Processes and workflows in an organization involve tasks that require cooperation
and collaboration. There are tasks in a stage of the workflow that are dependent
on the previous stage. Failure to complete the activities at a certain stage
can cause delays and may lead to failure in meeting critical deadline. To ensure
that the workflow runs smoothly, organizations impose specifics rules and guidelines
to govern the processes (Skarbek, 2012; Ahmad,
2004). In addition to these rules and guidelines, some organizations also
rely on norms in coordinating the tasks (Castelfranchi et
al., 2000). Norms represent the soft constraints or unwritten rules
that regulate the behaviors of role actors in an organization (Barka
and Sandhu, 2000).
In many organizations, various plans are executed to avoid delays in task completion.
One of the strategies is to decompose and delegate the tasks (Barka
and Sandhu, 2000; Ahmad et al., 2009). However,
in an organization where an individual may be responsible for a few roles and
has to complete many tasks, the process of delegating tasks may not be trivial.
In many organizations, individuals are organized and grouped based on roles.
Based on the assigned role, their tasks and responsibilities vary. As the organization
grows, the assigned roles might correspondingly grow and become more complex.
In line with this concept, role and task delegation are useful in ensuring that
the organizational goal is achieved. In large organizations, the delegation
of roles and tasks are complex especially when multiple roles are involved.
This creates problems of coordination between different tasks and allocation
of resources, which are examples of challenges in this area. Among others, solutions
from the literature use norms to address the coordination problem of delegating
resources and tasks allocation in organizations (Ahmad et
al., 2009; Conte et al., 1999).
Consequently, this paper discusses how norms can be exploited to resolve such
coordination problem. We select the domain of Final Examination Paper Preparation
(FEPP) for the Diploma Programs at the Faculty of Computer and Mathematical
Sciences of Universiti Teknologi Mara (UiTM) for this research project. This
involves the discussion on the roles of norms as soft constraint or unwritten
rules and guidelines in regulating individuals behavior to complete their
tasks in a selected domain. We then model task delegation process in a normative
multi-agent environment and simulate the dynamism of task delegation to achieve
some organizational goal.
The objectives of this work are (1) To analyse and develop existing multi-agent
systems and the delegation of roles in FEPP systems, (2) To develop and implement
a framework for the delegation of roles in cooperative agents for a multi-agent
system based on the concept of normative governance and (3) To assess the performance
of the cooperative agents in multi-agent systems.
The achievement of these objectives would provide the answers to the following
||Is the inclusion of the principles of normative governance
of individual and society clusters in the delegation of roles in multi-agent
systems a valid proposition?
||How one does developed a framework for the delegation of roles based on
the above principles?
||Does the inclusion of such principles in multi-agent systems improves
This research contributes to the multi-agent workflow process by improving
the availability and quality of task performances. The improvement is realized
through the use of agents in identifying task overload situation and attempt
to reduce the problem through task delegation.
Our contribution in this paper is two-fold. Firstly, we propose a framework
for task delegation using software agents. Secondly, we incorporate a normative
framework that provides clear indication to humans on the urgency of their scheduled
In multi-agent systems environment, norms are deployed to facilitate agents
actions which subsequently improve their coordination and cooperation. Norms
have been said to increase the predictability of multi-agents behavior
via the compliance of agent actions to the norms. Many frameworks have been
developed that show how norms facilitate and regulate agents rational
behavior in multi-agent systems (Ahmad et al., 2009;
Conte et al., 1999; Rao and
George, 1995; Lopez and Marquez, 2004). They are
modeled according to the BDI (Belief-Desire-Intention) architecture. Such agents
have the advantage of making decisions similar to humans decision-making
(Thangarajah et al., 2008). For example, if an
agent needs to choose a task to prioritize, its rational behavior should enable
it to choose between a normative goal and its individual goals which are based
on its beliefs of the environment.
An autonomous agent with norms is called a normative autonomous agent. Such
agent is able to adopt some norms (i.e., norm instances) and select which norms
to comply with (i.e., intended norms) and which norms to reject (i.e., rejected
norms) (Savarimuthu et al., 2005). Conte
et al. (1999) propose that objects that enter the agents mental
process and interact with its beliefs, aims and plans represent norms. Consequently,
the agent is able to decide either to follow or violate a norm and react to
norms violation. Normative agents are aware of the enacted norms
so much so that they are able to either obey or violate the norms in specific
situations (Conte et al., 1999), thus enabling
them to complete a task within the given duration.
