Content uploaded by Pedro Antunes
Author content
All content in this area was uploaded by Pedro Antunes on Jun 10, 2018
Content may be subject to copyright.
Supporting People-Driven, Dynamic and Geo-Located
Work Processes
Pedro Antunes
Victoria University of Wellington
P.O. Box 600, Wellington 6140
New Zealand
pedro.antunes@vuw.ac.nz
Nelson Baloian
Computer Science Department (DCC)
Universidad de Chile
Chile
nbaloian@dcc.uchile.cl
Gustavo Zurita
Dept. of Management Control and
Information System. Economics and
Business Faculty. Universidad de Chile
gzurita@fen.uchile.cl
José A. Pino
Computer Science Department (DCC)
Universidad de Chile
Chile
jpino@dcc.uchile.cl
ABSTRACT
Some work scenarios foster the adoption of people-driven,
dynamic and geo-located processes. To support such
scenarios, we suggest two fundamental changes in process
structure and control. Regarding structure, we move away
from traditional process models towards process contexts,
which can be organized around geographical locations.
Regarding control, we move away from model-based
control-flow towards dynamic activities defined by the
participants as processes unfold. This research makes the
following unique contributions: 1) It provides the first
implementation of people-driven dynamic processes; 2) It
provides the first implementation combining people-
driven dynamic processes and geographical context; 3)
Finally, it provides a unique approach to build process
context, which leverages the possibilities brought by
microblogging platforms in exchanging semi-structured
and unstructured messages.
CCS CONCEPTS
• Applied Computing → Enterprise computing;
Business process management
KEYWORDS
Dynamic definition of activities; Process context;
Contextually-supported human control; Twitter use.
ACM Reference format:
P. Antunes, N. Baloian, G. Zurita, J. A. Pino., 2018. Supporting
People-Driven, Dynamic and Geo-Located Work Processes. In
Proceedings of S-BPM ONE 2018, Linz Austria, April 2018 (S-BPM
ONE 2018), 10 pages.
1 INTRODUCTION
People-driven dynamic processes are those in which the
activities and their execution order are determined as the
process unfolds, depending on the interactions and
decisions made by human workers [1]. This type of
process is particularly adequate to scenarios where some
parts of the work either cannot be foreseen or are left
open for people to decide. Many of these scenarios also
involve the geographical context, as illustrated below.
Consider for instance, occasional rubbish collection.
Since orders from clients are defined on a daily basis, the
collection process cannot be completely planned.
However, it also cannot be completely geo-referenced or
simply ad hoc. Some degree of control-flow is necessary to
ensure quality of service and adequate performance.
Maintenance work done by electricity distribution
companies follows the same pattern. Even though the
maintenance activities can be carefully planned and
optimized, many events may occur during the day that
prevent a process to evolve as expected. For instance,
clients may not be on site to give access to contractors. A
request to cut distribution may be suddenly cancelled
because the client found a payment receipt. And of course,
urgent repairs may overtake other planned activities.
Another scenario considers firefighting. To avoid
uncertainty, firefighting is based on extensive training,
experienced professionals, and multiple contingency
plans. However, each fire is a unique case. Depending on
how it evolves, firefighters may be forced to dynamically
change the command structure, responsibilities, goals,
coordination protocols, etc. [39].
All above scenarios involve humans in making
dynamic changes in process execution, as the work
context changes over time. Furthermore, geographical
context also plays an important role in determining the
process execution, since activities are location-dependent
and the participants are constantly on the move to
accomplish their work.
The adequate support to these scenarios has generated
considerable interest in the BPM (Business Process
Management) community regarding their implications to
process modelling, enactment, management, and
S-BPM ONE 2018, April 2018, Linz, Austria
P. Antunes et al.
2
execution. Various alternatives have been studied,
including how to increase the flexibility of model-based
execution [2], how to handle unique cases [3], how to
model people-driven activities [4], and also how to
integrate geographical context in process models [5].
However, a more recent concern is how to democratize
BPM [6] by empowering the process participants to
determine the process evolution at their own discretion,
while at the same time preserving some important
characteristics of the BPM approach such as control,
visibility and traceability.
So far, such democratization of BPM has been mainly
addressed in a conceptual way [4, 6]. In this paper, we
discuss the implementation of a BPM system that brings
together the following features:
• Support people-centered process enactment and
execution, where processes can be enacted and
activities can be defined by the process
participants without any model-based constraints;
• Support dynamic process execution, where
activities can be defined and changed as the
process unfolds;
• Support geo-location as a complementary way to
structure process execution.
