=Paper=
{{Paper
|id=None
|storemode=property
|title=Embodied Organizations: a Unifying Perspective in Programming Agents, Organizations and Environments
|pdfUrl=https://ceur-ws.org/Vol-627/coin_7.pdf
|volume=Vol-627
|dblpUrl=https://dblp.org/rec/conf/mallow/PiuntiBHR10
}}
==Embodied Organizations: a Unifying Perspective in Programming Agents, Organizations and Environments==
https://ceur-ws.org/Vol-627/coin_7.pdf
Embodied Organizations: a unifying perspective
in programming Agents, Organizations and
Environments
Michele Piunti1 , Olivier Boissier2 , Jomi F. Hübner3 , and Alessandro Ricci1
1
Universitá di Bologna, Italy - {michele.piunti,a.ricci}@unibo.it
2
Ecole des Mines St-Etienne, France - boissier@emse.fr
3
University of Santa Catarina, Florianópolis, Brazil - jomi@inf.furb.br
Abstract. MAS research pushes the notion of openness related to sys-
tems combining heterogeneous computational entities. Typically, those
entities answer to different purposes and functions and their integration
is a crucial issue. Starting from a comprehensive approach in developing
agents, organizations and environments, this paper devises an integrated
approach and describes a unifying programming model. It introduces the
notion of embodied organization, which is described first focusing on the
main entities as separate concerns; and, second, establishing different in-
teraction styles aimed to seamlessly integrate the various entities in a co-
herent system. An integration framework, built on top of Jason, CArtAgO
and Moise (as programming platforms for agents, environments and or-
ganizations resp.) is described as a suitable technology to build embodied
organizations in practice.
1 Introduction
Agent based approaches consider agents as autonomous entities encapsulating
their control, characterized (and specified) by epistemic states (beliefs) and mo-
tivational states (goals) which result in a goal oriented behavior. Recently, or-
ganization oriented computing in Multi Agent Systems (MAS) has been advo-
cated as a suitable computation model coping with the complex requirements
of socio-technical applications. As indicated by many authors [8, 2, 6], organiza-
tions are a powerful tool to build complex systems where computational agents
can autonomously pursue their activities exhibiting social attitudes. The orga-
nizational dimension is conceived in terms of functionalities to be exploited by
agents, while it is assumed to control social activities by monitoring and changing
those functionalities at runtime. Being conceived in terms of human organiza-
tions, i.e., being structured in terms of norms, roles and global objectives, this
perspective assumes an organizational layer aimed at promoting desired coordi-
nation, improving control and equilibrium of social dynamics. Besides, the need
for openness and interoperability requires to cope with computational environ-
ments populated by several entities, not modellable as agents or organizations,
which are supposed to be concurrently exploited by providing functionalities
98
supporting agents objectives. These aspects are even more recognized in current
ICT, characterized by a massive interplay of self-interested entities (humans
therein) developed according to different models, technologies and programming
styles. Not surprisingly, recent approaches introduced environment as pivotal di-
mension in MAS development [22, 14]. Such a multifaceted perspective risks to
turn systems into a scattered aggregation of heterogenous elements, while their
interplay, as well as their interaction, is reduced to a problem of technological
interoperability. To prevent this, besides the different mechanisms and abstrac-
tions that must be considered, there is a strong need of binding these elements
together in a flexible and clear way.
Providing a seamless integration of the above aspects places the challenge to
conceive the proper integration pattern between several entities and constructs.
A main concern is agent awareness, namely the need for agents to exhibit special
abilities and knowledge in order to bring about organizational and environmen-
tal notions—which typically are not native constructs of their architectures [21,
15]. Once the environment dimension is introduced as an additional dimension,
a second concern is how to connect in a meaningful way the organizational en-
tities and the environmental ones, thereby (i ) how the organization can ground
normative measures as regimentation and obligations in environments, and (ii )
how certain events occurring in environments may affect the global organiza-
tional configuration. These aspects enlighten a series of drawbacks on existing
approaches, either on the conceptual model and on the programming constructs
to be adopted to build systems in practice.
Taking a programming perspective, this work describes an infrastructural
support allowing to seamlessly integrate various aspects characterizing an open
MAS. In doing so, the notion of Embodied Organization is detailed, aimed at
introducing each element in the MAS as an integral part of a structured infras-
tructure. In order to reconcile organizations, agents and environments, Embodied
organization allows developers to focus on the main entities as separate concerns,
and then to establish different interaction styles aimed to seamlessly integrate
the various entities in a coherent system. In particular, the proposed approach
defines a series of basic mechanisms related to the interaction model:
i. How the agents could profitably interact with both organizational and other
environmental entities in order to attain their design objectives;
ii. How the organizational entities could control agent activities and regiment
environmental resources in order to promote desired equilibrium;
iii. How environmental changes could affect both organizational dynamics and
agents activities;
The rest of the paper is organized as follows: Section 2 provides a survey
of situated organization as proposed by existing works. Starting from the de-
scription of the basic entities characterizing an integrated perspective, Section 3
presents a unified programming model including agents, organizations and envi-
ronments. The notion of Embodied Organization is detailed in Section 4, while
Section 5 discusses a concrete programming model to implement it in practice.
99
Finally, Section 6 concludes the paper discussing the proposed approach and
future directions.