Ahmad et al. (2009) developed a normative multi-agent
framework called the Obligation-Prohibition-Recommended-Neutrality-Disliked
(OP-RND) framework to regulate rules and norms effectively. Their agents perform
tasks from a set of pre-compiled tasks based on their beliefs of the reward
and penalty associated with the selected tasks. In their work, Obligation, O,
is a command imposed by some authority agent. In such environment, an agent
is obligated to perform an action, gets rewarded for doing it or penalized for
leaving it. They defined Prohibition, P, as a command, which an agent has to
avoid an action, hence gets rewarded for leaving it or penalized for doing it.
They consider Obligation and Prohibition (OP) as rules imposed by the authority
in a normative environment due to absolute consequences (reward or penalty)
upon conformation or violation of some action (Barka and
Sandhu, 2000). In our work, we exploit the OP-RND normative framework to
incorporate it with the workflow and to eventually enhance the system performance.
This study is based on the selected domain of Final Examination Paper Preparation
(FEPP) process for the Diploma Programs at the Faculty of Computer and Mathematical
Sciences of Universiti Teknologi Mara (UiTM). Since the Diploma of Computer
Science program is offered in several branch campuses, the common process of
FEPP is implemented throughout the faculty. Coordinators for the courses are
selected by the Dean among the UiTMs branch campuses that offer the facultys
courses. These coordinators coordinate and prepare the examination questions
of assigned courses in camera-ready form before sending them to
the examination office at the facultys main branch. The selection of a
branch campus for a particular courses examination paper preparation is
done based on the number of experienced lecturers teaching the course. The appointment
for each branch campus is valid for two years. The hierarchical relations for
the FEPP process are shown in Fig. 1.
The FEPP process starts at the beginning of a new semester. The Dean via the
Deputy Dean (Academic) instructs all lecturers to prepare sets of final examination
papers. At each of the branch campus, the Program Coordinator (PC) receives
the preparation instruction for all the courses under his/her responsibility.
He/She extends this instruction to the Course Coordinators (CC). Since each
course is offered in more than one branch campuses and to ensure the quality
of the questions prepared, the CC informs all Branch Course Coordinators (BCC)
at other branch campuses to prepare a set of final examination paper (for a
course) as a contribution set. This set consists only of raw questions and solutions
and is not formatted as required of the final examination paper. From the contributions
sets received from all the branch campuses, the CC and the Moderation Committee
finalize three sets. These three sets are then sent to the facultys main
branch before they proceed to the next step at the examination department.
||Hierarchical relations for FEPP process
The preparation for other courses examination papers follows the same
instructional and assignment structure.
UiTMs unique FEPP entails several processes, which are considered as
norms. These norms form the basic foundation for the FEPP framework. At the
beginning of each semester, every lecturer is aware that they have to prepare
questions for the final examination:
||The duration for preparing examination questions depends on
instruction from higher management, which might be issued on the first day
of the semester or later
||The CC must also get the examination sets from other branches through
the BCC of other branches (society clusters norm)
||The BCC/CC divides the task of preparing the questions based on the syllabus
and format among other lecturers (L) who teach the same subject
||The BCC/CC gets the examination questions from L and compiles them
||The BCC may submit the examination sets directly to the CC (without sending
through the PC)
||The BCC may submit the examination set through the PC, who submits to
||If only one lecturer, who is also a BCC/CC, teaches the course, he/she
prepares it alone
||If more than one lecturer teaches one course, the BCC/CC delegates the
task among the other lecturers
While some of these norms are entrenched within the processes of the FEPP framework,
some other norms could be improved, especially in the process of task assignment,
delegation and examination paper submission.
FINAL EXAMINATION PAPER PREPARATION (FEPP) FRAMEWORK
Due to the many number of preparation tasks (i.e., two or three courses) that
each lecturer receives from his/her Course Coordinators, they may not be able
to avoid delays in completing their tasks. However, since they are required
to submit their papers before the deadline, they are obliged to collaborate
among themselves. To model such collaboration, our framework considers the relations
between software agents and their human counterparts with other humans and their
agents. An overview of our framework is shown in Fig. 2.
||Agent-based FEPP framework
||An instance of the agent-based FEPP framework
Figure 3 shows an instance of the framework for a particular
courses examination preparation. We call the group of lecturers who prepares
a particular courses examination paper as a society cluster.