The system has been developed using the example
scenarios already discussed. In all these scenarios, events,
locations, emergent needs, and other contextual factors
may lead human workers to override model-based
control-flow (e.g. B can only be executed after A finishes).
Even though process execution may be foreseen, it cannot
be statically defined: it must be dynamic, people-driven
and context aware.
The research reported in [7] discussed a set of
requirements for integrating geo-located activities with
BPM, highlighting in particular the conflict between
spatial and task dependencies in control-flow. The authors
suggested that the conflict could be solved by giving
predominance to spatial dependencies. This paper
presents a concrete implementation of that proposition
plus some new ideas. The new ideas emphasize the
people-driven approach, which centers process enactment,
management and execution on the process participants.
Our implementation provides a graphical, easy to use
interface, which combines geographical and process
visualization. It also implements a people-driven solution
for efficient assignment of activities in which workers
self-manage the process execution.
The next section discusses previous work that has been
done in this area. Section 3 discusses in more detail the
system features. Section 4 describes the implemented
system, and section 5 presents results from a preliminary,
formative evaluation. Section 6 concludes the paper.
2 PREVIOUS WORK
The BPM approach is very attractive for structuring most
types of work in organizations. The work activities are
usually presented in a descriptive model that is typically
simple, understandable and elegant. Furthermore, these
descriptive models can be used to control the process
execution (what we designate by control-flow), usually by
process aware information systems (PAIS) [8]. The
coordination of activities is normally predetermined by
analysts at design time and controlled by PAIS at runtime.
Process participants then interact with worklist handlers
to execute the activities assigned to them.
However, for long it has been recognized that model-
based execution faces many challenges, especially in
organizations needing some degree of flexibility [2]. A
large body of research has been devoted to this problem,
which because of its complexity cannot be detailed in this
paper (we recommend [9] and [10] for a more
comprehensive overview). Nevertheless, we can
summarize the existing viewpoints and approaches to
address the problem.
Well-structured processes with ad hoc activities.
This approach regards processes as essentially well-
structured and model-based. However, sometimes
variations and exceptions occur, which require additional
rules supporting ad hoc interventions. Two well-known
solutions have been proposed: exception handling and
flexible BPM. The former integrates rules and mechanisms
to handle expected and unexpected events, typically
suspending or cancelling a process to execute an
alternative handling procedure [2, 11, 12]. The latter
brings constraints to process models, thus allowing
alternative activities and flows, provided they do not
violate the specified constraints [13, 14, 40]. All in all, the
two approaches support a degree of improvisation,
process variants and ad hoc activities, however under the
scope of model-based process execution.
Event-driven processes. Event-driven BPM suggests
that we should model process events instead of activities
[15, 16]. Since activities are not defined, the process
participants effectively have ample latitude to perform the
activities in their own ways. Nevertheless, process
execution is still model-based.
Adaptive processes. An approach designated adaptive
process modelling suggests the use of process fragments
in process models [17]. Process fragments define regions
where optional models can be dynamically selected to
execute work in different ways. This approach does not
mean we can model people-driven dynamic work, but
instead that we can define regions where work may be
changed depending on contextual factors. However,
outside these regions, work is still confined to model-
based execution.
Supporting People-Driven, Dynamic and Geo-Located Work Processes S-BPM ONE 2018, April 2018, Linz, Austria
3
Semi-structured processes with collaborative
activities. This approach combines the typical model-
based BPM with social media, a concept often designated
as social BPM [18, 19]. The shared space provided by
social media can be used to assign activities to people in a
participative way. This approach definitely differs from
the previous ones by emphasizing the role of people in
process execution. It can therefore be seen as a first, albeit
limited approach to people-driven dynamic processes.
Dynamic subject-oriented processes. Social
Business Process Management (S-BPM) has recently
gained attention as a BPM alternative which emphasizes
how humans participate and collaborate within the
process scope, instead of just seeing humans as actors
designated to execute specific activities [6]. Taking this
viewpoint, researchers started to investigate how to model
dynamic behavior using particular model constructs [20].
Researchers have also started to investigate how to
support these dynamic processes [4, 21]. Even though
promising, we have not seen an actual implementation of
these concepts.
Knowledge-intensive people-driven processes. The
main focus of this approach is on the people driving
process execution. Its initial impetus was brought by the
concept of case management. In opposition to traditional
BPM, which concerns repetitive, systematic work, case
management deals with unique and knowledge-intensive
work. With case management, control is moved from
activities to a case file [22, 23]. Two more specific
approaches can be found in this category: adaptive case
management [3, 24] and emergent case management [25].
