Conflict Resolution in Overlapping Information Fields
for Context-based Activity Design
Zuraini Zainol and Keiichi Nakata
Informatics Research Centre, University of Reading, RG6 6UD, Reading, U.K.
Keywords: Norm Analysis, Information Field, Conflict Resolution Strategies.
Abstract: Norms are a set of rules that govern the behaviour of human agent, and how human agent behaves in
response to the given certain conditions. This paper investigates the overlapping of information fields (set of
shared norms) in the Context State Transition Model, and how these overlapping fields may affect the
choices and actions of human agent. This paper also includes discussion on the implementation of new
conflict resolution strategies based on the situation specification. The reasoning about conflicting norms in
multiple information fields is discussed in detail.
1 INTRODUCTION
Information field is a set of shared social norms that
governs the behaviour of a group member in an
organised fashion (Stamper et al., 2004). For
instance, a human agent may belong to different
social groups and each has its own shared set of
norms referred to as fields of norms. The
overlapping of these fields of norms may introduce
conflicts. So, in order to resolve these conflicts, a
strategy is required to decide which norm should be
applied. The aim of this paper is to resolve the
conflicting norms, which may affect the transition of
activity states in the Context State Transition Model
(CSTM). The reminder of this paper presents a brief
overview of the Context-based Activity Design
(CoBAD), followed by the representation of norms,
then conflict resolution strategies, the analysis of
conflicting norms in multiple information fields,
discussion and conclusion.
2 CONTEXT-BASED ACTIVITY
DESIGN (CoBAD)
Context Ontology Model (COM) is one of the
important elements in this study that provides a
well-structured scheme for semantic representation
of context identifiers, which enables context
reasoning. The COM consists of three top-level
entities (e.g., Extrinsic Context, Intrinsic Context,
and Interface Context), which corresponds to an
activity system. The Extrinsic Context refers to its
surrounding environment, the Intrinsic Context
describes the attributes of human agent, and the
Interface Context refers to its activities involving the
interactions with its environment. This captures an
activity system as an interaction (Interface Context)
between the agent (Intrinsic Context) and its
surroundings (Extrinsic Context) that potentially
results in changes to all three contexts that can be
applied into common CoBAD (Zainol and Nakata,
2010). The CoBAD represents the use of context and
its dynamic changes in the interactive systems. Two
inference mechanisms have been introduced in this
study: (i) Activity Reasoning (AR) rule specifies the
activity reasoning of human agent; (ii) State
Transition (ST) rule specifies the possible activity
states that human agent can perform, and thus will
affect the choices of the next activity state.
Typically, both of them are defined as a set of a
condition-action rule. The AR rule may be described
as having the following general form:
[ExtrinsicContext ∧ IntrinsicContext] →
[newExtrinsicContext ∧ newInterfaceContext ∧
newIntrinsicContext]
when newInterfaceContext ≠ 0;
The left hand side (LHS) of the AR rule refers to
the situational conditions, while the action on the
right hand side (RHS) of the AR rule consists of a
new context of any category when new interface
190
Zainol Z. and Nakata K..
Conflict Resolution in Overlapping Information Fields for Context-based Activity Design.
DOI: 10.5220/0004124901900195
In Proceedings of the International Conference on Knowledge Engineering and Ontology Development (KEOD-2012), pages 190-195
ISBN: 978-989-8565-30-3
Copyright
c
2012 SCITEPRESS (Science and Technology Publications, Lda.)
context is not empty. A special type of AR rule is ST
that can be expressed as follows:
[ExtrinsicContext ∧ InterfaceContext ∧
IntrinsicContext]→ [newInterfaceContext]
The LHS of the ST rule refers to the current activity
state (Interface Context) of a human agent in a
specific situation. The RHS of the ST rule represents
the updates to the current activity state, which refers
to next activity state (new Interface Context). If the
LHS of a rule become true then the action
corresponding to the specific rule is triggered. This
in turn results in the changing of activity states. In
order to design the transitions of activity states, we
apply the method of state space representation. We
proposed the Context State Transition Model
(CSTM) that consists of a set of activity states and
ST rules. However, this model lacks the ability to
capture and represent the meta-level contextual
specification. Therefore, a semiotics theory, such as
norms and information field (IF) theory is
incorporated in the CSTM, which represent the
aspect of default and dynamic norms that governs
the behaviour of a human agent in a specific
situation. Such set of norms should be modular and
should be dynamically added and removed with a
possibility to specify preferences between the
conflicting norms.