2 Organizations situated in MAS Environments
Although early approaches in organization programming have not been ad-
dressed at modeling environments explicitly, recent trends are investigating the
challenge to situate organizations in concrete computational environments. In
what follows, a survey on related works is discussed, enlightening strengths and
drawbacks of existing proposals.
2.1 Current Approaches
Several agent based approaches allow to implement situated organizations instru-
menting computational environments where social interactions are of concern. A
remarkable example of situated organization is due to Okuyama et al. [12], who
proposed the use of “normative objects” as reactive entities inspectable by agents
working in “normative places”. Normative objects can be exploited by the orga-
nization to make available information about norms that regulate the behavior
of agents within the place where such objects can be perceived by agents. Indeed,
they are supposed to indicate obligations, prohibitions, rights and are readable
pieces of information that agents can get and exploit in computational environ-
ments. The approach envisages a distributed normative infrastructure which is
assumed to control emergent dynamics and to allow agents to implicitly interact
with a normative institution. The mechanism is based on the intuition that the
reification of a particular state in a normative place may constitute the realiza-
tion of a particular institutional fact (e.g., “being on a car driver seat makes an
agent to play the role driver”). This basic idea is borrowed from John Searle’s
work on speech acts and social reality [16, 17] Searle envisaged an institutional
dimension rising out of collective agreements through special kind of rules, that
he refers as constitutive rules. Those rules constitute (and also regulate) an ac-
tivity the existence of which is logically dependent on the rules themselves, thus
forming a kind of tautology for what a constitutive rule also defines the notion
that it regulates. In this view, “being on a car driver seat makes an agent to
play the role driver” strongly situate the institutional dimension on the environ-
mental one, both regulating the concept of role adoption and, at the same time,
defining it.
Constitutive rules in the form X counts as Y in C are also at the basis of
the formal work proposed by Dastani et al. [5]. Here a normative infrastructure
(which is referred as “normative artifact”) is conceived as a centralized envi-
ronment that is explicitly conceived as a container of institutional facts, i.e.,
facts related to the normative/institutional states, and brute facts, i.e. related
to the concrete/ “physical” workplace where agents work. To shift facts from
the brute dimension to the normative one the system is assumed to handle con-
stitutive rules defined on the basis of “count-as” and “sanctioning” constructs,
100
which allows the infrastructure to recast brute facts to institutional ones. The
mechanism regulating the application of “count-as” and “sanctioning” rules is
then based on a monitoring process which is established as an infrastructural
functionality embedded inside the normative system. Thanks to this mechanism,
agents behavior can be automatically regulated through enforcing mechanisms,
i.e. without the intervention of organizational agents.
A similar approach is proposed in the work by Tinnemeier et al. [20], where
a normative programming language based on conditional obligations and pro-
hibitions is proposed. Thanks to the inclusion of the environment dimension in
the normative system, this work explicitly grounds norms either on institutional
states either on specific environmental states. In this case indeed the normative
system is also in charge of monitoring the outcomes of agent activities as per-
formed in the work environment, in so doing providing a twofold support to the
organizational dimension and to the environmental one.
With the aim to reconcile physical reality with institutional dimensions, an
integral approach has been proposed with the MASQ approach, which introduces
a meta-model promoting an analysis and design of a global systems along several
conceptual dimensions [19]. The MASQ approach relies on the less recent AGR
model, extended with an explicit support to environment as envisaged by the
AGRE and AGREEN [1]. Four dimensions are introduced, ranging from endoge-
nous aspects (related to agent’s mental attitudes) to exogenous aspects (related
to environments, society and cultures where agents are immersed). In this case,
the same infrastructure used to deploy organizational entities is also regulated
by precise rules for interactions between agents and environment entities. The
resulting interaction model relies on the theory of influences and reactions [9],
in the context of which several interaction styles can be established among the
heterogenous entities dwelling the system.
Besides conceptual and formal integration, few approaches have accounted a
programming approach for situated organizations. By relating situated activities
in the workplace, the Brahms platform endows human work practices and allows
to represent the relations of people, locations, agent systems, communication and
information content [18]. Based on existing theories of situated action, activity
theory and distributed cognition, the Brahms language promotes the interplay
of intelligent software agents with humans their organizations. A similar idea is
provided by Situated Electronic Institutions (SEI) [4], recently proposed as an
extension of Electronic Institutions (EI) [7]. Besides providing a runtime man-
agement of the normative specification of dialogic interactions between agents,
the notion of observability of environment states is at the basis of SEI. They
are aimed at interceding between real environments and EI. In this case, special
governors, namely modelers, allow to bridge environmental structures to the in-
stitution by instrumenting environments with “embodied” devices controlled by
the institutional apparatus. Participating agents can, in this case, perform indi-
vidual actions and interactions (either non message based) while operating upon
concrete devices inside the environment. Besides, SEI introduces the notion of
staff agents, namely organization aware agents which role is to monitor ongoing
101
activities performed by agents which are not under the direct control of the in-
stitution. Staff agents are then assumed to bridge the gap between participating
agents and the institutional dimensions: they typically react to norm violations,
possibly ascribing sanctioning and enforcements to disobeying agents. Institu-
tional control is also introduced by the mean of feedback mechanisms aimed at
comparing observed properties with certain expected values. On the basis of pos-
sible not standard properties detected, an autonomic mechanism specifies how
reconfigure the institution in order to re-establish equilibrium.