The former adds to-do lists to case files, which can be seen
as an equivalent to process models but without control-
flow [24]. The latter emphasizes the collaborative
management of case files, using in particular social media
and microblogging for communication. When compared
to the previous approaches, case management represents a
radically different way to handle work, as it becomes fully
unstructured. However, it seems too radical for the work
scenarios discussed in this paper: case management is
clearly centered on the specific case of highly-skilled
workers performing knowledge-intensive work.
Interaction-intensive people-driven processes. In
this approach, we also find a significant concern with the
human involvement in the process dynamics [26-28].
However, instead of giving primacy to knowledge-
intensive work, the emphasis is on human interaction
[20]. One innovative solution that has been suggested to
manage the execution of unstructured work consists in
using machine learning to automatically identify
dependencies between activities and to suggest who
should execute a certain activity [26, 27].
In summary, we observe that dynamic process
execution has been addressed with a range of solutions
that extend from the model-driven to the people-driven.
The former case emphasizes model-based control, which
in special circumstances can accommodate ad hoc
activities, while the latter emphasizes human discretion in
determining what to do next.
The system described in this paper adopts the latter
viewpoint with a specific focus on process execution,
which seems to be a current gap in the research literature.
Even though researchers already equated how to integrate
people-driven dynamic processes in process modelling,
from a conceptual perspective, in this research we are
concerned with process execution.
Since our system integrates geo-location with process
execution, it also seems relevant to briefly overview
research in that area. However, little work has been done
so far. [29] developed a framework for integrating
visualization applets into worklist handlers, which allows
workers to select work items taking geo-location into
consideration. [30] adopts a similar approach, integrating
GIS tools with worklist handlers. However, these two
studies do not actually address the broader problem of
integrating geo-location into process management, as they
only concern worklist handing. [5, 31, 32] extended BPM
modelling to include location-dependencies in control-
flow, e.g. using location-dependent parallel splits and
synchronizations, but they did not address the use of geo-
location in dynamic processes. To the best of our
knowledge, our research is unique in integrating geo-
location into people-driven dynamic processes beyond
worklist handing.
3 FEATURES
We now briefly discuss some key features of the
developed system.
Process structure. Process structure concerns the way
in which activities are pulled together in a process. The
most common approach to process structure is to define
process models at design time, which define the sequences
in which activities take place during execution [33].
(However, other approaches exist, e.g. based on visual
narrative [34-36].) Such models are then used by PAIS to
enact, execute and manage the processes [37]. However,
as discussed in Section 2, this approach constrains
dynamic changes during process execution. In order to
avoid this constraint, we decided to include geo-location
as an alternative process structure. Prior research [38] into
dynamic work scenarios highlighted that spatial data
provide an adequate frame for situation awareness and
action, which can be used to structure work. Therefore,
the first feature considers that:
People-driven, dynamic, and geo-located processes are
primarily structured by geographical locations.
S-BPM ONE 2018, April 2018, Linz, Austria
P. Antunes et al.
4
As with the interaction-intensive people-driven
approach discussed in Section 2, which moved process
structure away from models towards communication, or
the knowledge-intensive people-driven approach, which
moved process structure to case files, in our approach we
move process structure away from models towards
geographical locations.
Process control. Process control concerns how the
activities defined by a process are actually managed
during execution time. The most common approach is to
use model-based control-flow, where the process model is
used to determine how the sequence of activities is
executed (However, other approaches exist, e.g. adaptive
case management [24].) It derives from this viewpoint that
PAIS are effectively in control of the process. However, as
already discussed in Section 2, this approach reduces the
involvement of humans in making dynamic changes to
processes, while they are being executed. In order to avoid
this constraint, we decided to relinquish control from
PAIS and instead give that control to humans. Likewise
the knowledge-intensive people-driven approach,
processes become unframed [37]. The decisions on how to
execute activities will then depend on human
communication and determination. Therefore, the second
feature we consider is:
People-driven, dynamic, and geo-located processes are
primarily controlled by the process participants;
information systems will only have a supportive role and
will not constrain any dynamic changes required by the
process participants.
Unlike the knowledge-intensive people-driven
approach, which adopted the collaborative management of
case files, we propose combining process communication
with process context.
Process communication. In order to control the
process, the process participants will have to
communicate more (than with the model-based approach).
In particular, semi-structured messages have to be
exchanged to enact processes, initiate activities, request
and pass control over activities, notify that an activity has
been completed, etc. Unstructured messages are also
necessary, e.g. to discuss who can take responsibility for
an activity, who could substitute a participant unable to
complete an activity, or even what should be done in case
of inappropriate behavior. We suggest such
communication can be supported by microblogging:
People-driven, dynamic, and geo-located processes can be
supported by microblogging, which provides a conduit for
exchanging semi-structured and unstructured messages
about a process; process participants rely on these messages
to dynamically manage processes.
