NORMATIVE CONFLICTS

Patterns, Detection and Resolution

Georgios K. Giannikis and Aspassia Daskalopulu

Department of Computer and Communications Engineering

University of Thessaly, Gklavani 37, 38221 Volos, Greece

Keywords: Conflict Patterns, Conflict Detection, Conflict Resolution, e-Contracts, Default Logic.

Abstract: The analysis, representation and management of normative conflicts have been the focus of much research

in recent years in commercial and business applications. In this paper we are concerned with normative

conflicts that arise for agents engaging in electronic contracting. First, we identify a set of primitive conflict

patterns and present some patterns that have not been identified in other proposals. Secondly, we use a

representation of e-contracts as Default Theories, which afford us both detection and resolution of such

conflict patterns.

1 INTRODUCTION

In this work we are concerned with normative

conflicts that arise for agents engaging in electronic

contracting, within an electronic marketplace, and

we investigate an alternative representation in which

we use Reiter’s Default Logic (DfL) (Reiter,1980).

In (Giannikis & Daskalopulu, 2006, Giannikis &

Daskalopulu, 2007b) we proposed the representation

of contractual norms as default rules, which are

constructed dynamically from temporal

representations. The resulting default theories afford

us both temporal defeasible reasoning and conflict

management (Giannikis & Daskalopulu, 2006).

Here, we identify a set of primitive patterns for

normative conflicts and show how the conflicts

identified by other researchers may be seen as

instances of these primitives. We also identify some

patterns of normative conflict that have not been

identified in other proposals. Furthermore, we

discuss the way the representation of contractual

norms as default rules facilitates conflict detection

and resolution.

2 CONFLICT DETECTION

For the purposes of illustration consider an

electronic marketplace, populated by software

agents that establish and perform e-contracts on

behalf of some real world parties. Let the set

Agents={Agent1, Agent2, Agent3,…..} denote distinct

identifiers for the various agents, and the set

Roles={RA, WA, MA, CA, …} denote distinct roles that

agents may assume in the e-market (where

RA, WA,

MA, CA denote retailer, wholesaler, mediator and

carrier respectively).

Consider a two-party business transaction.

Agent1

that acts as a retailer orders some goods from the

wholesaler

Agent3. The terms of the agreement

between these two agents are:

Agent3 should see to it

that the goods be delivered to

Agent1 within 10 days

from commencement (e.g., the date that the order

takes place).

Agent1, in turn, should see to it that

payment be made within 21 days from the date it

receives the goods. If

Agent3 does not deliver on time,

then a fixed amount is to be deducted from the

original price of the goods for each day of delay and

it should see to it that delivery be made by a new

deadline. If

Agent1 does not perform payment on

time, then a fixed amount is to be added to the

original price of the goods for each day of delay and

it should see to it that payment be made by a new

deadline.

Following (Daskalopulu, 2000), we may take an

informal, process view of the business transaction

that is regulated by the agreement. Each state offers

a (possibly partial) description of the factual and

normative propositions that hold true. A transition

between states corresponds to an event that takes

place, i.e. an action that one of the parties performs

527

Giannikis G. and Daskalopulu A.

NORMATIVE CONFLICTS - Patterns, Detection and Resolution.

DOI: 10.5220/0001835505270532

In Proceedings of the Fifth International Conference on Web Information Systems and Technologies (WEBIST 2009), page

ISBN: 978-989-8111-81-4

or omits to perform. Normative propositions of the

form:

ΝN(agent1, role1, action, agent2, role2)

express that agent1 that acts as role1 is in legal relation

ΝN towards agent2 that acts as role2 to perform action,

where ΝN may be Obligation, Prohibition, Permission and

legal Power.

We use Reiter’s Default Logic (Reiter, 1980) to

represent the norms of an agreement as default rules.

A default rule has the form

P:J1,J2,…Jn/C, where P is

the prerequisite,

J={J1,J2,… Jn} is a set of justifications

and C is the derived consequent. If J coincides with

C, the default rule is called normal. The semantics of

a default rule is: If

P holds and the assumption J is

consistent with our current knowledge, then C may

be inferred.