The ORA4MAS approach [11] proposed a programming model for concretely
building systems integrating organizational functionalities in instrumented work
environment. In ORA4MAS organizational entities are viewed as artifact based
infrastructures. Specialized organizational artifacts (OAs) are assumed to en-
capsulate organizational functions, which can be exploited by agents to fulfill
their organizational purposes. Using artifacts as basic building blocks of orga-
nizations, allows agents to natively interact with the organizational entity at a
proper abstraction level, namely without being constrained to shape external
actions as mechanism-level primitives needed to work with middleware objects.
The consequence is that the infrastructure does not rely on a sort of hidden com-
ponents, but the organizational layer is placed beside the agents as a suitable
set of services and functionalities to be dynamically exploited (and created) as
an integral part of the MAS work environment. On the other side, ORA4MAS
does not provide an explicit support to environmental resources which are not
included in the organizational specification. Two types of agents are assumed
to evolve in ORA4MAS systems: (i) participating agents, assumed to join the
organization in order to exploit its functions (i.e., adopting roles, committing
missions etc.), while (ii) organization aware agents, assumed to manage the
organization by making changes to its functional and structural aspects (i.e.,
creating and updating functional schemes or groups) or to make decisions about
the deontic events (i.e. norm violations).
2.2 Open Issues and Challenges
Despite the richness of the models proposed for organizations of agents situated
in computational environments, many aspects are still under discussion and have
still to converge in a shared perspective between the different research lines. In
the literature, this variety of approaches have been dealt with separately, each
forming a different piece of a global view, with few consideration for how they
could fit all together. On these basis, we here enlighten a series of current issues
and challenges which our approach, described later on, is going to face with.
Agents/Organisations/Environments Interactions Typically interactions
are based on a sub-agentive level, and are founded on protocols and mechanisms,
instead on being based on the effective capabilities and functionalities exhibited
by the entities involved in the whole system. Different approaches are provided
for the interaction model between environment, agents and their organizations.
Besides, there is not a clear vision on how environment and organizational en-
102
tities should support agents in their native capabilities, as for instance the ones
related to action and perception.
Grounding Goals The computational treatments of goals clashes different ap-
proaches once they are referred to agents and their subjective goals, and when
they are related to organizations and their global goals. For instance, approaches
as MASQ, ORA4MAS describe in a rather abstract terms (i) how the subjective
and global goals should be fulfilled in practice; (ii) which brute state has to be
reached in order to consider a goal as achieved. By considering environments ex-
plicitly, either agents and organizations should be able to ground goals to actual
environment configurations, thus recognizing the fulfillment of their objectives
once the pursued goals have been reached in practice (this approach is adopted,
for instance, in [5]). Other approaches, as for instance ORA4MAS [11], do not
assume organizations able to automatically detect the fulfillment of global goals
in terms of environment configurations.
Grounding Norms As for goals, a weak support is provided for grounding
norms in concrete application domains, thus allowing to establish how and when
a norm has been fulfilled or violated. Furthermore few studies have been ad-
dressed at managing norm lifecycle with respect to distributed and (highly)
dynamic environments. No agreement is then established on which kind of mon-
itoring and sanctioning mechanisms must be adopted. Some approaches envisage
the role of organizational/staff agents [4], other approaches propose the sole au-
tomatic regulation provided by a programmable infrastructure [5, 20].
Agent Awareness It is not clear which kind of capability, and which grade
of awareness, is required for agents to exploit the functionalities provided by
the (situated) entities embedding organizational and environmental resources.
Related to organizations, some approaches propose agents able to automatically
internalize organizational specifications (i.e. MASQ, “normative objects”), other
approaches, as (ORA4MAS and SEI) assume agents’ awareness to be encoded at
a programming level.
Openness Concerns about interoperability and openness cross each of the above
mentioned aspects. Few approaches account technological integration, for in-
stance with respect to varying agent architectures, protocols and data types.
Besides, the described proposals typically focus on a restricted set of interaction
styles (i.e. dialogical interactions supported by an institutional infrastructure
in SEI, environment mediated interactions in normative objects, an hybrid ap-
proach in ORA4MAS).
With the aim to respond the above mentioned challenges, the next sections
describe an integrated approach aimed at devising a unified programming model
seamlessly integrating agents, organizations and environments.
3 Unifying Agents, Organizations and Environments
Programming
This section figures out the main elements characterizing an Embodied Orga-
nization. It envisages an integrated MAS in terms of societies of agents, envi-
103
������� �����
����
������ ������� ������
���� ���� ����
����� �����
����� �����
�������� ������������ ����
�����
������
����� ����������� �����������
�����������
�������� ������������
�����������
�������������
����
���������
�����
���� �������������
����
���������������������������� �������������������������
Fig. 1. Structural (a) and Normative (b) specifications for the hospital scenario, rep-
resented using the Moise graphical notation.
ronmental and organizational entities. In doing this, we refer to the consistent
body of work already addressed at specifying existing computational models,
while only the aspects which are relevant for the purposes of this work will be
detailed. In particular, we refer to Jason [3] as agent development framework,
CArtAgO [14] for environments and Moise [10] for organizations.