Process context. Since we move control from PAIS to
the process participants, the process participants need to
access the process context to decide which activities to
take, and when and where to executed them. Furthermore,
since activities are dynamically executed, the process
context must also track the process enactment and
evolution:
People-driven, dynamic, and geo-located processes are
supported by context, which tracks where, who and what
activities have been done.
In a way, in the scope of dynamic processes, we can regard the
process context as an alternative to the process model. Unlike
the knowledge-intensive people-driven approach, which is
centered on the case file, our approach is centered on the process
context, which includes geographical locations. Additionally, the
process context supports visibility and traceability of process
execution.
4 PEOPLE-DRIVEN, DYNAMIC AND GEO-
LOCATED PROCESS MANAGEMENT SYSTEM
A prototype was developed to explore the feasibility of
implementing a system which complies with the features
discussed in the previous section. For this reason, we
decided to implement a system having only the most
important functionalities we thought were necessary for
that. Therefore, we consider only one human role in the
system, which is able to enact processes, dynamically add
and modify activities, pick and execute activities, and
communicate with other process participants. Next, we
describe in detail the system functionality and discuss in
particular how the system uses the Twitter platform for
process communication.
4.1 Process enactment
Users login into the system using their Twitter
credentials. When logging in for the first login, they can
declare their expertise (roles), so the system can push
activities that best fit their profile. A user enacts a process
by giving it a name. This is the only required information
at the beginning, because the rest can be dynamically
specified during execution. In particular, activities and
flows can be re/defined at any time after enactment. It is
also possible to add an activity to a process at any time,
and existing activities can also be modified. Fig. 1 shows
the main view of the user interface during process
enactment. This functionality fosters open participation in
the process, which in turn depends on personal
responsibility rather than machine control to execute the
process.
Supporting People-Driven, Dynamic and Geo-Located Work Processes S-BPM ONE 2018, April 2018, Linz, Austria
5
Figure 1: Process enactment: a list of enacted processes is
shown on the upper left part of the window. The timeline
at the bottom-right shows the so far defined activities for a
selected process. The map shows the locations of activities
defined for a selected process (if location information is
defined).
According to the decision to give predominance to the
spatial context of activities, a map showing the location of
defined activities takes a prominent part of the user
interface. Along with this, the user interface also shows
the list of enacted processes and a timeline with activities.
Fig. 1 shows a process enacted with the name “Industrial
Electricity Maintenance”; it was created by user
@waldouribe with the hashtag #item_427, which follows
Twitter conventions. The timeline shows six activities.
The first one, named “Turn off the generator”, appears
slightly brighter than the others, indicating that it is the
only one that can currently be executed. We see that two
other activities can be executed in parallel, but only after
the first one is completed. When an activity is completed,
it is shown in a dark color.
After enacting a process, the following actions can be
done:
• Defining roles: a role corresponds to a specific set of
skills that may be required to execute an activity. Fig.
2 shows the user interface that can be used to define
roles for a process. Roles are part of the process
context. Since we allow to dynamically change the
process, roles are considered as part of context rather
than core process information. Furthermore, roles can
be created either before or after activities have been
defined.
Figure 2: Defining roles for activities.
• Creating an activity: activities can be created by
clicking a location on the map or the timeline. When
using the map, a pin will appear showing the place
where the activity should be executed. Activities
always appear in the timeline.
• Defining activity attributes: By selecting an
activity in the timeline or map, the following
attributes can be defined and/or changed:
o Description: brief description of what
should be done.
o Priority: a number between 1 and 10, which
suggests the importance of this activity. This
information is intended to help users
making decisions about which activities to
pick first.
o Dependencies: this is a (possibly empty)
list containing names of other activities
which are required to be completed before
this one can start. This information is taken
in order to compute the right place for the
activity in the timeline. All activities with no
dependencies are displayed in parallel at the
beginning of the timeline. The system
checks for possible inconsistencies in the
definition of dependencies, like circular
references and deadlocks.
o Roles: this is a (possibly empty) list
mentioning the roles that workers should
have to pick and execute an activity. When
more than one role is defined, the system
assumes any worker having at least one of
the listed roles can perform the activity.
When no role is specified, any worker can
perform the activity.
o Starting and ending times: these
attributes are not set by users, because they
correspond to the times when the activity
was actually executed. These attributes are
set by the system.