3 TAXONOMY OF NORMS
Norm is a field of force that makes inhabitants of a
community to behave or think in a particular way
(Stamper et al., 2000). Humans are seen as agents
and their actions are influenced by the forces that are
present in information fields (IFs), and these forces
originate from the norms that are shared within the
community (Gazendam, 2004). In this study, the
concept of norms and the IFs paradigm are adopted
to represent the aspect of norms, which govern the
behaviour of a human agent in the CSTM. There are
four types of norms: perceptual, evaluative,
cognitive and behavioural (Stamper et al., 2000).
Based on this, we further define them as follows: (i)
perceptual norms are concerned with how human
agent acts in accordance with his/her perception
based on facts. They can be represented as
condition
; (ii) cognitive norms represent the
aspect of human agent’s belief about actions. They
can be represented as activities
; (iii)
behavioural norms determine how a human agent
should behave and define what a human agent is
expected to do under a given situation. These norms
are represented by ST rules:
∧
→
, where
both
and
are activity states and
is a
condition; and finally, (iv)evaluative norms are
used to represent the aspect of choices or
preferences of a human agent to choose his/her next
action based on the available context information.
They can be represented by a set of activities or
conditions
, …,
,
where
is the most preferred activity than
,…,
, or
, … ,
,
where
is the most preferred condition than
,…,
. Based on our observation, the
intersections of two or more IFs in the CSTM may
introduce conflicts among norms. In other words, an
agent can be affected by more than one information
fields (IFs) at one time. One possible way to resolve
a conflict is to set precedence of norms in the shared
norms. By setting the precedence of norms in the
IFs, a human agent is expected to be able to comply
with an appropriate norm based on the situation.
4 CONFLICT RESOLUTION
STRATEGIES
In order to resolve a conflict, a strategy is required to
select a rule from the conflict set for firing. The most
popular strategies used in many of the existing
production system (PS) are random, recency,
specificity, and refractoriness. However, none of
them support the preference setting of norms. To
overcome this problem, we proposed new strategies
and divided them into three main situations:
Situation 1: when the information field 1(IF 1)
is more dominant than the IF 2
or vice versa then apply dominant IF.
The strategies are: (i) DominantIF: choose the
rule from dominant IF; (ii) Dominant-
PreferredOutcome: choose the rule that result
in preferred outcome specified by evaluative
norms in dominant IF; (iii) Dominant-
PreferredCondition: choose the rule that
contains preferred condition specified by
evaluative norms in dominant IF; (iv)
DominantRuleCondition: choose the rule that
contains condition in dominant IF.
Situation 2: If no dominant IF specified
then apply any IF. These
strategies are listed as follows: (i)
PreferredOutcome: choose the rule that result
in preferred outcome in any IF; (ii)
PreferredCondition: choose the rule that result
in preferred condition in any IF.
ConflictResolutioninOverlappingInformationFieldsforContext-basedActivityDesign
191
Situation 3: If the strategies specified in both
situation 1 and 2 failed then apply the standard
strategies, such as random, recency,
specificity, and refractoriness.
5 THE ANALYSIS OF
CONFLICTING NORMS
In order to analyse the reusability of CSTM into
different set of norms, we present three simple
diagrams. Firstly, we analyse the base information
field (IF) (see figure 1), followed by the analysis of
the overlapping a non-dominant IF (figure 2), and a
dominant IF (figure 3) onto the base IF. To perform
reasoning, we applied a production system (PS) and
set the ordering of strategies as follows:
DominantIF, DominantPreferredOutcome, Domi-
nantPreferredCondition, DominantRuleCondition,
PreferredOutcome, PreferredCondition, Random,
Recency, Specificity, and Refractoriness. The
sequences of rule firing and actions are presented in
Table 1, 3 and 5.