The operation of the framework is based on the symbiosis between humans and
their agents. Humans mainly perform the offline task of examination paper preparation
while their agents continuously monitor the states of the environment to which
they autonomously respond. Specifically, the agents perform common mundane actions
such as alerting humans to complete the task, reminding imminent deadlines,
forwarding completed documents to the right destination and all other actions
that are necessary to achieve the FEPPs goal. With such functions given
to the agents, the humans can ignore the deadlines of documents submissions
and their destinations. Alerting service provided by the agents ensures continuous
reminder of the deadlines. The agents then submit the documents automatically
to the intended recipients when humans upload the documents to his/her folder.
All these events are recorded in an event log file as part of the agents
environment for tracking the process flow.
INCORPORATING NORMS IN DELEGATION
In the examination paper preparation process, the norm is to complete the preparation
and submit the paper before the deadline. However, task overload problems cause
norms violation when some examination papers are not submitted by the deadline.
With task delegation and norms incorporated in a multi-agent system, such problem
could be avoided. We exploit Ahmad et al. (2009)
work on the OP-RND normative framework, which suggests that a rule is a mutually
exclusive state of Obligation, O and Prohibition, P, imposed on an agent. Figure
4 shows an abstraction of the OP-RND framework from Ahmad
et al. (2009).
The domain discussed in (Ahmad et al., 2009)
is suitable for our main objective of enhancing the performance of FEPPs
workflow. Consequently, we incorporate their normative framework in our domain
with some required modifications which are appropriate to our work. While the
FEPP process entails similar conditions and parameters associated with the deadlines,
we incorporate the idea of rewards and penalties that are imposed for norm compliance
and violation. There are two reasons for deploying the OP-RND framework:
||The OP-RND framework provides a clear delineation or boundary
of normative performance of humans and agents based on the Recommended,
Neutrality and Disliked periods
||The RND periods can be specified or adjusted according to the abilities
of the performers given all the necessary resources
If α is an agent, δ is an action, Ω is a reward, Π is a
penalty, Γ is neutral (no reward and no penalty) then according to (Ahmad
et al., 2009):
||O is a state in which an agent must perform an action and
is rewarded for doing it but penalized otherwise:
||P is a state in which the agent must avoid an action and is
rewarded for leaving it but penalized otherwise:
||A norm is a mutually exclusive state of Recommended, R, Neutrality,
N and Disliked, D, where:
||R is a state in which the agent is rewarded for performing
an action but is not penalized otherwise
||N is a state in which the agent is neither rewarded nor penalized
for performing or avoiding an action
||D is a state in which the agent is rewarded for avoiding an
action but is not penalized otherwise
Normative governance in FEPP framework: Due to the task overload problem,
the FEPP workflow occasionally encounters delays. Since the lecturers need to
collaborate in preparing the contribution sets in limited time, failure to complete
the task leads to failure to meet the deadline. Hence, this overload problem
of the FEPP and personal goals (i.e., other responsibilities of the lecturers)
demand the delegation of tasks with normative governance.
The implement the normative governance, we incorporate the OP-RND in the FEPP
framework at the lecturers level. To concur with the OP-RND frameworks
terms, we called the examination paper submission as the normative goal, GN,
which needs to be achieved by the given deadline. At the same time there are
other goals, which we called the personal goals, GP, based on each
individual lecturers obligation such as performing administration tasks,
marking quiz papers, attending meetings and so on. In this case, both GN
and GP need to be arranged within two weeks based on the OP-RND framework
The OP-RND normative framework implements the normative governance for our
agent-based FEPP framework as follows:
||We propose the Recommended Period, PR, for an agent,
α, to motivate or urge a lecturer, L, to complete his/her task, T,
within the Recommended period slot (PR = S1), to get
a reward, Ω
||We propose the Neutrality Period, PN, for an agent,
α, to alert a lecturer, L, to complete his/her task, T, within the
Neutrality period slot (PN = S2), to avoid a penalty,
Π, but no reward
||We propose the Disliked Period, PD, for an agent,
α, to inhibit a lecturer from such delay in the future that effectively
subject the FEPP to high risks as this process is associated with students
examination schedule. In this period, the lecturer completes his/her task,
T, within the Disliked Period slot (PD = S3) and incur
a penalty, Π
In this work, the reward and penalty actions are given points (+1, 0, -1). Completing
the task in the following periods, incur the corresponding points:
||In the Recommended Period, P , a lecturer receives +1 point,
reward (Ω = +1)
||In the Neutrality Period, PN, a lecturer receives
0 point, Neutrality (Γ = 0)
||In the Disliked Period, PD, a lecturer receives
-1 point, Disliked (Π = -1)
Figure 5 shows the three periods, the green area represents
the recommended period, in which a lecturer receives +1 point. The blue area
represents the neutrality period, in which a lecturer receives 0 point and the
red area represents the disliked period, in which a lecturer receives -1 point.