For instance, the following default rule expresses

that if an order from

Agent1 (acting as a retailer)

towards Agent3 (acting as a wholesaler) holds, and it

is consistent to assume that

Agent1 will become a

regular client, then we may infer that Agent3 is legally

obliged towards Agent1 to perform delivery:

Order(Agent1, RA, Agent3, WA)

BecomeRegularClient(Agent1)

Obligation(Agent3, WA, Delivery, Agent1, RA)

A Default Theory (DfT) is a pair of the form (W,

, where W is a set of closed formulae that hold, and

D is a set of defaults. Rules may be used to compute

extensions

E of the default theory. A rule is

applicable to a set of formulae W⊆E if and only if P

∈

and ¬J

,…, ¬J

∉

E. We consider grounded DfTs and we

derive extensions in the manner presented in (Error!

Reference source not found.), i.e. by maintaining

consistent sets of formulae. This derivation may be

conducted in stepwise manner. Thus, an agent that

engages in a transaction governed by some

agreement, essentially reasons with a default theory.

At each time point during the business transaction

the agent attempts to compute the extensions of its

current DfT. Note that whenever multiple extensions

are computed for a Default Theory these represent

possible world views. Depending on its chosen

action an agent is committed to a particular

extension. The DfT contract representation allows us

to detect normative conflicts by examining

extensions. A normative conflict may be detected

either between multiple extensions or between some

extension and the current knowledge (

W) of the

agent. Where a conflict is detected between multiple

extensions, the latter represent alternative futures for

the agent; let us call these inter-extension conflicts.

Where a conflict is detected between an extension

and the current knowledge of the agent, it represents

a state in which some normative violation will

eventually occur; let us call these intra-extension

conflicts. The role of conflict detection is, thus, to

assist an agent to choose a course of action so that

normative violations may be predicted and avoided.

The first step of conflict management involves

the detection of conflicts. To this end, in section 2.1,

we identify primitive patterns of normative conflict

that may be spotted during the derivation of

extensions of the default theory representation of an

agreement. In section 2.2 we identify additional

cases of normative conflicts which are not discussed

already in the existing literature.

2.1 Primitive Patterns of Normative

Conflicts

In what follows we use Obligation, Permission, Prohibition

and Power as predicates that express normative

relations between agents. We do not employ the

axiomatization of any particular system of Deontic

logic; specifically, we do not employ the

axiomatization of Standard Deontic Logic (SDL), in

which these notions are modeled as operators that

are inter-defined. This is because in Standard

Deontic Logic (and any system where the D scheme

¬O⊥, where O denotes obligation, is valid) it is not

possible for an agent to bear conflicting obligations

because of the D scheme. Yet, in most realistic

situations, indeed in our everyday life, agents do

find themselves in normative conflict. Moreover, if

we were to employ SDL, permission, obligation and

prohibition would be interdefined, and so all of the

patterns we present in this section (section 2.1)

would be reduced to three of all six patterns

(Giannikis & Daskalopulu, 2007a); thus the

representation would be less distinguishing.

A. Conflict between a normative notion (

NN) and

its negation. The general pattern is:

NN(agent1, role1, action, agent2, role2)

¬NN(agent1, role1, action, agent2, role2)

This is the common syntactical conflict that

arises when an agent has contradictory

knowledge. All other approaches, without any

exception, refer to this type of conflict. In

policy-based approaches, when the normative

notion is Obligation it is called positive-

negative conflict of modalities (Moffett &

Sloman, 1993). This type of conflict never

actually arises in our representation, where

WEBIST 2009 - 5th International Conference on Web Information Systems and Technologies

528

norms are represented as defaults, because the

derivation of extensions preserves consistency.

It may, however, be identified as a potential

conflict, when multiple extensions are

computed.