In order to ease the description, the approach will be sketched in the context
of an hospital scenario. It summarizes the dynamics of an ambulatory room, and
can be seen as an open system, where heterogenous agents can enter and leave in
order to fulfill their purposes. In particular, two types of agents are modeled as
organization participants. Staff agents (namely physicians and medical nurses)
are assumed to cooperate with each other in order to provide medical assistance
to visitors. Accordingly, visitor agents (namely patients and escorts) are assumed
to interact themselves in order to book and exploit the medical examinations
provided by the staff.
3.1 Organizations
The first considered dimension concerns the organization. We do adopt the
Moise model, which allows to specify an organization based on three differ-
ent dimensions referred as (i ) structural, (ii ) functional, and (iii ) normative4 .
The Structural Specification (SS) provides the organizational structure in terms
of groups of agents, roles and functional relations between roles (links). A role
defines the behavioral scope of agents actually playing it, thus providing a stan-
dardized pattern of behavior for the autonomous part of the system. An inheri-
tance relation can be specified, indicating roles that extend and inherit properties
from parent roles. As showed in Fig. 1 (left), visitor agents can adopt two roles,
patient and escort, both inheriting from a visitor abstract role. The doctor role
4
We here provide a synthesis of the Moise approach showing the specification of the
hospital scenario. For a more detailed description, see [10].
104
is assumed to be played by a physician. It extends the properties of a more
generic staff role, which is assigned in support and administration activities in-
side the group. Relationships can be specified between roles to define authorities,
communication channels and acquaintance links. Groups consist in a set of roles
and related properties and links. In the hospital scenario escorts and patients
form visit groups, while staff and doctor from staff groups. The specification
allows taxonomies of groups (i.e., escorts and patients forming visit group), and
intra-group links, stating that an agent playing the source role is linked to all
agents playing the target role. Notice that the cardinalities for roles inside a
group are specified, indicating the maximum amount of agents allowed to play
that role. The constraints imposed by the SS allow to establish global properties
on groups, e.g. the well-formedness property means to complain role cardinality,
compatibility, and so on.
The Functional Specification (FS) gives a set of functional schemes specifying
how, according with the SS, various groups of agents are expected to achieve
their global (organizational) goals. The related schemes can be seen as goal
decomposition trees, where the root is a goal to be achieved by the overall group
and the leafs are goals that can be achieved by the single agents. A mission
defines all the goals an agent commits to when participating in the execution of
a scheme and, accordingly, groups together coherent goals which are assigned to a
role in a group. The FS for the hospital scenario (Fig. 2) presents three rehearsed
schemes. The visitor scheme (visitorSch) describes the goal tree related to the
visitor group. It specifies three missions, namely mVisit as the mission to which
each agent joining the visit group has to commit, mPatient as the mission to be
committed by the patient who has to undergo the medical visit, and mPay as
the mission to be committed by at least one agent in the visit group. Notice that
the goals “do the visit” (which is related to the mission mPatient) and “pay
visit” (which is related to the mission mPay) can be fulfilled in parallel. The
monitorSch describes the activities performed by a staff agent. These plans are
aimed at verifying if the activities performed by the visitors follow an expected
outcome, namely if the visitors fulfill the payment committing the mPay mission
(which includes the “pay visit” goal). Finally, the docSch specifies the activities
to which a doctor has to commit, namely to perform the visit to every patient.
Notice that each mission has a further property specifying the maximum amount
of time than an agent has to commit to the mission (“time to fulfill”, or ttf
value). The FS also defines the expected cardinality for every mission in the
scheme, namely the number of agents inside the group who may commit a given
mission without violating the scheme constraints.
The Normative Specification (NS) relates roles (as they are specified in the
SS) to missions (as they are specified in the FS) by specifying a set of norms.
Moise norms result in terms of permissions or obligations to commit to a mission.
This allows goals to be indirectly related to roles and groups, i.e. through the
policies specified for mission commitment. Fig. 1 (right) shows the declarative
specification of the norms regulating the hospital scenario, and refers to the
missions described in Fig. 2. “Time to fulfill” (ttf ) values refer to the maximum
105
amount of time the organization expects for the agent to fulfill a norm. For
instance, norms n1 and n2 define an obligation for agents playing either patient
and escort roles to commit to the mVisit mission. A patient is further obliged
to commit to mPatient mission (n3). The norm n10 is activated only when the
norm n6 is not fulfilled: It specifies an obligation for a doctor to commit the
mStaff mission, if no other staff agent is committing to it inside the group.
Based on the constraints specified within the SS and FS, the NS is assumed to
include an additional set of norms which are automatically generated in order
to control role cardinality, goal compliance, deadline of commitments, etc.
The concrete computational entities based on the above detailed specifica-
tion have been developed based on an extended version of ORA4MAS [11]. This
programming approach envisages organizational artifacts (OA) are those non-
autonomous computational entities adopted to reify organizations at runtime,
thereby implementing the institutional dimension within the MAS. In particu-
lar, ORA4MAS adopts two types of artifacts, referred as scheme and group arti-
facts, which manage the organizational aspects as specified in Moise’s functional,
structural and normative dimensions. The resulting system has been referred as
Organizational Management Infrastructure (OMI), where the term infrastruc-
ture can be understood from an agent perspective: it embeds those organizational
functionalities exploitable by agents to participate the organizational activities
and to access organization resources possibly exploiting, creating and modifying
OAs on the need. Of course, in order to suitably exploit the OMI functionali-
ties, agents need to be equipped with special capabilities and knowledge about
the organizational structures, that is what in Subsection 2.2 we refer as agent
awareness.