S-BPM ONE 2018, April 2018, Linz, Austria
P. Antunes et al.
6
Fig. 3 shows the user interface when changing the
attributes of an activity. It illustrates the definition of
attributes for activity “Clean F-20 board”. This activity has
priority 3. The activity should be performed by an
electrician. This activity can also be defined as dependent
on other activities defined for this process: “Turn off
generator” and “Replace T-450 cable”. In the darkened
background, we see that the current activity can only start
after “Replace T-450 cable” has been completed. We also
see that the disposition of the bars representing activities
in the timeline shows that “Replace T-450 cable” can start
only after “Turn generator off” is completed. We also see
that there is a fourth activity defined with the name “Turn
generator on”, which is displayed after the current one in
the timeline, meaning it has been defined to be dependent
on the completion of “Clean F-20 board”. This last activity
is not displayed when defining dependencies for “Clean F-
20 board” to avoid circular dependencies and deadlocks.
Figure 3: Defining attributes for an activity, including
dependencies between activities.
4.2 Process execution
After describing the process enactment, we now
describe the process execution. Users can pick activities
from the worklist handler shown in Fig. 4. Users can select
a process from the list shown at the top left part of the
user interface, after which a list of available activities
matching the user’s roles are displayed in a list with the
header “Work to do”. Below this list, a map shows the
locations where the activities should be performed. By
clicking on an activity, the attributes are displayed in a
pop-up window, similar to the one shown in Fig. 3. The
user can then commit to execute the activity and the
system will register the starting time. The activity will be
immediately removed from the other users’ worklist
handlers.
The order activities appear on the “Work to do” can be
determined by two different criteria: priority number, and
distance between user and activities.
Although the user interface shows all activities that can
be executed by a user, not all of them might actually be
selectable. Those activities which are dependent on other
activities that have not been completed are shown with a
darker color and are not selectable.
4.3 Twitting and process context
In Section 3, when discussing the process context, we
noted that the Twitter platform could not only be used to
support communication between the process participants
but also as a way to attach contextual information to a
process. In other words, instead of having a process
model, we have a collection of tweets explaining how the
process execution evolves over time. The tweets are
automatically sent by the system to the Twitter platform.
For this purpose, every time a user notifies the system that
an activity has started or ended, the system contacts the
REST API of Twitter and posts a tweet on the worker’s
account. The structure of these messages is shown in Fig.
5.
Figure 4: Worklist handler. The “Work to do” only shows
the activities matching the user’s’s skills. The two
activities shown in darker color have to wait the
completion of other activities.
Fig. 6 is an actual screenshot from Twitter, which
shows how messages are sent. We can see that user
frankjenson has tweeted six messages while dealing with
the #item_407 process. (Note that date and time are
implicit in the tweets.)
Of course, besides the semi-structured messages related
to the #item_407 process, we should also expect to see
unstructured messages needed to communicate about the
process execution, like “I will start activity X late because
of a traffic jam”, as explained in Section 3.
Supporting People-Driven, Dynamic and Geo-Located Work Processes S-BPM ONE 2018, April 2018, Linz, Austria
7
Figure 5: Semi-structured messages sent to Twitter on
behalf of the user.
Figure 6: Twitter page of user frankjayson with tweets
originated from the system communicating the starting
and ending of activities for process item_427.
4.4 Twitting and process control
We specify process control in the following way:
@<username> with role <role list> must <activity> at
<address> when <activity list> finishes
The character @ marks the beginning of an activity
description, followed by the username that controls the
activity execution. If no user is assigned, a question mark
(?) is used instead. The keyword “with role” marks the
beginning of the roles list which can perform the activity,
separated by commas. The “at” keyword marks the
beginning of the string containing the location where the
activity has to be executed. The <activity list>, which is
delimited by the keywords “when” and “finishes”, contains
the dependencies of this activity, i.e. other activities which
have to be finished before this one can start. Except for
the @ keyword, all other keywords in an activity
description are optional.
It should be noted that the process name does not
appear in these messages because the Twitter messages
are stored as part of the process context. In Twitter, the
messages have an associated hashtag which corresponds
to the process name.
Fig. 7 illustrates process control using an example. The
example consists of 5 activities with names Task1 to Task5
in which Task1 has to be completed before Task2, Task3
and Task4 can start, and Task5 has to wait for Task2,
Task3 and Task4 before starting. The upper half of the
figure shows the classical model-based control-flow
specification, while the lower half shows the five
sentences required to describe the same type of control
using tweets. A fundamental difference between the two
approaches is that the former is usually specified at design
time, while the later can be dynamically defined as the
process unfolds, and can be dynamically changed by
sending tweets referring to the same activities but with
different attributes.
The tweets are sent when an activity is created or
edited (for changing attributes). They are automatically
posted on Twitter using a hashtag with the process name.
