanction [15] which

causes a self-disciplining effect thereby preventing from disregarding explicit or

implicit agreements.

2.2 Petri Nets

In the area of process modeling a large number of modeling languages have been

discussed. While semi-formal methods such as event-driven process chains [21] are

popular because of their easy handling, Petri nets combine the advantages of a graphi-

cal language with a formal foundation. They can be also applied for analysis (e.g.,

verification) and simulation purposes [2].

Petri nets are bipartite directed graphs that consist of two disjunctive classes of

nodes: places (depicted by circles) and transitions (depicted by squares). Directed

edges are connecting a place and a transition or vice versa. In elementary Petri nets

[8] places can either be interpreted as conditions (that can be true or false) or as con-

138

tainers for (undistinguishable) objects. Transitions represent activities that depend on

the respective input and output places connected with the transition. The state of the

Petri net is represented by the marking, a set of tokens (depicted by black dots) as-

signed to the respective places. The marking determines which activities can be exe-

cuted and changes with the execution of each activity.

reputation information management system

evaluate

object

evaluation

send

evaluation

sent

evaluation

transform

updated

history

update

algorithm policy

history

...

actor

transition~

edge

place

place with

token

legend:

reputation information management system

evaluate

object

evaluation

send

evaluation

sent

evaluation

transform

updated

history

update

algorithm policy

history

...

actor

transition~

edge

place

place with

token

transition~

edge

place

place with

token

legend:

Fig. 1. Petri net for the evaluator's process

Figure 1 shows a simple Petri net. It describes a part of the evaluation process from

the evaluator's perspective: An actor evaluates an object. He sends the result of his

evaluation to the reputation information management system. The "sent evaluation" is

then transformed according to the predefined algorithm and policy. The existing his-

tory is finally updated.

While elementary Petri nets are not well suited for the description of individual ob-

jects, high-level Petri nets [12] allow for different interpretations of the process rele-

vant objects. In predicate/transition nets [13] places represent relation schemes ac-

cording to which the marking of the net assigns a relation to each place. A transition

represents a class of operations on the relations in the adjacent places. In colored Petri

nets [14] individual objects (e.g., resources, goods, or humans) are distinguished by

assigning identifying tokens. Motivated by the increasing importance of the XML

standard [25] so-called XML nets [16, 17], are proposed. XML nets allow for an

integrated modeling of processes and the process relevant XML documents. They are

related to SGML nets described in [27], that combine Petri nets with the Standard

Generalized Mark-up Language (SGML).

3 Managing Reputation Information

This section gives a critical overview of existing institutions and systems, followed

by the derivation of requirements on the reputation information management system.

139

3.1 Existing Institutions and Systems for Managing Reputation Information

Centralized institutions such as the "Better Business Bureau", "Trustee", and other

independent institutions evaluate objects based on available (central) standards for

quality criteria thereby providing information based on "objective" evaluations. Cer-

tificates and other seals guarantee competence and are supposed to signal trustworthi-

ness according to predefined criteria. The reputation mechanism Sporas [28] which is

used for the implementation of the "Better Business Bureau-Service" is based on the

principle of an evaluation by pairs of transaction partners. The underlying principle is

comparable to that of the reputation database of the online auction Ebay [11]: Both

mechanisms provide for centralized reputation services based on "subjective" experi-

ences instead of standardized quality criteria.

The basic principle of decentralized systems is that so-called (external) "reviewer

agents" provide their expert knowledge. An example for a decentralized trust-based

approach is the internet-based reputation mechanism of Abdul-Rahman and Heiles

[4]. It is a personalized system of reputation relationships where every actor main-

tains his own database of "direct trust relationships" or "recommender trust relation-

ships". Nevertheless, the realization of the mechanisms is not feasible due to the in-

herent necessity of a globally standardized trust category. Furthermore, the system

lacks for assigning temporal information to the recommender and trust values in order

to enhance to management of the system accordingly.

3.2 Requirements on a Demand-Driven Reputation Information Management

System

The use of centralized service institutions seem to be appropriate only for high-value

transactions since it involves high cost. Furthermore, evaluation criteria from inde-

pendent institutions and centralized or decentralized systems often do not correspond

to individual evaluation criteria. A demand-driven and process-based reputation in-

formation management system can comprise own evaluations, ("subjective") experi-

ences and observations, and experiences of third parties. Third parties can provide for

("objectified") experiences and standardized ("objective") expert evaluations or even

generate reputation information.