5.1 The Base IF
Figure 1 shows a diagram that represents the default
activity state transitions for base IF. It consists of a
set of activity states
,…,
,
conditions
,…,
, and state transition (ST)
rules
,…,
that specify the set of possible
activity states. Depending on which activity state the
human agent is in, different ST rules are available
for a human agent to be triggered. We assume
activities are mutually exclusive. i.e., a human agent
cannot be engaged in more than one activity at any
time. Table 1 illustrates an example of PS model
solution for base IF. It summarises the sequences of
rule firings and actions in the example, and the
stages of working memory (WM) in the execution
along with the directed graph of the state space
(refer figure 1). The first column shows the cycle of
PS. The second column describes the information
content in the WM, which can be further extended
into two subsections: the first sub column contains
current context state (facts) and the second sub
column contains the derived state as the result of
executing the ST rule in the preceding cycle.
Next, the third column refers to the potential
rules that can be fired in the conflict set, while the
fourth column describes possible strategies that
would be employed to fire the chosen rule, however,
if no rules are applicable then stop. Finally, the rule
that has been fired is presented in the last column,
while the action (facts) in the ST rule is then added
into the WM to be reused in the next cycle.
Table 1: Trace of a production system for base IF
.
Cycle Working Memory Conflict
set
Conflict
Resolution
Rule
fired
IF
0
derived facts
0 C
2
,
C
5
A
1
(initial state) ST
2
NULL ST
2
1 C
2
,
C
5
A
2
ST
5
NULL ST
5
2 C
2
,
C
5
A
5
NULL NULL HALT
Figure 1: An example of CSTM with the rules of state transition for base information field.
State Transition Rules
1
:
1
∧
1
→
3
2
:
1
∧
2
→
2
3
:
3
→
7
4
:
2
∧
3
→
4
5
:
2
→
5
6
:
2
∧
4
→
6
7
:
7
∧
5
→
8
8
:
4
∧
6
→
8
9
:
7
∧
8
→
9
10
:
5
→
9
11
:
6
∧
9
→
3
12
:
8
→
9
13
:
5
∧
7
→
4
KEOD2012-InternationalConferenceonKnowledgeEngineeringandOntologyDevelopment
192
Assume that when we start the rules and facts are
loaded into production rules and WM, respectively
(refer table 1). Given that
represents the starting
activity state. Based on the known facts in the WM,
we apply a forward chaining, which is reasoning
from facts to the conclusions resulting from those
facts. The inference begins from the top of the rule,
,
and goes on downward until the first true
condition is found. In the first iteration (cycle 0), the
recognise-act cycle (RAC) matches the current state
(conditions):
,
and
in WM against the ST
rules. At this stage, the rule,
matches with the
facts (conditions) in the WM. Therefore, the rule,
is fired and its action,
, is asserted in the WM.
This in turn affects the transition from activity state,
to
. The second iteration (cycle 1) uses this
information and the updated facts:
,
and
would match with the rule,
. The rule,
is then
fired and the activity state
is then added in the
WM; indicating a transition from activity state,
to
. Finally, in cycle 2, the execution halts as no
more rules to fire.
5.2 The Overlaying of Non-dominant
IF to Base IF
Next, we analyse the overlapping of a non-dominant
information field, e.g., information field X
to
base information field
. When these two IFs
overlap with each other, the
will brings different
set of norms (e.g., facts, rules, preferences) into the
existing set of norms of
, as presented in table 2.
As a result, these set of norms will then introduce
conflicts to some extent.
Table 2: A set of norms added from the
.
Types of norms
Norms added from the
Behavioural norms -
Evaluative norms
Cognitive norms
Perceptual norms
,
Table 3 illustrates an example of PS model
solution for activity state transition based on both
and
. In this table, another sub column is
added into WM’s column, which contains temporary
context information introduced by another IF.
Again, given that
represents the starting activity
state, the inference begins from the top of the rule,
,
and goes on downward until the first true
condition is found. In the first iteration (cycle 0), the
RAC matches the current state (conditions):
,
,
,
and
in the WM against the ST rules.