B. Conflict between the prohibition to perform an

action and the simultaneous permission or

obligation to perform the same action. Once

again, all previous research approaches refer to

this type of conflict. In (Moffett & Sloman,

1993) and (Lupu & Sloman, 1999) these

conflicts are called conflicts between authority

policies (sub-pattern B1: Prohibition vs

Permission) and conflict between authority and

imperatival policies (sub-pattern B2:

Prohibition vs Obligation) respectively.

Consider, for instance, the following default

theory

(W, D) where:

W={Order(Agent1, RA, Agent3, WA)}

and D={

Order(Agent1, RA, Agent3, WA)

WellKnownDebtor(Agent1)

Prohibition(Agent3, WA, Delivery, Agent1, RA)

Order(Agent1, RA, Agent3, WA)

Permission(Agent3, WA, Delivery, Agent1, RA)

}

Permission(Agent3, WA, Delivery, Agent1, RA)

The first default denotes that if an order from

Agent1 (acting as retailer) towards Agent3 (acting

as wholesaler) holds, and it is consistent to

assume that

Agent1 is related to a well known

debtor then we may infer that

Agent3 is

prohibited to perform delivery. Similarly, the

second default expresses that if an order from

Agent1 towards Agent3 holds, and it is consistent to

assume that

Agent3 is permitted to perform

delivery, then we may infer that

Agent3 is

permitted to perform delivery towards

Agent1.

Agent3 may find itself in a conflicting state after

applying the two defaults sequentially. We

denote this type of conflict as B1. Note that

special terms, such as

WellKnownDebtor(agent),

BecomeRegularClient(agent) or IsRegularClient(agent)

among others, are used only for the purposes of

illustration and are not binding to the

characterization of domain-independent conflict

patterns.

In the same spirit, let us replace the second

default shown above with the following:

Order(Agent1, RA, Agent3, WA)

Obligation(Agent3, WA, Delivery, Agent1, RA)

Once again Agent3 is in conflict. We denote this

conflict between Prohibition and Obligation as

B2.

C. Conflict between an obligation to perform

action

and the simultaneous obligation or permission

to perform

¬action. Here ¬action denotes a negative

action, and the issue of representing negative

actions has concerned researchers (e.g.

(Royakkers, 1998) regards them as actions that

do not lead to the successful fulfillment of a

norm). We have not developed special

semantics for the representation of negative

actions; we merely regard such expressions as

denoting either performance of some action

other than the negative one, or as idleness (non

performance of any action). This case arises,

also, in Lee (Lee, 1988) and Abrahams

(Abrahams & Bacon, 2002) who use the term

Waive.

For example consider the following DfT where:

W={Order(Agent1, RA, Agent3, WA)}

and D={

Order(Agent1, RA, Agent3, WA)

BecomeRegularClient(Agent1)

Obligation(Agent3, WA, Delivery, Agent1, RA)

Order(Agent1, RA, Agent3, WA)

WellKnownDebtor(Agent1)

}

Obligation(Agent3, WA, ¬Delivery, Agent1, RA)

D. Conflict between the power to perform an

action and the simultaneous prohibition to

perform the same action. This type of conflict is

also noted in (Abrahams & Bacon, 2002).

For instance consider the following DfT:

W={Order(Agent1, RA, Agent3, WA)}

and D={

Order(Agent1, RA, Agent3, WA)

Power(Agent3, WA, Delivery, Agent1, RA)

Order(Agent1, RA, Agent3, WA)

WellKnownDebtor(Agent1)

}

Prohibition(Agent3, WA, Delivery, Agent1, RA)

NORMATIVE CONFLICTS - Patterns, Detection and Resolution

529

One may argue that in this case there is no

conflict and, consequently, that there is no need

for conflict resolution. Indeed, legal power to

perform an action goes hand-in-hand with

permission to exercise it, according to formal

definitions of institutional power ((Makinson,

1986, Jones & Sergot, 1996)). Hence, there is a

conflict here, albeit some may perceive it as a

conflict between permission and prohibition to

exercise a certain power.