The figure also shows the black area, the Prohibited Period (PP),
which this research attempts to resolve.
Normative process with delegation: We use the norms with delegation
in our framework as an interesting prospect to improve the clusters predictability.
Since time is very important for the lecturers, they need to help each other
by cooperating between the group members to gain time and get the rewards.
Since we have two weeks for each lecturer to submit his/her tasks, the agent
always recommend submitting in earlier time. Therefore, if the lecturer has
more than three tasks (i.e., preparing examination papers for four courses),
he/she should request the CC to delegate one of the courses:
His/her agent calculates the duration, μ, for preparation from the moment
the request is made to the CC to request for delegation or to increase his/her
efforts, F, by doing his/her best to submit in the Recommended or Neutral period:
FEPP-OP-RND FRAMEWORK MAPPING
According to the OP-RND framework, we need to set the duration of each of the
Recommended, neutrality and disliked periods for task submissions. UiTM sets
the Obligation for completing the final examination paper preparation to 15
days (Ahmad et al., 2009). We propose the first
six days (S1 = 6 days) as the Recommended period. Assume that a lecturer
gets the maximum average of the tasks (e.g., three tasks). If he/she gets more
than three, his/her agent will request for delegation. Hence, he/she can complete
each task in two days and submit all the three tasks in six days. Therefore,
he/she can submit in the Recommended period and get the reward point.
We propose the second six days (S2 = 6 days) as the Neutrality period,
considering that the lecturer has to complete some other tasks (i.e., their
personal goals) together with the FEP and he/she still has time to submit before
the Disliked period. We propose the Disliked period within the last three days
(S3 = 3 days) before the submission deadline. As for the Prohibition
period, it is set immediately after the deadline. Figure 6
shows the mapping of our FEPP Framework to the RND periods.
The norms obligation in our multi-agent system gives the expected time to complete
the task for humans and this is based on the number of tasks received by the
agent by using the formula below to calculate the number of tasks:
where, i = 1..., n, Xi is the task counter and Yi represents
the number of preparation tasks that a lecturer has been assigned to from a
For each lecturer, a counter counts the number of tasks he/she gets when a
Course Coordinator assigns the tasks to the lecturers. The agent calculates
the maximum number of days for each task based on the normative periods and
on this basis determines the number of days needed by the lecturer to complete
the task by relying on the following formula:
where, Q is the number of days for each task (average achievement) and VQ
is the value of V based on Q. where V is the expected number of days. Then:
If R is the actual value of average achievement by a lecturer, VR
is the value of V based on R, then:
||FEPP mapping with OP-RND framework
From Q and R:
where, the value of Q is less than or equal to R, the lecturer gets a reward:
otherwise the lecturer loses the reward, for example, if a lecturer has three
tasks and the reward period based on the framework is 6 days. Then:
||VQ = 6 days
||X = 3 tasks
||Q = VQ/X
||Q = 6/3 = 2
||Assume that the lecturer completes the preparation in VR
= 5 days. Then:
||VR = 5 days
||X = 3 tasks
||R = VR/X
||R = 5/3 = 1.66
Since Q>R, then the lecturer gets a reward.
The agent compares it with the UiTMs norms Obligation period by using
the Boolean operator ≥ (greater than or equal) and ≤
(less than or equal) on the above formula, Table 1 shows the
UiTMs Obligation norms.
TESTING THE NORMATIVE FRAMEWORK
We implement the normative OP-RND framework with Delegation in our system using
Win-Prolog and its extended module Chimera (Joeris, 2000).
When the FEPP starts to exchange messages between the agents and requests to
prepare the final examination paper, for each task, the agent calculates and
shows the R, N or D periods in which the submission is possible so that it helps
the human to increase his/her ability to submit at an earlier period. The message
appears to the humans as in Fig. 7.
||UiTM Obligation norms
||Norms report 1
If the lecturer has two tasks, the agent allocates each task the maximum time
to complete it, which is five days. So for two tasks, the lecturer needs 10
days to complete them and that allows him/her to complete in the Neutrality
period. Therefore, the agent informs the lecturer to increase his/her ability
to complete it in less than six days to be in the Recommended period (Fig.
8). If the lecturer has three tasks, he/she needs 15 days to submit the
FEP. Then, the agent indicates to the lecturer to submit in the Recommended
to gain reward or Neutrality period to avoid penalty as in Fig.
The lecturer submits in the Prohibited period if he/she has more than three
tasks because he/she needs more than 15 days. In this case, the agent requests
for delegation and changes the status from Prohibited period to Request for
Delegation as shown in Fig. 10.