3.2 Environments
As said in Subsection 2.1, the ORA4MAS approach does not support environ-
ments besides organizational functionalities. To this end, dually to the OMI, an
Environment Management Infrastructure (EMI) is introduced to embed the set
of environmental entities aimed at supporting pragmatic functionalities. While
artifacts are adopted as basic building blocks to implement the EMI, environ-
ments also make use of workspaces (e.g., an Hospital workspace is assumed
to contain the hospital infrastructures). Artifacts are adopted in this case to
provide a concrete (brute) dimension – at the environment level – to the global
system. Workspace are adopted in order to model a notion of locality in terms
of an application domain.
As Fig. 2 shows, it is quite straightforward to find a basic set of Environment
Artifacts (EA) building the EMI. Taking an agent perspective, the developer here
simply imagines which kind of service may be required for the fulfillment of the
various missions/goals, thus mapping artifact functionalities to the functional
specification given by the Moise FS.
Designing an EMI is thus not dissimilar to instrumenting a real workplace
in the human case: (i ) to model the hospital room it will be used a specialized
hospital workspace, (ii ) to automate bookings it will be provided a Desk artifact,
106
���������� ���������� ������
�������� ������� ������
������ ������ ������
������ ������ ����� ������ ������ ����
����� ����� ���� ����������� ������
�������
�������� ��������� �������
�������� ����
������ ��� ���� ����
����� ����� ���� �����
���� ���
��� ���
���� �������������� ����� ���
��������� ��� ����� �������������
��� �������������� ��� ��� �������
������������� ������������� ������������� �������� ��������
�������� ������� �������� �������� �������
�������� �������� ������������
������
����������� ��������� ������� ��������
���������� ��� ��������� ������� �������
��������������
������� �������������� ���� ������������� ��������
Fig. 2. (Above) Moise Functional Specification (FS) for the hospital scenario. Schemes
are used to coordinate the behavior of autonomous agents. (Below) FS is used to find
a set of environmental artifacts, and to map their functionalities in the EMI.
(iii ) to finalize visits it will be provided a (program running on an) Surgery
Tablet artifact, (iv ) to automate payments it will be provided a Billing Machine
artifact, and (v ) to send fees and bills it will be provided a Terminal artifact.
3.3 Agents
Besides the abstract indication of the different artifacts exploitable at the en-
vironment level, the Fig. 2 also shows the actions to be performed by agents
for achieving their goals. Thanks to the CArtAgO integration technology, several
agent platforms are actually enabled to play in environments: seamless interoper-
ability is provided by implementing a basic set of actions, and related perception
mechanisms, allowing agents to interact with artifacts and workspaces [14, 15].
Those actions are directly mapped into artifact operations (functions), or ad-
dressed to the workspace: in the case of the EMI, a Jason agent has to perform
a joinWorkspace("Hospital") action to enter the room (which is related to
the mVisit mission); to book the visit (related to the mVisit mission) the ac-
tion bookvisit()[artifact name("Desk")] has to be performed on the desk
artifact, and so on (see Fig. 2, below).
The same semantic mapping agents’ actions into artifact operations is adopted
to describe interactions between agents and OMI: e.g., commitMission is an op-
eration that can be used by agents upon the scheme artifact to notify mission
commitments; adoptRole (or leaveRole) can be used by an agent upon the
group artifact in order to adopt (leave) a given role inside the group, etc.
Fig. 3 (left) shows a global picture of the resulting system. As showed, agents
fulfill their goals and coordinate themselves by interacting with EMI artifacts,
107
�����������
�����
��� �����
�����������
����� ��������� �������
�������� ������������
����� ������
�������� ��������
����������� �������������
���� ����������������� ���� ���������
�������������� �������������� ������������
��������� ��������
������������
�������
��������������
�������� �����
�������� ���� ���� ����
�������
���������
�����
���������
Fig. 3. (Left) Global view of the system presents an open set of agents at work with
infrastructures managing Environment and Organization. Functional relationships be-
tween EMI and OMI are established by count-as and enact rules. (Right) Meta-model
for Organizational Embodied Rules, used to implement count-as and enact rules.
while staff agents, which we assume as special agents aware of organizational
functionalities, can directly interact with the OMI. Both these dimensions are
an integral part of the global infrastructure and, most important, can be dynam-
ically exploited by agents to serve their purposes. From an agent perspective,
the whole system can be understood as a set of facts and functions, which are
exploited, from time to time, to the organizational and environmental dimen-
sions. Through artifacts, the global infrastructure provides observable states,
namely information readable by agents for improving their knowledge. Artifacts
also provide operations, namely process based functionalities, aimed at being ex-
ploited by agents for externalizing activities in terms of external actions. Thus,
the epistemic nature of observable properties can be addressed to the infor-
mational dimension of the whole infrastructure, while the pragmatic nature of
artifact operations is assumed to cover the functional dimension.