A prerequisite for the implementation, use, and maintenance of the reputation in-

formation management system is to agree upon common processes within a network.

Communication difficulties between heterogeneous actors using different communi-

cation standards on the one hand and the reputation information management system

on the other hand can thus be prevented. In order to correspond to the variety of re-

quirements of different actors, specific contexts, and domains, the system should be

adaptable and flexibly generate and manage reputation information according to indi-

vidual demands. Additionally, the reputation information management system should

be based on a modular technology.

In order to cope with the problem of subjective evaluations of a single person (in

contrast to objective or simultaneous evaluations of different actors) weights can be

assigned to the respective evaluations. Weights may indicate the importance and

usability of the respective evaluations (for example a lower weighting for evaluations

140

of private actors compared to a high weighting for expert evaluations). However, it is

not trivial to agree upon weights and to assure an identical interpretation of the

weights by all participating actors.

4 System for Reputation Information Management

In this section, we describe the basic concepts of the reputation information manage-

ment system, especially the different roles of participating actors and the architecture

of the reputation information management system.

4.1 Roles in Reputation Information Management

An actor of the reputation information management can be a human or a legal person

or an (autonomous) software agent. Each actor may have different roles such as

evaluator, specialized reputation information requestor, and conventional reputation

information requestor. The conventional reputation information requestor demands

for reputation information that he specifies by interacting with the reputation informa-

tion management system. For the interactive requirement specification, the reputation

information management system provides for adequate functionalities, for example

the communication interface. In contrast to the conventional reputation information

requestor, the specialized reputation information requestor is able to specify his re-

quirements individually by determining the transformation rules and information

sources and thus generates his own reputation information scheme. The reputation

information scheme determines the structure, the context, and the domain of the repu-

tation information. Moreover, existing reputation information schemes can also serve

as a proposal to a conventional reputation information requestor for his reputation

information specification.

The evaluator evaluates objects according to pre-determined instructions. He can

also specify the scheme of the evaluation, for example within a so-called initiative

evaluation. An evaluation is called "initiative" if the evaluator himself determines the

way and the time of evaluating an object and provides this information to the system

without being asked for it. Finally, an actor can also have the role of an evaluation

object or the reputation information source for other evaluators.

4.2 Architecture of the Reputation Information Management System

In the following section, we describe the basic concept of reputation information

management. Figure 2 outlines the interaction of actors in their specific roles with the

reputation information management system and their document-based processes (de-

picted by black arrows). For the exchange, information and data are presented as

documents.

The evaluation mechanism of the reputation information management system starts

with the reputation information request of an actor. The actor can be either a conven-

141

tional or a specialized reputation information requestor. The conventional requestor is

identified by ID_cR comprising IP-address and his email. He requests reputation

information of the target object with an object_ID. The specialized requestor is identi-

fied by ID_sR. His reputation request additionally comprises transformation rules, the

reputation information scheme, and the information source. The reputation informa-

tion management system checks if the required reputation information about the

evaluation object is already stored in the history database or if the reputation informa-

tion can be derived from stored evaluations. The history database contains reputation

information reports (with temporal information) from which the reputation informa-

tion itself, reputation information schemes, transformation rules, and meta-

information about reputation information reports can be extracted. In the simplest

case, the stored reputation information can be sent back to the reputation information

requestor straight away as a new reputation information report (as RIMS service).

Otherwise, existing evaluations have to be examined by the reputation information

management system as to whether or not they can be transformed in order to generate

the requested reputation information. If a transformation is possible, the reputation

information scheme supports the transformation by serving as a specification base

and providing for substantiating rules and methods (algorithms and/or policies). In

the most unfavorable case, neither reputation information nor evaluations are stored

in the history database and the reputation information management system has to send

an evaluation request (comprising the ID of the reputation information management

system and the necessary information of the request) to (potential) evaluators.