E. Conflict between two obligatory distinct

actions, when it is impossible to do both at the

same time. This corresponds to Horty’s moral

dilemma (Horty, 1994).

For instance consider the following DfT where:

W={Order(Agent1, RA, Agent3, WA),

Order(Agent2, RA, Agent3, WA),

no simultaneous performance of actions is possible}

and D={

Order(Agent1, RA, Agent3, WA)

BecomeRegularClient(Agent1)

Obligation(Agent3, WA, Delivery1, Agent1, RA)

Order(Agent2, RA, Agent3, WA, T1)

IsRegularClient(Agent1)

}

Obligation(Agent3, WA, Delivery2, Agent2, RA)

Agent3 bears two obligations that cannot be

simultaneously satisfied.

F. Conflict between an obligation and the negation

of the agent’s permission or power to perform it.

The negation of an agent’s permission/power to

perform an action may be explicitly derived

from the agent’s knowledge base (sub-pattern

F1) or it may be derived from a possibly

incomplete knowledge base, through the

absence of explicit information (sub-pattern F2).

For instance consider the following default

theory where:

W={Order(Agent1, RA, Agent3, WA)}

and D={

Order(Agent1, RA, Agent3, WA)

BecomeRegularClient(Agent1)

Obligation(Agent3, WA, Delivery, Agent1, RA)

Order(Agent1, RA, Agent3, WA)

WellKnownDebtor(Agent1)

}

¬Permission(Agent3, WA, Delivery, Agent1, RA)

Now consider a DfT that contains the first of the

defaults above and in place of the second, the

following:

Order(Agent1, RA, Agent3, WA)

¬Permission(Agent3, WA, Delivery, Agent1, RA)

If the agent’s knowledge base does not contain

an explicit permission, then the justification of

this default will be satisfied, and hence its

conclusion will be drawn.

2.2 Additional Patterns

Here are some additional cases of normative conflict

that we have identified, which are not discussed

already in the existing literature. We mention them

separately because, although they may be reduced to

the primitive patterns, there is additional information

that may be exploited to facilitate conflict resolution.

Type of action-based conflicts

. A common

feature of e-contracts is the so called Contrary-to-

Duty structures (Prakken & Sergot, 1996). An

agent’s contractual obligations may be divided in

two types. Prima facie obligations that concern the

performance of actions that are in principle

stipulated by the agreement and secondary

obligations that concern the performance of

reparatory actions; the latter apply only when

violations of prima facie obligations happen.

An agent may, thus, bear two distinct obligations

(for instance of the kind described by E), where one

is primary and the other is secondary. This

qualification may be helpful in conflict resolution.

The general pattern is:

Obligation(agent1, role1, action, agent2, role2)

Obligation(agent1, role1, reparatoryaction, agent3, role3)

Agreement-based conflicts. An agent may find

itself in a conflicting state because it is engaged in

multiple contracts. For instance a wholesaler may be

obliged to perform two distinct deliveries to two

distinct retailers as dictated by two distinct

agreements. This situation may be regarded as the

generalization of pattern E discussed earlier, because

in this case the important information is the

distinction between the contracts. The additional

information that the two norms stem from two

agreements, may be exploited for the purposes of

conflict resolution. The general pattern is:

Obligation(contract1, agent1, role1, action1, agent2, role2)

Obligation(contract2, agent1, role1, action2, agent3, role3)

WEBIST 2009 - 5th International Conference on Web Information Systems and Technologies

530

where normative propositions of the form

ΝN(contract,

agent1, role1, action, agent2, role2)

express that according

contract, agent1 that acts as role1 is in legal relation

ΝN towards agent2 that acts as role2 to perform action.

Note that this conflict pattern is different form

the one presented in (Herrestad & Krogh, 1995). The

key notion here is the different contracts an agent

has to comply with. Different contracts may be

established towards different agents or even towards

the same agent.

Conflicts between assumptions and knowledge

A conflict may arise not only as a result of an

agent’s explicit knowledge but also between its

knowledge and its current assumptions or even

between distinct assumptions.