At this stage, only two rules:
and
matched
with the facts (conditions) in the WM. At this stage,
only two rules:
and
matched with the facts
(conditions) in the WM. Hence, the conflict set
consists of the following information:
〈
,
,
〉
,
〈
,
,
〉
. In order to resolve the
conflict, a strategy PreferredCondition is applied to
select
for firing because of the following
explanation:
is
more preferred than
,
, where
is a new fact that
has been
introduced by the
. Therefore,
is
fired and its action,
, is asserted in the WM. This
in turn affects the transition from activity state,
to
.
Figure 2: The overlaying of
to
in the CSTM.
The second iteration (cycle 1) uses this
information and the updated facts:
,
,
,
and
would match with
. The rule,
is then fired
and the activity state
is then added in the WM;
indicating a transition from activity state,
to
.
In the third iteration (cycle 2), the RAC again
matches the updated facts:
,
,
,
and
in
the WM against the rules in the production rule. At
this stage, two rules,
and
are enabled for
firing, and as for now, the conflict set consists of the
following information:
〈
,
,
〉
,
〈
,
,
〉
.
In order to resolve the conflict, a strategy
PreferredOutcome is applied to select
for firing
because its conditions matches with the preferred
outcome. As a result, the rule
is fired and its
action
is asserted into WM, indicating a transition
from activity state,
to
. Finally, in cycle 3, the
execution halts as there are no more rules to fire.
Information
Field X
ConflictResolutioninOverlappingInformationFieldsforContext-basedActivityDesign
193
Table 3: Trace of a production system based on the overlapping
and
, where
.
Cycle
Working Memory
Conflict set Conflict Resolution Rule fired
IF
0
IF
x
derived facts
0 C
2,
C
5
C
1,
C
8
A
1
(initial state) ST
1
ST
2
PreferredCondition ST
1
1 C
2,
C
5
C
1,
C
8
A
3
ST
3
NULL ST
3
2 C
2,
C
5
C
1,
C
8
A
7
ST
7
ST
9
PreferredOutcome ST
9
3 C
2,
C
5
C
1,
C
8
A
9
NULL NULL HALT
5.3 The Overlaying of a Dominant IF
to Base IF
In the next example (see figure 3), we introduce the
information field Y
, which is more dominant
than base information field
.Note that, the
replacement of
, will, thus, bring another set of
norms (e.g., facts, rules, preferences) into existing
set of norms (see table 4). The overlapping of these
IFs may introduce conflicts, which can be further
summarised in table 5.
Table 4: Set of norms added from the
.
Types of norms
Norms added from the
Behavioural norms
:
∧
→
:
→
:
∧
→
Evaluative norms
Cognitive norms
Perceptual norms
,
Again, given that
represents the starting
activity state, the inference begins with rule,
,
and goes on downward until a rule that fires is
found. In the first iteration (cycle 0), the RAC
matches the current state (conditions):
,
,
,
and
in the WM against the ST
rules in the production rule. At this stage, a conflict
occurs when two rules,
and
are matched
with the current conditions in the WM. Hence, the
conflict set consists of the following information:
〈
,
,
〉
,
〈
,
,
〉
. To resolve a conflict,
a strategy DominantRuleIF is applied to select a rule
from the dominant IF. Therefore,
is fired and
its action,
is placed in the WM, which in turn
moves the activity state from
to
. In the
second iteration (cycle 1), the RAC again matches
the updated facts:
,
,
,
and
in the
WM with the production rule. At this stage, only
matches with the current facts. Hence,
is
fired and its action,
is then added into WM,
indicating a move from an activity state,
to
.
In the third iteration (cycle 2),
and
are
matched with the updated facts:
,
,
,
and
.
The conflict set contains the following
information:
〈
,
,
,
〉
,
〈
,
,
〉
. To
resolve the conflict, a strategy,
DominantPreferredCondition is employed to select
and fire the
based on the following preferences:
is more preferred than
,
, which then placed
in the WM. The
transition of activity state is now changed from the
activity state,
to
.
Figure 3: The overlaying of
to
in the CSTM.