ANALYSIS OF RESULTS
Table 2 below shows the number of preparation tasks assigned
to each lecturer, (L1-L3), before and after the delegation for seven simulation
runs, (S1-S7). The intersections of columns L1-L3 and rows S1-S7 (where L is
a lecturer and S is a simulation run) are the number of preparation tasks assigned
to the corresponding lecturers (both before and after delegation). For example,
at S6, L1 and L3 are assigned four tasks before the delegation.
Table 3 below shows the number of successful submissions
from each lecturer before and after the RND for seven simulation runs (S1-S7
from Table 2). The intersections of rows and columns are the
number of successful submitted tasks in the corresponding periods for each lecturer,
LI, L2 and L3 based on Table 2 (both before and after RND).
Each column contains seven runs, for example, L1 during the seven runs has submitted
twice in the Recommended, Neutral, Disliked periods and once in the Prohibited
||Norms report 2
||Norms report 3
||Norms report 4
||Distribution of tasks before and after delegation
||Distribution of submissions before and after RND
|Before RND (R = 23.81%, N = 28.57%, D = 33.33%, P = 14.29%),
After RND (R = 33.33%, N = 38.10%, D = 23.81%, P = 4.76%). The R percentage
(23.81%) is calculated as follows: The successful submissions for 3 lecturers
as shown in Table 3 = 2+2+1 = 5, Total submissions for
3 lecturers = 7+7+7 = 21, Percentage value of achievement = (The successful
submissions/Total submissions)*100 Percentage value of achievement = (5/21)*00
||Percentage of submissions before and after RND
||No. of submissions in each period for 3 lecturers before and
We calculate the percentage value of successful submission as follows:
As shown in Table 3, the distribution of submissions before
and after RND shows that submissions in the Recommended periods are achievable
in five tasks before RND for three lecturers, in contrast to seven submissions
achieved by the same three lecturers after RND. Within the Neutrality period,
the achievement increases from six submissions before RND to eight submissions
after RND. While in the Disliked and Prohibited period, lecturers managed to
decrease the submissions from seven to five and from three to one, respectively.
Figure 11 shows the percentage of submissions for each period
before and after RND. The percentage of submissions in the Recommended period
increases from 23.81 to 33.33%. In the Neutrality period increases from 28.57
to 38.10%, while in Disliked period it decreases from 33.33 to 23.81% and in
Prohibited period, it decreases from 14.29 to 4.76%.
Figures 12 shows the number of submissions in each period
for three lecturers before and after RND.
CONCLUSION AND FURTHER WORK
Educational institutions face numerous administrative problems in meeting academic
obligations and assigning workload to lecturers for examination paper preparation
is no exception. While other AI techniques have been applied to resolve task
scheduling problems, we have contributed here a novel technique of conflict
resolution involving the distribution of workload in examination paper preparation.
In this technique, a multi-agent system is deployed to enable intelligent software
agents to assist humans in augmenting the workflow process by performing mundane
tasks of message and document passing between agents for humans and within a
society. Clusters of such individuals and societies operate and collaborate
to determine a fair distribution of workload by delegating tasks to lightly
We incorporate a normative framework that provides clear indication to humans
on the urgency of their scheduled tasks. In this framework, humans are well-informed
about the status of their deadlines when their agents alert and remind them
of impending deadline breach based on the three normative periods allocated
for each task.
We also demonstrate the role of norms in coordinating agents behavior
in a multi-agent system. In the context of FEPP, the integration of norms with
a multi-agent system helps to improve humans collaboration and teamwork
quality in performing their tasks which is submitting completed examination
papers within the stipulated deadline. The utilization of norms and effective
delegation strategy enables the system to produce a fair distribution of workload
by delegating tasks to lightly loaded individuals. However, while agent-based
technology such as this could provide humans with the tool to improve performance
by providing information and performing the necessary actions, the ultimate
success is only achieved by the due diligence of humans in completing their
part of the work within the stipulated deadline.
In summary, results from the study provide some evidence on the significance
roles played by norms within a multi-agent syste m especially for the coordination
and regulation of agents behavior. In our future work, we shall investigate
the emergence of emotions within the normative workflow process and identify
emotive functions to further augment the performance of humans and agents.
This research project is sponsored by the Ministry of Higher Education (MOHE)
of Malaysia under the project code 600-RMI/ST/FRGS 5/3/Fst (165/2010) and led
by Universiti Teknologi Mara (UiTM) with collaboration from Universiti Tenaga