REPUTATION INFORMATION

MANAGEMENT SYSTEM

evaluation mechanism

TRANSFORMATION

REPUTATION

INFORMATION

HISTORY

actor

evaluation

request:

<ID_RIMS,

object_ID, trans-

formation rule,

evaluation

scheme,

information

source>

EVALUATOR

evaluation

document:

<ID_E, object_ID,

evaluation scheme,

evaluation

result>

ID_E; object_ID;

evaluation

document:

evaluation scheme

(type, context,

structure, domain);

information source,

evaluation result

actor

RI-request:

<ID_cR,

object_ID>

RIMS service:

<RI-report>

ID_sR (IP- address,

e-mail); RI-request*

(object_ID;

transformation

rules: algorithm,

policy; RI-scheme:

structure, type,

context, domain;

information source)

specialized

RI-REQUESTOR

RI-request*:

<ID_sR, object_ID,

transformation rule,

RI-scheme,

information

sources>

RIMS service:

<RI-report>

conventional

RI-REQUESTOR

ID_cR (IP-

address, email);

RI-request

(object_ID)

RI-query:

<transformation

rule, infor-

mation source>

actor

evaluation

request:

<ID_RIMS,

object_ID, trans-

formation rule,

evaluation

scheme,

information

source>

EVALUATOR

evaluation

document:

<ID_E, object_ID,

evaluation scheme,

evaluation

result>

ID_E; object_ID;

evaluation

document:

evaluation scheme

(type, context,

structure, domain);

information source,

evaluation result

actor

evaluation

request:

<ID_RIMS,

object_ID, trans-

formation rule,

evaluation

scheme,

information

source>

EVALUATOR

evaluation

document:

<ID_E, object_ID,

evaluation scheme,

evaluation

result>

ID_E; object_ID;

evaluation

document:

evaluation scheme

(type, context,

structure, domain);

information source,

evaluation result

actor

RI-request:

<ID_cR,

object_ID>

RIMS service:

<RI-report>

ID_sR (IP- address,

e-mail); RI-request*

(object_ID;

transformation

rules: algorithm,

policy; RI-scheme:

structure, type,

context, domain;

information source)

specialized

RI-REQUESTOR

RI-request*:

<ID_sR, object_ID,

transformation rule,

RI-scheme,

information

sources>

RIMS service:

<RI-report>

conventional

RI-REQUESTOR

ID_cR (IP-

address, email);

RI-request

(object_ID)

RI-query:

<transformation

rule, infor-

mation source>

actor

RI-request:

<ID_cR,

object_ID>

RIMS service:

<RI-report>

ID_sR (IP- address,

e-mail); RI-request*

(object_ID;

transformation

rules: algorithm,

policy; RI-scheme:

structure, type,

context, domain;

information source)

specialized

RI-REQUESTOR

RI-request*:

<ID_sR, object_ID,

transformation rule,

RI-scheme,

information

sources>

RIMS service:

<RI-report>

conventional

RI-REQUESTOR

ID_cR (IP-

address, email);

RI-request

(object_ID)

RI-query:

<transformation

rule, infor-

mation source>

RI-REQUEST

EVALUATION

Fig. 2. Interaction between actors and the reputation information management system for the

execution of reputation information requests and evaluations

The evaluator can either be an (institutionalized) expert (comparable to a rating

agency like Standard & Poor’s [24]) or a non-expert that evaluates the required object

according to respective instructions. He is identified by ID_E and assigns the evalua-

tion document (comprising information such as evaluation scheme, information

142

source, and evaluation result) to the target object. Instructions are delivered within the

document of an evaluation request. They contain a detailed description of the evalua-

tion process and the evaluation scheme (according to the reputation information

scheme of the reputation information requestor by "inverting" the transformation

rules). Finally, the generated evaluation document is sent to the reputation informa-

tion management system. The evaluation mechanism of the reputation information

management system in turn transforms the evaluation into the required reputation

information and delivers the reputation information report to the reputation informa-

tion requestor. If evaluations are invoked by other reasons than concrete reputation

information requests (e.g., good will of the actor or monetary incentives of the sys-

tem), the evaluator himself can specify the evaluation scheme. For such an initiative

evaluation, the evaluation document includes the evaluation scheme and thus allows

for assigning this evaluation to potential reputation information requests in the future.

5 Reputation Information Management Processes

Instead of focusing on technical issues, we will emphasize the process aspects. The

example in Figure 1 has illustrated that elementary Petri nets are not well suited for

the modeling of document-based processes.