For example, according to the following DfT the

prohibition that derives from the second default

contradicts not only with obligation that derives

from the first default, but also with the assumption

of the first default (

Permission):

W={Order(Agent1, RA, Agent3, WA)}

and D={

Order(Agent1, RA, Agent3, WA)

Permission(Agent3, WA, Delivery, Agent1, RA)

Obligation(Agent3, WA, Delivery, Agent1, RA)

Order(Agent1, RA, Agent3, WA)

WellKnownDebtor(Agent1)

}

Prohibition(Agent3, WA, Delivery, Agent1, RA)

3 CONFLICT RESOLUTION

Conflict resolution in DfL may be performed using

Brewka’s (Brewka, 1994) proposal that enables us to

define and apply priorities on default rules

dynamically.

Brewka defines a Preferential Default Theory

(PDfT) as a triple

(W, D, name), where name is a

function that assigns names to default rules D. The

extension of a PDfT is derived in the same way as in

a DfT.

What makes PDfTs really useful is that the

ascription of priorities to default rules may, itself, be

done dynamically. Using dynamic priorities, we

generate preferred extensions, each of which

indicates a transaction plan. Priorities amongst

ground defaults may be defined dynamically either

by making different assumptions or by specifying

domain-dependent criteria. The general pattern for

ascribing priorities dynamically takes the form of a

default rule:

Rule(d1, v1)

∧

Rule(d2, v2)

∧

criterion

Assumptions

d1<d2

Here d1, d2 are variables that denote names of ground

defaults; Rule(d1,v1) denotes a ground default d1 and

its set of entities of interest

v1. The intended

interpretation of this rule is: if two defaults d1 and d2

apply and some criterion is satisfied between entities

of interest, then

d1 takes priority over d2, if certain

assumptions may consistently be made.

Three general strategies for defining such

criteria have been discussed in the literature, namely

hierarchies of entities of interest, time and

specificity of norms. Thus, given a particular

normative conflict, different resolution strategies

may be applied depending on our specific criterion

of interest.

In (Giannikis & Daskalopulu, 2006) we

presented a temporal representation of normative

relations in DfL, which takes into account the

external time of a norm (i.e. the time at which it

comes into force) and the internal time of a norm

(i.e. the time stipulated for its satisfaction, its

deadline). Here, a formula of the form:

NN(agent1, role1, action, time2, agent2, role2, time1)

denotes that at time point time1 agent1 (acting as role1)

is in legal relation NN towards agent2 (acting as role2)

to perform

action by time2. For instance, consider the

case where two norms (D1 and D2) that define

conflicting obligations for Agent3 are active. The first

one is initiated at time

ET1 and it is towards retailer

Agent1 who is a regular client. It sets an obligation to

perform delivery until

IT1. The second one is towards

retailer

Agent2, it is initiated at ET2 and defines a

reparatory obligation to perform delivery until

IT2.

The relation between time points is as follows:

ET1 <

ET2 < IT2 < IT1

. There is information that can be used

to determine different conflict resolution criteria.

The strategy of temporality based on external time

may give priority to

D1 as it was initiated first. On

the other hand, temporality based on internal time

may give priority to

D2 since it has a shorter

deadline. Another alternative, using the strategy of

hierarchy is to give precedence to

D1, because Agent1,

as a regular client, takes precedence over

Agent2. Or,

we may give precedence to

D2, because it concerns a

reparatory action, if we choose to assign higher

priority to secondary norms over primary ones. It

should be clear that various combinations of these

NORMATIVE CONFLICTS - Patterns, Detection and Resolution

531

criteria may also be defined based on the agent’s

current knowledge and the assumptions it makes.