4 Embodied Organizations
As far as the global system is conceived, EMI and OMI are situated side by
side inside the same work environment, but they are conceived as separated sys-
tems. They are assumed to face distinct application domains, the former being
related to concrete environment functionalities and the latter dealing specifi-
cally with organizational ones. The notion of Embodied Organization provides
a more strict integration: it further identifies and implements additional mech-
anisms and conceives a unified infrastructure enabling functional relationships
between EMI and OMI. As some of the approaches discussed in Section 2, we
theoretically found this relationship on Searle’s notion of constitutive rules. Dif-
ferently from other approaches, we ground the notion of Embodied Organization
108
on a concrete programming model, as the one who lead us to the implementation
of EMI and OMI. As explained below, Embodied Organizations rely on a revised
management of events in CArtAgO, and can be specified by special programming
constructs referred as Emb-Org-Rules.
4.1 Events
A crucial element characterizing Embodied Organizations is given by the re-
newed workspace kernel based on events. Events are records of significant changes
in the application domain, handled at a platform level inside CArtAgO. They are
referred to both state and processes to represent the transitions of configurations
inside workspaces. Each event is represented by a type,value pair (�evt , evv �):
Event type indicates the type of the event (i.e., join req indicating agents join-
ing workspace, op completed indicating the completion of an artifact operation,
signal indicating events signalled within artifact operation execution, and so
on); Event value gives additional information about the event (i.e., the source
of the event, its informational content, and so on). Due to the lack of space, the
complete list of events, together with the description of the mechanism underly-
ing event processing, can not be described here. The interested reader can find
the complete model, including the formal transition system, in [13]. We here em-
phasize the relevance of events, which have the twofold role (i ) to be perceived
or triggered by agents (i.e. focusing/using artifacts) and (ii ) to be collected and
ranked within the workspace in order to trace the global dynamic of the system.
4.2 Embodied Organization Rules
While the former role played by events refers to the interaction between agents
and artifacts, the second role is exploited to identify, and possibly govern, intra-
workspace dynamics. On such a basis, the notion of Embodied Organization refers
to the particular class of situated organization structured in terms of artifact
based infrastructures and governed by constitutive rules based on workspace
events. Events are originated within the infrastructure, being produced by envi-
ronmental and organizational entities. Computing constitutive rules is realized
by Emb-Org-Rule, which consist of a programmable constructs “gluing” together
organizational and environmental dimensions. An abstract model of this pro-
cess is shown by the dotted arrows between EMI and OMI in Fig. 3 (right).
Structures defining Emb-Org-Rule refer to count-as and enact relations.
Count-as rules state which are the consequences, at the organizational level,
for an event generated inside the overall infrastructure. They indicate how, since
the actions performed by the agents, the system automatically detects relevant
events, thus transforming them to the application of a set of operators aimed
at changing the configuration of the Embodied Organization. In so doing, either
relevant events occurring inside the EMI (possibly triggered by agents actions),
either events occurring in the context of the organization itself (OMI) can be
vehicled to the institutional dimension: these events can be further translated in
109
+join_req(Ag)
+ws_leaved(Ag)
-> make("visitorGroupBoard",
-> apply("visitorGroupBoard",
"OMI.GroupBoard",
leaveRole(Ag, "patient")).
["moise/hospital.xml","visitGroup"]);
make("visitorSchBoard",
+op_completed("BillingMachine",
"OMI.SchemeBoard",
Ag, pay)
["moise/hospital.xml","visitorSch"]);
-> apply("visitorSchBoard",
apply("visitorGroupBoard",
setGoalAchieved(Ag, pay_visit)).
adoptRole(Ag, "patient"));
include(Ag).
+op_completed("Terminal",
+op_completed("visitorGroupBoard", _,
Ag, sendFee)
adoptRole(Ag, "patient"))
-> apply("monitorSchBoard",
-> apply("visitorSchBoard",
setGoalAchieved(Ag, send_fee)).
commitMission(Ag, "mPat")).
Table 1. Example of Emb-Org-Rule (count-as) in the hospital scenario.
+signal("monitorSchBoard",
goal_non_compliance,
+signal("visitorGroupBoard",
obligation(Ag,
role_cardinality, visitor)
ngoa(monitorSch,mRew,send_bill),
-> disable("Desk", bookVisit).
achieved(monitorSch,send_bill,Ag), TTF)
-> exclude(Ag).
Table 2. Example of Emb-Org-Rule (enact) in the hospital scenario.
the opportune institutional changes inside the OMI, that is assumed to update
accordingly.
Enact rules state, for each institutional event, which is the control feedback
at the environmental level. Hence, enact rules express how the organizational
entities automatically control the environmental ones. The use of enact rules
allows to exploit organizational events (i.e. role adoption, mission commitment)
in order to elicit changes in the environment.
5 Programming Embodied Organizations
Embodied Organizations enable a unified perspective on agents, organizations
and environments by conceiving an interaction space based on a twofold in-
frastructure governed by events and constitutive rules (Emb-Org-Rules). In this
section examples of programming such rules are discussed.
Programming Count-as Rules According to the Moise FS previously de-
fined, the organization expects that an agent vaid joining the hospital workspace
is assumed to play the role visitor, which purpose is to book a medical visit and
possibly achieve it. Thus, an event join req, �vaid , t�, dispatched once an agent
vaid tries to enter the workspace, from the point of view of the organization
“count-as” creating a new position related to the visit group. Making the event
join req to “count as” vaid adopting the role visitor, is specified by the first rule
in Table 1 (left): it states that since an event signalling that an agent Ag is join-
ing the workspace, an Emb-Org-Rule must be applied to the system. The body
of the rule specifies that two new instances of organizational artifacts related to
the visit group will be created using the make operator. In this case the new
110
artifacts will be identified by visitorGroupBoard and visitorSchBoard. The
following operator constitutes the new role inside the group: apply acts on the
visitorGroupBoard artifact just created by automatically making the agent Ag
to adopt the role patient. Finally, once the adopt role operator succeeds, the last
operator includes the agent Ag in the workspace.