Combining these tweets with the tweets notifying
when an activity starts and ends (Fig. 8), allows to
reconstruct the process enactment and execution, thus
sharing the process context between the process
participants. Currently, we have not yet implemented
functionality to reenact a previously enacted process.
Figure 7: Above: model-based control-flow of 5 activities in
which there are two “and” gates, one for diverging and
another for converging the flow. Below: the same control
using Twitter messages.
Starting activity:
@<username> twits
#<process id> <name of activity> started at
<address of activity> <date and time>
Ending activity:
@<username> twits
#<process id> <name of activity> ended <date and
time>
Task
Task
Task
Task
4
Task
@user1 must Task 1 at address1
@user2 must Task 2 at address2 when Task1 finishes
@user3 must Task 3 at address3 when Task1 finishes
@? must Task 4 at address4 when Task1 finishes
@? must Task 5 at address5 when Task3, Task4, Task5
finishes
S-BPM ONE 2018, April 2018, Linz, Austria
P. Antunes et al.
8
Figure 8: The tweets appear while editing an activity. They
are automatically sent to Twitter under the hashtag
representing the process.
5 PRELIMINARY, FORMATIVE EVALUATION
At this development stage, the main aspect we wanted to
understand was the utility of having a system which can
support people-driven processes for which the
geographical location of activities is an important
component. On a second level of importance, we also
wanted to gather feedback on how to improve usability,
which will be fundamental for conducting more formal
evaluations in the future. Considering these requirements,
a formative evaluation based on a focus group was
selected.
The participants in the evaluation were recruited
among computer engineering students who had already
passed a course where they studied business process
modelling, so they had background knowledge about the
concepts of process structure and control. Five students
were selected for the evaluation, 3 male and 2 female, all
between 21 and 24 years old.
The evaluation was organized as follows. First, the
system was described to the participants in a session of 30
minutes, and then they were asked to 1) think about a
scenario in which the system would be helpful and 2)
model in detail a concrete work process for such scenario.
For this purpose, they had two days’ time, after which the
focus group session was organized to discuss the system
support to the various scenarios and work processes
elaborated by the participants.
The scenarios chosen by the participants included
building a personal computer using parts from different
stores, servicing a truck fleet, doing car repairs on the
road, transportation of luxury cars, and even an arcade
game with activities taking place in different parts of the
world. To spark the focus group session, we raised two
initial questions: “Do you think it is a good idea to geo-
localize activities in work processes?” and “Do you think
the system can be applied to the scenarios you
developed?”.
All participants agreed that geo-located processes were
a good idea and could be applied to several scenarios. The
variety of the scenarios suggested and discussed by the
participants reinforced the positive feedback.
To further discuss the system features, we also asked:
“What do you think about adding priorities to activities?”,
and “What do you think about adding roles to activities
and filtering users according to them?”. Regarding
priorities, the participants agreed that their value was not
so clear, except for a specific scenario in which the users
that create activities are not those that perform them.
They regarded priorities as a kind of price that those who
create activities are ready to pay to those performing
them. Regarding roles, the participants agreed they could
be of help in most scenarios.
To obtain feedback on the use of tweets to control
processes, we also asked: “Do you think you could
understand a process just by looking at the tweets
generated by the system?”. Most participants agreed the
tweets were understandable, but two of them noted that it
takes some effort to reconstruct the whole process by only
using the messages. They noted that a graphical view
should be developed in the future.
To get feedback on how processes are enacted, we
asked if process enactment was an easy task, to which
most agreed it was not difficult at all. However, when
asked if the user interface was intuitive enough, they said
that it required some time to fully understand how to use
it and that an initial explanation before using it was
absolutely necessary.
When asked about which new features should the
system provide, the participants suggested: to show on the
map green pins for completed activities; to implement
"drag-and-drop" to move activities on the map; to have a
more accurate definition of activity locations; to be able to
export a process as a BPMN model; and to check if users
are near the activities’ locations when reporting
completion.
6 CONCLUSIONS
In this work, we present a system supporting people-
driven, dynamic and geo-located work processes. The
system allows the process participants to define activities
in a dynamic way, while processes are being executed. To
accomplish this, we substituted process models with
process context; and substituted model-based control-flow
with contextually-supported human control. We defined
process context as a combination of activities and other
attributes such as geographical location. This way, typical
process models have been effectively substituted by
process context. Process context ensures some desirable
properties of BPM, such as visibility and traceability,
while at the same time avoiding model-based control-flow,
which constrains dynamic changes in process execution.