4 CONCLUSIONS AND FUTURE

WORK

In (Giannikis & Daskalopulu, 2007b) we proposed

the representation of e-contracts as default theories

that can be constructed dynamically from event

calculus representations. This technique affords us

the ability to perform temporal reasoning, defeasible

reasoning and conflict management. In this work, we

presented a set of normative conflict patterns that

may be encountered in e-contracts, and recorded

some conflicts that have not been identified yet in

other proposals. Our current work focuses on

developing a computational tool based on Reiter’s

DfL and its major variations, that supports temporal

defeasible reasoning as well as conflict detection and

resolution as presented in this paper and in

(Giannikis & Daskalopulu,2006).

REFERENCES

Abrahams, Alan S., & Bacon, Jean M. 2002 (May). The

Life and Times of Identified, Situated, and Conflicting

Norms. Pages 3–20 of: DEON'02: Deontic Logic in

Computer Science, 6th International Workshop on

Deontic Logic in Computer Science.

Antoniou, Grigoris. 1999. A tutorial on default logics.

ACM Computer Surveys, 31(4), 337–359.

Brewka, Gerhard. 1994. Reasoning about priorities in

default logic. Pages 940–945 of: AAAI'94:

Proceedings of the twelfth national conference on

Artificial intelligence (vol. 2). Menlo Park, CA, USA:

American Association for Artificial Intelligence.

Daskalopulu, Aspassia. 2000. Modeling Legal Contracts

as Processes. Pages 1074–1079 of: Legal Information

Systems Applications, 11th International Workshop on

Database and Expert Systems Applications

(DEXA'00). IEEE Computer Society.

Giannikis, Georgios K., & Daskalopulu, Aspassia. 2006.

Defeasible Reasoning with e-Contracts. Pages 690–

694 of: IEEE/WIC/ACM International Conference on

Intelligent Agent Technology (IAT 2006). Hong Kong,

China: IEEE Computer Society.

Giannikis, Georgios K., & Daskalopulu, Aspassia. 2007a

(February). Normative Conflict Management with

Default Logic. Tech. rept. 5154/1, A006.3ΓΙΑ.

Department of Computer and Communications

Engineering, University of Thessaly.

Giannikis, Georgios K., & Daskalopulu, Aspassia. 2007b.

The Representation of e-Contracts as Default

Theories. Pages 963–973 of: Okuno, H.G., & Ali, M.

(eds), Proceedings of 19th International Conference

on Industrial, Engineering and Other Applications of

Applied Intelligent Systems (IEA/AIE 2007). LNAI

4570. Kyoto, Japan: Springer-Verlag Berlin

Heidelberg.

Herrestad, H., & Krogh, C. 1995. Deontic Logic

Relativised to Bearers and Counterparties. Pages 453

– 522 of: Anniversary Anthology in Computers and

Law, Ed. J. Bing. and O. Torrund.

Horty, John F. 1994. Moral Dilemmas and Nonmonotonic

Logic. Journal of Philosophical Logic, 23(1), 35–65.

Jones, Andrew J.I., & Sergot, Marek J. 1996. A formal

characterisation of institutionalised power. Journal of

the IGPL, 4(3), 427–443.

Lee, Ronald M. 1988. Bureaucracies as Deontic Systems.

ACM Trans. Inf. Syst., 6(2), 87–108.

Lupu, Emil, & Sloman, Morris. 1999. Conflicts in Policy-

Based Distributed Systems Management. IEEE Trans.

Software Eng., 25(6), 852–869.

Makinson, David. 1986. On the formal representation of

rights relations. Journal of Philosophical Logic, 15(4),

403–425.

Moffett, Jonathan D., & Sloman, Morris S. 1993. Policy

Conflict Analysis in Distributed System Management.

Journal of Organizational Computing.

Prakken, Henry, & Sergot, Marek J. 1996. Contrary-to-

Duty Obligations. Studia Logica, 57(1), 91–115.

Reiter, Raymond. 1980. A Logic for Default Reasoning.

Artif. Intell., 13(1-2), 81–132.

Royakkers, Lamber M. M. 1998. Extending Deontic Logic

for the Formalisation of Legal Rules. Kluwer

Academic Publishers.

WEBIST 2009 - 5th International Conference on Web Information Systems and Technologies

532