In the above described scenario, the effect of the application of the rule
provides an institutional outcome to the joinWorkspace actions. Besides joining
the workspace, a sequence of operators is applied establishing what this event
means in organizational terms. When the effects of the role-adoption are com-
mitted, as previously described, a new event is generated by the group board:
�op completed, �"visitorGroupBoard", vaid , adoptRole, patient ��. For the
organization, such an event may “count-as” committing to mission mP at on
the visitorSchBoard. This relation is specified by the second rule in Table 1,
where a commitMission is applied to the visitorSchBoard for the mission
mPat. Similarly, an event �ws leaved, �vaid , t��, signalling that the visitor agent
has left the workspace, from an organizational perspective “count-as” leaving
the role patient. This relation is specified by the first rule in Table 1 (right),
where a leaveRole is applied to the visitorGroupBoard for the role patient.
At the same time, an event like �op completed, �BillingMachine, vaid , pay, t��
signals that a visitor agent has successfully finalized the pay operation upon
the billing machine. Such an event “count-as” having achieved the goal pay
visit on the visitorSchBoard (second rule in Table 1, right). Finally, an event
�op completed, �Terminal, said , sendFee , t��, signalling that a staff agent has
successfully used the terminal to send the fee to a given patient, “count-as”
having achieved the goal send fee (third rule in Table 1, right).
Programming Enact Rules Enact effects are defined to indicate how, from the
events occurring at the institutional level, some control feedback can be applied
to the environmental infrastructure. As far as the execution of the operations
is conceived in CArtAgO, the OMI automatically dispatches events signalling
ongoing violations. Violations are thus organizational events which may suddenly
elicit the application of some enact rule used to regiment the environment.
In Table 2, a regimentation is installed by the organization thanks to the en-
act rule stating that an event �signal, �visitorGroupBoard, role cardinality,
∅, t�� signalled by the visitorGroupBoard indicates the violation for the norm
role cardinality. The related enact rule is given in Table 2 (left), where the re-
action to this event is specified in order to disable the book operation on the desk
artifact, for all the agents inside the workspace. The absence of any parameter
related to agent identifier in the disable("Desk", bookVisit) operator makes
the disabling to affect the overall set of agents inside the workspace. Similarly, vi-
olating the obligation imposed to the staff agent to fulfill sanctioning and reward-
ing missions elicits the scheme board assigned to the monitorSch to signal the
event �signal, �monitorSchBoard, goal non compliance,obligation(Ag,ngoa(
monitorSch,mRew,send bill),achieved(monitorSch,send bill,Ag),TTF), t��.
This event is generated thanks to a special norm (called goal non compliance)
which is automatically generated since the Moise specification and stored in-
111
side the OMI. Due to the enact rule specified in Table 2 (right), this causes the
exclusion for the Ag agent from the hospital workspace.
6 Conclusion and Perspectives
In this paper Embodied Organizations have been introduced as a unified pro-
gramming model promoting a seamless integration of environmental and organi-
zational dimensions of a MAS. A series of responses to the challenges envisaged
in Subsection 2.2 could be listed: Infrastructures. Either environmental and
organizational entities are implemented in concrete infrastructures instrument-
ing workspaces, decentralized in specialized artifacts which serve informational
and operational functions. Interaction. The approach establishes a coherent
semantic for agent - infrastructure interactions, Embodied Organizations define
functional relationships between the heterogenous entities at the basis of orga-
nizations and environments. These are placed in terms of programmable con-
structs (Emb-Org-Rules), governed by workspace events and inspired by Searle’s
notion of constitutive rules. Goals and Norms. Implementing organizations in
concrete environments allows to deal explicitly with goals and norms, which ful-
fillment can be structurally monitored and promoted at the organizational level
through the use of artifacts. Awareness. Embodied Organizations are aimed to
fit the work of agents and accordingly to allow them to externalize pragmatic
and organizational activities. The use of Emb-Org-Rule automates and promotes
specific organizational patterns, to which agents may effortlessly participate sim-
ply by exploiting environmental resources. Artifacts can be used in goal oriented
activities, and, most important, without the need to be aware of organizational
notions like roles, norms, etc. Openness. Technological interoperability is en-
sured at a system level, by providing mechanisms for agent-artifact interactions
which are based on a coherent semantic. Besides, several interaction styles can
be established at an application level, being agents mediated by infrastructures
which can be modified, replaced and created on the need.
Future work will be addressed at covering missing aspects, such as the dia-
logical dimension of interactions, and the inclusion of real embodied entities in
the system (i.e., humans, robots, etc.). An important objective is the definition
of a general purpose approach, towards the full adoption of the proposed model
in the context of concrete application domains and mainstream agent oriented
programming.