Our implementation uses the Twitter platform as a
communication mechanism, allowing process participants
Supporting People-Driven, Dynamic and Geo-Located Work Processes S-BPM ONE 2018, April 2018, Linz, Austria
9
to exchange semi-structured and unstructured messages
about processes using familiar technology. Furthermore,
we also use Twitter to support the process context: all
messages exchanged about a process are stored in the
platform and can be used both by the system and the
process participants to understand how a process was
enacted and how it evolved until completion.
This work makes the following unique contributions.
Firstly, it provides an implementation of people-driven
dynamic processes. Secondly, it provides an
implementation combining people-driven dynamic
processes with geographical locations. Thirdly, it provides
a unique solution to building process context, leveraging
the possibilities brought by microblogging platforms.
A preliminary, formative evaluation action conducted
by a focus group suggested that the geographical
contextualization of processes could be of great utility.
Also, the obtained feedback supports using tweets for
enacting and communicating about process, and assigning
priorities to activities. However, the obtained feedback
suggests that, even though it is not difficult to use the
system, it requires some learning time to master it.
ACKNOWLEDGMENTS
This work was partially supported by Fondecyt Regular
1161200 and the 3rd Internal Research Contest FEN.
REFERENCES
[1] Barukh, M. and Benatallah, B. Processbase: A hybrid process
management platform. In International Conference on Service-
Oriented Computing (2014), 16-31.
[2] Antunes, P. and Mourão, H. Resilient business process
management: Framework and services. Expert Systems With
Applications, 38, 2 (2011), 1241-1254.
[3] Motahari-Nezhad, H. and Swenson, K. Adaptive case
management: Overview and research challenges. In IEEE 15th
Conference on Business Informatics (2013), 264-269.
[4] Gottanka, R. and Meyer, N. Modelasyougo:(re-) design of s-
bpm process models during execution time. In International
Conference on Subject-Oriented Business Process Management
(2012), 91-105.
[5] Zhu, X., Recker, J., Zhu, G., and Santoro, F. Exploring
location-dependency in process modeling. Business Process
Management Journal, 20, 6 (2014), 794-815.
[6] Meyer, N. and Schiffner, S. Democratizing business process
management. In IEEE 16th Conference on Business Informatics
(2014), 93-100.
[7] Antunes, P., Zurita, G., and Baloian, N. Characterization of an
ad-hoc geo-referenced bpm supported by microblogging. IEEE
Latin America Transactions, 13, 6 (2015), 1989-1994.
[8] Kabicher-Fuchs, S., Rinderle-Ma, S., Recker, J., Indulska, M.,
Charoy, F., Christiaanse, R., and Mendling, J. Human-centric
process-aware information systems (hc-pais). (2012).
[9] Schonenberg, H., Mans, R., Russell, N., Mulyar, N., and Van
der Aalst, W. Process flexibility: A survey of contemporary
approaches. In Advances in enterprise engineering i Heidelberg:
Springer (2008), 16–30.
[10] Reichert, M. and Weber, B. Enabling flexibility in process-
aware information systems: Challenges, methods, technologies.
Heidelberg: Springer (2012).
[11] Kurz, M., Fleischmann, A., Lederer, M., and Huber, S. Planning
for the unexpected: Exception handling and bpm. In
International Conference on Subject-Oriented Business Process
Management (2013), 123-149.
[12] Antunes, P. Bpm and exception handling: Focus on
organizational resilience. IEEE Transactions on System, Man,
and Cybernetics Part C: Applications and Reviews, 41, 3 (2011),
383-392.
[13] Pesic, M. Constraint-based workflow management systems :
Shifting control to users Technische Universiteit Eindhoven. ,
Eindhoven (2008).
[14] Pesic, M. and Van der Aalst, W. A declarative approach for
flexible business processes management. In International
conference on business process management (2006), 169-180.
[15] Krumeich, J., Weis, B., Werth, D., and Loos, P. Event-driven
business process management: Where are we now? A
comprehensive synthesis and analysis of literature. Business
Process Management Journal, 20, 4 (2014), 615-633.
[16] Scheer, A., Thomas, O., and Adam, O. Process modeling using
event-driven process chains. In Process-aware information
systems (2005), 119-146.
[17] Weidmann, M., Koetter, F., Kintz, M., Schleicher, D., Mietzner,
R., and Leymann, F. Adaptive business process modeling in
the internet of services (abis). In Proceedings of the Sixth
International Conference on Internet and Web Applications and
Services (2011).
[18] Brambilla, M., Fraternali, P., and Vaca, C. Bpmn and design
patterns for engineering social bpm solutions. In Bpm 2011
international workshops, clermont-ferrand, france, august 29,
2011, Daniel, F., Barkaoui, K., and Dustdar, S., Eds. Heidelberg:
Springer (2012), 219-230.