5.1 XML Nets for the Description of Document-Based Processes

All information that is exchanged between actors and the reputation information

management system has a structure and thus can be represented as an XML docu-

ment. XML schemes exactly define the structure of the respective documents and are

subject to reputation information management (see [25] for XML and [26] for XML

Schema). For an integrated description of evaluation processes and its relevant docu-

ments we propose XML nets [17], a variant of high-level Petri nets. In XML nets,

schemes of documents can be graphically specified for documents like evaluations or

reputation information reports.

legend:

element

attribute

association

1..*

at least 1~

element

filter

arbitrary

evaluation

object

EO-No

name

instruction

annotation duration

policy

P-No

for evaluation

E-No

instruction

list (P)

1..*

algorithm

A-No

requirement

for evaluation

EV-No

calculation annotation duration

instruction

list (A)

1..*

RI-report

RNo

status

time

contact

name

address

status

legend:

element

attribute

association

1..*

at least 1~

element

filter

arbitrary

evaluation

object

EO-No

name

instruction

annotation duration

policy

P-No

for evaluation

E-No

instruction

list (P)

1..*

algorithm

A-No

requirement

for evaluation

EV-No

calculation annotation duration

instruction

list (A)

1..*

RI-report

RNo

status

time

contact

name

address

status

143

Fig. 3. Graphical XML scheme for the XML documents representing reputation information

reports

Figure 3 shows a graphical XML scheme of a reputation information report. In ac-

cordance to this scheme, a reputation information report with its identifier "RNo"

consists of the following elements: the "contact", the evaluation object, the appropri-

ate transformation rules consisting of the "policy" and the "algorithm", the reputation

information, and the date.

In XML nets, XML schemes define the types of the places they are assigned to. As

a result, the places can be seen as containers of XML documents according to the

respective scheme, e.g., reputation information reports or descriptions of physical

objects. The direction of the edge determines whether the relevant documents are

input or output documents of the corresponding activity. The edge inscriptions, so-

called filter schemes, allow for filtering relevant documents from the set of available

documents and for determining how to manipulate documents during the execution of

the activity. A filter scheme may contain a manipulation filter, depicted by a black bar

on the left side of the element (see Figure 4). The manipulation filter defines the ma-

nipulation: for example the deletion of old evaluations, insertion of a new evaluation

or updating of the reputation information report. The manipulation operation can be

applied either to the whole document or to a part of the document. A filter scheme

without manipulation filter expresses that the respective document (e.g., the policy) is

only read.

update

RI-report

evaluation

E-No

contact

evaluation

value

evaluation

object

EO-No

name

address

name

status

value time

status

instruction annotation duration

policy

P-No

for evaluation

E-No

instruction list

(P)

1..*

algorithm

A-No

requirement

for evaluation

EV-No

calculation annotation

duration

instruction

list (A)

1..*

transform

E‘

E‘‘

EV EO

EON

RIC Date

EON

...

legend:

element

attribute

manipula-

tion filter

association

1..*

at least 1~

element-

filter

arbitrary

element

place holder

EON

history

H-No

RI-No

...

update

RI-report

evaluation

E-No

contact

evaluation

value

evaluation

object

EO-No

name

address

name

status

value time

status

instruction annotation duration

policy

P-No

policy

P-No

for evaluation

E-No

instruction list

(P)

1..*

algorithm

A-No

requirement

for evaluation

EV-No

calculation annotation

duration

instruction

list (A)

1..*

transform

E‘

E‘‘

EV EO

EON

EV EO

EON

RIC DateEO

EON

RIC Date

EON

...

legend:

element

attribute

manipula-

tion filter

association

1..*

at least 1~

element-

filter

arbitrary

element

place holder

EON

history

H-No

RI-No

...

history

H-No

RI-No

...

Fig. 4. XML net for the generation of a reputation information report

Figure 4 shows an XML net describing the part of the evaluation mechanism that

generates reputation information reports (as a document-based and therefore more

detailed description compared to the process in Figure 3). Relevant documents for the

execution of the transformation are "evaluation", "policy", "algorithm", and "reputa-

tion information report". The respective schemes (except that for the reputation in-

formation report which is omitted due to space limitation) are shown in Figure 4. The

144

activity "transform" describes the transformation of an evaluation and the composi-

tion of the document for the reputation information report. For a detailed description

of the activities during the transformation, the activity itself can be replaced by an

appropriate detailed XML net. An adequate use of variables of the edge inscriptions,

especially the one for the identifier of the evaluation object, ensures that only the

corresponding policy and algorithm are taken into account for the transformation of

the evaluation object.

The example illustrates that XML nets allow for a document-based process model-

ing. Their expressive power and formal semantics represent the basis for a systemati-

cal analysis of the process schemes. XML nets are a suitable means for the visualiza-

tion and communication of processes (e.g., between experts and other actors), espe-

cially since the graphical representation can be adapted to specific skills and require-

ments. Moreover, XML nets support the execution of processes and therefore allow

for a better control of (semi-) automatic process executions.