References
1. José-Antonio Báez-Barranco, Tiberiu Stratulat, and Jacques Ferber. A unified
model for physical and social environments. In Environments for Multi-Agent
Systems III, Third International Workshop (E4MAS 2006), volume 4389 of Lecture
Notes in Computer Science, pages 41–50. Springer, 2006.
2. Olivier Boissier, Jomi Fred Hübner, and Jaime Simão Sichman. Organization Ori-
ented Programming: From Closed to Open Organizations. In Engineering Societies
112
for Agent Worlds (ESAW-2006). Extended and Revised version in Lecture Notes
in Computer Science LNCS series, Springer, pages 86–105, 2006.
3. Rafael H. Bordini, Jomi Fred Hübner, and Michael Wooldrige. Programming Multi-
Agent Systems in AgentSpeak using Jason. Wiley Series in Agent Technology. John
Wiley & Sons, 2007.
4. Jordi Campos, Maite Lòopez-Sànchez, Juan A. Rodrı̀guez-Aguilar, and Marc Es-
teva. Formalising Situatedness and Adaptation in Electronic Institutions. In
COIN-08, Proc., 2008.
5. Mehdi Dastani, Nick Tinnemeier, and John-Jules CH. Meyer. A programming
language for normative multi-agent systems. In Multi-Agent Systems: Semantics
and Dynamics of Organizational Models. IGI-Global, 2009.
6. Virginia Dignum, editor. Handbook of Research on Multi-Agent Systems: Semantics
and Dynamics of Organizational Models. IGI-Global, 2009.
7. Marc Esteva, Juan A. Rodrı́guez-Aguilar, Bruno Rosell, and Josep L. AMELI: An
agent-based middleware for electronic institutions. In Proceedings of International
conference on Autonomous Agents and Multi Agent Systems (AAMAS’04), pages
236–243, New York, 2004. ACM.
8. Jacques Ferber, Olivier Gutknecht, and Fabien Michel. From Agents to Organiza-
tions: An Organizational View of Multi-agent Systems. In Proceedings of (AOSE-
03), volume 2935 of Lecture Notes Computer Science (LNCS). Springer, 2003.
9. Jacques Ferber and Jean-Pierre Müller. Influences and Reaction: a Model of Situ-
ated Multi-Agent Systems. In Proc. of the 2nd Int. Conf. on Multi-Agent Systems
(ICMAS’96). AAAI, 1996.
10. Jomi F. Hübner, , Jaime S. Sichman, and Olivier Boissier. Developing organised
multi-agent systems using the MOISE+ model: Programming issues at the system
and agent levels. International Journal of Agent-Oriented Software Engineering,
1(3/4):370–395, 2007.
11. Jomi F. Hübner, Olivier Boissier, Rosine Kitio, and Alessandro Ricci. Instrument-
ing Multi-Agent Organisations with Organisational Artifacts and Agents. Journal
of Autonomous Agents and Multi-Agent Systems, April 2009.
12. Fabio Y. Okuyama, Rafael H. Bordini, and Antônio Carlos da Rocha Costa. A
Distributed Normative Infrastructure for Situated Multi-Agent Organisations. In
Decl. Agent Lang. & Techn. (DALT-VI), volume 5397 of LNCS. Springer, 2009.
13. Michele Piunti. Designing and Programming Organizational Infrastructures for
Agents situated in Artifact-based Environments. PhD thesis, ALMA MATER
STUDIORUM Universitá di Bologna, April 2010.
14. Alessandro Ricci, Michele Piunti, and Mirko Viroli. Environment programming
in multi-agent systems: An artifact-based perspective. Autonomous Agents and
Multi-Agent Systems, 2010. Springer, ISSN 1387-2532 (Print) 1573-7454 (Online).
15. Alessandro Ricci, Andrea Santi, and Michele Piunti. Action and Perception in
Multi-Agent Programming Languages: From Exogenous to Endogenous Environ-
ments. In Proceedings Programming Multiagent Systems (PROMAS-10), 2010.
16. John R. Searle. Speech Acts, chapter What is a Speech Act? Cambridge University
Press, 1964.
17. John R. Searle. The Construction of Social Reality. Free Press, 1997.
18. M. Sierhuis. Modeling and Simulating Work Practice; Brahms: A multiagent mod-
eling and simulation language for work system analysis and design. PhD thesis,
University of Amsterdam, SIKS Dissertation Series, 2001.
19. Tiberiu Stratulat, Jacques Ferber, and John Tranier. MASQ: Towards an Integral
Approach of Agent-Based Interaction. In Proc. of 8th Conf. on Agents and Multi
Agent Systems (AAMAS-09), 2009.
113
20. Nick Tinnemeier, Mehdi Dastani, J.-J.Ch. Meyer, and L. van der Torre. Pro-
gramming normative artifacts with declarative obligations and prohibitions. In
IEEE/WIC/ACM International Joint Conference on Web Intelligence and Intel-
ligent Agent Technology (WI-IAT 2009), 2009.
21. M. Birna van Riemsdijk, Koen Hindriks, and Catholijin Jonker. Programming
organisation-aware agents: a research agenda. In In 10th Engineering Societies for
Agents Worlds (ESAW 09), 2009.
22. Danny Weyns, Andrea Omicini, and James J. Odell. Environment as a first-class
abstraction in multi-agent systems. Autonomous Agents and Multi-Agent Systems,
14(1):5–30, February 2007. Special Issue on Environments for Multi-agent Systems.
114