In the next iteration (cycle 3), the RAC again
matches the updated facts:
,
,
,
and
in
the WM against the rules in the production rule.
Two rules:
and
are matched with the
updated facts. Hence, the conflict set contains
〈
,
〉
,
〈
,
,
〉
. To resolve the conflict,
a strategy DominantPreferredOutcome is
implemented to choose
as
is more preferred
outcome than
,
. When
the
rule is fired, its action,
is added into
WM, and the transition of activity state is now
changed from the activity state,
to
. Finally in
cycle 4, the execution halts as there are no more
rules to fire.
Information
Field Y
KEOD2012-InternationalConferenceonKnowledgeEngineeringandOntologyDevelopment
194
Table 5: Trace of a production system for CSTM based on the overlapping
and
, where
.
Cycle
Working Memory
Conflict set Conflict Resolution Rule fired
IF
0
IF
y
derived facts
0 C
2
,
C
5
C
y1
,
C
y2
A
1
(initial state) ST
2
ST
y1
DominantRuleIF ST
y1
1 C
2
,
C
5
C
y
1
,
C
y
2
A
y
1
ST
y
2
NULL ST
y
2
2 C
2
,
C
5
C
y
1
,
C
y
2
A
6
ST
11
ST
y
3
DominantPreferredCondition ST
y
3
3 C
2
,
C
5
C
y1
,
C
y2
A
5
ST
10
ST
13
DominantPreferredOutcome ST
10
4 C
2
,
C
5
C
y1
,
C
y2
A
9
NULL NULL HALT
6 DISCUSSION AND
CONCLUSIONS
We have incorporated norm specifications into
production system (PS) model that uses the forward
chaining over production rules as a way of
reasoning. The concept of information field (IF)
provides default context identifier values, and is also
capable of systematically capturing different set of
norms based on context depending on what type of
IF has been added into the Context State Transition
Model (CSTM). However, a conflict may occur in a
set of shared norms and a strategy is required. In this
paper, we presented new strategies, which are
capable of handling the preference setting of norms.
The selection of strategies is dependent on the
situation specification defined in this study. For
instance, the implementation of DominantIF’s
strategy is to choose a rule from the dominant IF for
firing. However, if the first strategy failed, then
another strategy, e.g., DominantPreferredOutcome
will be selected next. The process of selecting a
strategy will continue until one of them matches
with the preference setting. The standard strategy
will, then, be implemented by the production system
(PS) when new strategies fail to resolve a conflict.
Besides, these new strategies offer a semantic
strategy, which enable the PS to resolve the
conflicting norms based on their meaning. While the
traditional strategies are only syntactic which
resolve the conflicts in accordance to the form and
occurrence of rules and conditions. However, the
application of strategies in the PS model is limited in
terms of user’s choices. As such, a user should be
able to select their own strategies in resolving the
conflicts. Furthermore, in order to make the
inference more effective, the random strategy should
be considered first, as this strategy is the most
selected by other PS. Future work will be focused on
the development of multiple scenarios of real-world
problems, and then followed by the implementation
based on a rule-based expert system, such as the
Jess.
REFERENCES
Gazendam (2004). Organizational Semiotics: A State of
the Art Report. Vol. 1(1). Retrieved from http://www.
semioticon.com/semiotix/semiotix1/sem-1-05.html
Stamper, R., Liu, K., Hafkamp, M., & Ades, Y. (2000).
Understanding the Roles of Signs and Norms in
Organization - A Semiotic Approach to Information
System Design. Behaviour and Information
Technology, 19, 15-27.
Stamper, R. K., Liu, K., Sun, L., Tan, S., Shah, H., Sharp,
B., et al. (2004). Semiotic Methods for Enterprise
Design and IT Applications. Proceedings of the 7th
International Workshop on Organisational Semiotics
2004, pp. 190-213.
Zainol, Z., & Nakata, K. (2010). Generic Context
Ontology Modelling: A Review and Framework. Paper
presented at the 2nd International Conference on
Computer Technology and Development, Cairo,
Egypt, pp. 126-130.
ConflictResolutioninOverlappingInformationFieldsforContext-basedActivityDesign
195