[19] Bider, I., Johannesson, P., and Perjons, E. A strategy for
merging social software with business process support. In
Business Process Management Workshops (2010), 372-383.
[20] Dorn, C., Dustdar, S., and Osterweil, L. Specifying flexible
human behavior in interaction-intensive process
environments. In International conference on business process
management: Springer (2014), 366-373.
[21] Rothschädl, T. Ad-hoc adaption of subject-oriented business
processes at runtime to support organizational learning. In S-
BPM ONE – Scientific Research (2012), 22-32.
[22] Van der Aalst, W., Weske, M., and Grunbauer, D. Case
handling: A new paradigm for business process support. Data
& Knowledge Engineering, 53, 2 (2005), 129-162.
[23] Marin, M., Hauder, M., and Matthes, F. Case management: An
evaluation of existing approaches for knowledge-intensive
processes. In International Conference on Business Process
Management (2015), 5-16.
[24] Herrmann, C. and Kurz, M. Adaptive case management:
Supporting knowledge intensive processes with it systems. In
S-BPM ONE-Learning by Doing-Doing by Learning (2011), 80-
97.
[25] Böhringer, M. Emergent case management for ad-hoc
processes: A solution based on microblogging and activity
streams. In Business process management workshops
Heidelberg: Springer (2011), 384-395.
[26] Dorn, C., Burkhart, T., Werth, D., and Dustdar, S. Self-
adjusting recommendations for people-driven ad-hoc
processes. Business Process Management Journal, (2010), 327-
342.
[27] Dorn, C. and Dustdar, S. Supporting dynamic, people-driven
processes through self-learning of message flows. In Advanced
information systems engineering Berlin: Springer (2011), 657-
671.
[28] Barukh, M. and Benatallah, B. Processbase: A hybrid process
management platform. In International conference on service-
oriented computing Heidelberg: Springer (2014), 16-31.
[29] Leoni, M., Adams, M., Van Der Aalst, W., and Hofstede, A.
Visual support for work assignment in process-aware
S-BPM ONE 2018, April 2018, Linz, Austria
P. Antunes et al.
10
information systems: Framework formalisation and
implementation. Decision Support Systems, 54, 1 (2012), 345-
361.
[30] Catarci, T., Leoni, M., Marrella, A., Mecella, M., Russo, A.,
Steinmann, R., and Bortenschlager, M. Workpad: Process
management and geo-collaboration help disaster response.
International Journal of Information Systems for Crisis Response
and Management, 3, 1 (2011), 32-49.
[31] Zhu, X., Zhu, G., and Guan, P. Exploring location-aware
process management. In Geo-informatics in resource
management and sustainable ecosystem Heidelberg: Springer
(2013), 249-256.
[32] Zhu, G. and Zhu, X. A study on geospatial constrained
process modeling using uml activity diagrams. In Asia-Pacific
Conference on Business Process Management (2014), 59-73.
[33] Benatallah, B., Sakr, S., Grigori, D., Motahari-Nezhad, H.,
Barukh, M., Gater, A., and Ryu, S. Business process
paradigms. In Process analytics: Springer (2016), 19-60.
[34] Simões, D., Antunes, P., and Carriço, L. Eliciting and
modelling business process stories: A case study. Business &
Information Systems Engineering, (2017).
[35] Antunes, P., Simões, D., Carriço, L., and Pino, J. An end-user
approach to business process modeling. Journal of Network
and Computer Applications, 36, 6 (2013), 1466-1479.
[36] Simões, D., Antunes, P., and Cranefield, J. Enriching
knowledge in business process modelling: A storytelling
approach. In Innovations in knowledge management: The
impact of social media, semantic web and cloud computing. vol.
95, Razmerita, L., Phillips-Wren, G., and Jain, L., Eds.
Heidelberg: Springer (2016), 241-267.
[37] Van der Aalst, W. Business process management: A
comprehensive survey. ISRN Software Engineering, (2013).
[38] Antunes, P., Zurita, G., Baloian, N., and Sapateiro, C.
Integrating decision-making support in geocollaboration
tools. Group Decision and Negotiation, 23, 2 (2014), 211-233.
[39] Monares, A., Ochoa, S.F., Pino, J.A., Herskovic, V., Rodriguez-
Covili, J. and Neyem, A. Mobile computing in urban
emergency situations: improving the support to firefighters in
the field. Expert Systems with Applications 38(2) (2011), 1255-
1267.
[40] Neyem, A., Franco, D., Ochoa, S.F. and Pino, J.A. Supporting
Mobile Workflow with Active Entities. In 11th International
Conference on Computer Supported Cooperative Work in Design
(2007), IEEE, 795-800.