5.2 Supporting Inter-Organizational Collaboration

An integration of reputation information generating processes of the reputation in-

formation management system to the (business) processes of its participating actors

allows for a step towards frictionless reputation information management. The ex-

change of reputation information and evaluation schemes together with a customiza-

ble representation results in a seamless collaboration between different actors and the

reputation information management system. Especially enterprises with streams of

requests and responds (e.g., simultaneous evaluations of several objects or of the

same object from different reputation information sources) can benefit from the inte-

gration of processes into their own business processes.

Despite the advantages of inter-organizational processes, difficulties can result

from integration, access, or execution if complex processes are not controlled. Work-

flow management systems allow for making complex application logic explicit where

processes are modeled in a high-level (typically graphical) language rather than en-

coded in a programming language. Due to the formal semantics Petri nets are suitable

for the graphical specification of workflows [1]. XML nets can be interpreted as

workflow descriptions. If an activity that is supposed to be accomplished is initiated,

the underlying workflow is automatically triggered by the workflow-engine. For

controlling the performance of the respective reputation information processes, so-

called fact transitions [13] are a suitable means. Moreover, they allow for the (graphi-

cal) declaration of rules that have to be obeyed during the process execution.

Besides the uncertainty reducing signaling effect of reputation information itself

there is an inherent sanctioning mechanism within the electronic network. Instead of

an explicit sanctioning instance, actors among themselves will prevent other partici-

pants to behave in an opportunistic manner within the electronic network. Inter-

organizational collaboration, however, also requires that each actor makes his process

or parts of it transparent [3] to the reputation information management system. Better

comprehension and understanding of each others' processes leads to a better demand-

oriented reputation information service. Moreover, the reputation information man-

agement system should be able to integrate arbitrary types of reputation information

145

requestors, consider different reputation information sources and different transfor-

mation rules for demand-specific requests.

6 Summary and outlook

The paper describes research work from an ongoing project which aims at developing

an architecture of a reputation information management system. In this paper we have

proposed a concept for generating and managing reputation information for reducing

behavior and quality uncertainties in electronic networks. Rather then coping with the

technical aspects of the system, we have concentrated on modeling aspects. We have

described evaluations and reputation information and gave a brief introduction into

Petri nets. A critical overview of existing concepts and systems for reducing uncer-

tainty led to the development of requirements on the reputation information manage-

ment. Disadvantages of existing concepts and systems, especially their restriction to

specific application domains, have motivated the introduction of a demand-driven,

process-based reputation information management system. The reputation informa-

tion management system is responsible for generating and managing reputation in-

formation, e.g., for handling evaluations or reputation information according to user

specific requirements. The concept of the reputation information management system,

especially the evaluation mechanism and different roles of actors were presented.

Processes of the reputation information management were described. Thereafter, a

process part and the relevant documents were modeled as an XML net.

Uncertainties in electronic networks that are less a result of technical or environ-

mental but rather of endogenous behavior and quality uncertainties between actors

can be reduced by providing indicative reputation information. The reputation infor-

mation management system helps to generate and manage reputation information in

order to deliver individual, demand-specific reputation information reports. It is de-

mand-driven, personalized, and adapts to different domains and contexts of individual

requirements of actors in different roles.

It is not necessary that reputation information has to be generated within the

evaluation mechanism of reputation information management system. In order to

enhance the interchange of relevant documents and data in electronic networks it is

necessary to develop appropriate interfaces for the integration of the processes. Spon-

taneous (especially short run) collaboration can be supported by making available a

repository of process fragments to the involved actors. The repository consists of

process schemes for relevant process parts that can be delivered to the requesting

actor on demand. The implementation of such a repository is facilitated by using a

common document-based process modeling language such as XML nets.

A prototype version of the reputation information management system is currently

under development.

146

References

1. van der Aalst, W. M. P.: The Application of Petri Nets to Workflow Management. In: The

Journal of Circuits, Systems and Computers, Vol. 8 (1) (1998) 21–66

2. van der Aalst, W. M. P., Desel, J., Oberweis, A. (eds.): Business Process Management –

Models, Techniques, and Empirical Studies, Lecture Notes in Computer Science, Vol. 1806,

Springer-Verlag, Berlin Heidelberg New York (2000)

3. van der Aalst, W. M. P., Kumar, A.: XML Based Schema Definition for Support of Inter-

organizational Workflow. In: Information Systems Research, Vol. 14 (1) (2003) 151–162

4. Abdul-Rahman, A.; Heiles, S.: Using Recommendations for Managing Trust in Distributed

Systems. In: Proc. IEEE Malaysia International Conference on Communications, Kuala

Lumpur, Malaysia (1997)

5. Akerlof, G.A.: The Market for Lemons: Quality Uncertainty and the Market Mechanism. In:

Quarterly Journal of Economics, 84 (3) (1970) 488–500

6. Alonso, G., Casati, F., Kuno, H., Machiraju, V.: Web Services: Concepts, Architectures and

Applications. Springer-Verlag, Berlin (2004)

7. Arrow, K.J.: The Economics of Agency. In: Pratt, J., Zeckhauser, R. (eds.), Principles and

agents: The structure of business, Harvard University Press, Boston (1985) 37–51

8. Brauer, W., Reisig, W., Rozenberg, G. (eds.): Advances in Petri Nets, Part I, Lecture Notes

in Computer Science, Vol. 254, Springer-Verlag, Berlin Heidelberg New York (1986)

9. Ciao, http://www.ciao.com

10. Dasgaputa, P.: Trust as a Commodity. In: Gambetta, D. (ed.): Trust Making and Breaking

Cooperative Relations, New York et al. (1988) 49–72

11. Ebay, http://www.ebay.com

12. Ehrig, H., Hoffmann, K., Padberg, J., et al.: High-level Net Processes. In: Brauer, W.,

Reisig, W., Rozenberg, G. (eds.): Formal and Natural Computing, Lecture Notes in Com-

puter Science, Vol. 2300, Springer-Verlag, Berlin Heidelberg New York (2002) 191–219

13. Genrich, H.J.: Predicate/Transition Nets. In: Brauer, W., Reisig, W., Rozenberg, G. (eds.):

Petri Nets: Central Models and Their Properties, Advances in Petri Nets – Part I, Lecture

Notes in Computer Science, Vol. 254, Springer-Verlag, Berlin Heidelberg New York (1987)

207–247

14. Jensen, K.: Coloured Petri Nets, Vol. 1 – Basic Concepts, Berlin (1992)

15. Kollock, P.: The Production of Trust in Online Markets. In: Lawler, E.J., Macy, M., Thyne,

S., Walker, H.A. (eds.): Advances in Group Processes, Vol. 16 (1999)

16. Lenz, K.: Modellierung und Ausführung von E-Business-Prozessen mit XML-Netzen.

Verlag für Wissenschaft und Forschung, Berlin (2003) (in German)

17. Lenz, K., Oberweis, A.: Inter-organizational Business Process Management with XML

Nets. In: Ehrig, H., Reisig, W., Rozenberg, G.; Weber, H. (eds.): Petri Net Technology for

Communication-Based Systems, Lecture Notes in Computer Science, Vol. 2472, Springer-

Verlag, Berlin Heidelberg (2003) 243–263

18. Reputation Research Network, http://databases.si.umich.edu/reputations/

19. Resnick, P.; Varian, H.R.: Recommender Systems. In: Communications of the ACM, 40 (3)

(1997) 56–58

20. Resnick, P.; Zeckhauser, R.; Friedman, E.; Kuwabara, K.: Reputation Systems. In: Com-

munications of the ACM, 43 (12) (2000) 45–48

21. Scheer, A.-W.: Business Process Modeling, 2

edition, Berlin et al. (1999)

22. Shapiro, C.: Premiums for High Quality Products as Returns to Reputations. In: Quarterly

Journal of Economics, Vol. 97 (1983) 659–679

23. Spence, M.: Informational Aspects of Market Structure: An Introduction. In: Quarterly

Journal of Economics, Vol. 90 (4) (1976) 591–597

24. Standard & Poor's, http://www.standardandpoors.com

147

25. World Wide Web Consortium: Extensible Markup Language, http://www.w3.org/XML

26. World Wide Web Consortium: XML Schema, http://www.w3.org/XML/Schema

27. Weitz, W.: Combining Structured Documents with High-level Petri-Nets for Workflow

Modeling in Internet-based Commerce. In: International Journal of Cooperative Information

Systems (IJCIS), Vol. 7 (4) (1998) 275–296

28. Zacharia, G.; Moukas, A.; Maes, P.: Collaborative Reputation Mechanisms in Electronic

Marketplaces. In: Proc. 32

Hawai'i Intl. Conference On System Science (1999) 1–7

148