=Paper=
{{Paper
|id=Vol-3182/paper15
|storemode=property
|title=Multi-paradigm Modelling for Policy-driven Socio-technical Systems
|pdfUrl=https://ceur-ws.org/Vol-3182/paper15.pdf
|volume=Vol-3182
|authors=Ada Diaconescu,Dominique Blouin,Alice Ludvig
|dblpUrl=https://dblp.org/rec/conf/jurix/DiaconescuBL21
}}
==Multi-paradigm Modelling for Policy-driven Socio-technical Systems==
https://ceur-ws.org/Vol-3182/paper15.pdf
Multi-paradigm Modelling for Policy-driven
Socio-technical Systems
Ada Diaconescu1 , Dominique Blouin1 and Alice Ludvig2
1
LTCI Lab, Telecom Paris, Institut Polytechnique de Paris, 19 place Marguerite Perey, 91120, Palaiseau, France
2
University of Natural Resources and Life Sciences BOKU, 33 Gregor-Mendel Street, 1180, Vienna, Austria
Abstract
Today’s socio-technical systems, intertwined with fast-changing cyber-physical infrastructures, are
becoming increasingly complex and must be able to cope with major unexpected events - e.g. health,
economic and ecology crisis; digital innovation, pandemics, etc. As a result, these systems, which are
governed by policies, experience rapid developments in their regulation requirements. In this context,
traditional policy-making processes are slow relative to the present pace of changes they face, often
leading to outdated inappropriate governance. We present a research plan for developing a highly
configurable framework and set of tools to help policy making by providing support to better specify,
analyse, monitor and assess socio-technical systems, taking into account governing policies and their
impacts. The framework is inspired from model-based systems engineering approaches, which have
been successful for Cyber-Physical Systems, to better formulate, characterise and analyse socio-technical
systems and their governing policies. It makes use at its heart of Multi-Paradigm Modelling, to develop,
reuse and integrate appropriate domain-specific modelling languages and tools, views and analyses to
better address policy making of nowadays rapidly changing complex heterogeneous socio-technical
systems.
Keywords
Socio-Technical Systems, Cyber-Physical Systems, Self-Adaptive Systems, Multi-Paradigm Modelling,
Model-Based Systems Engineering, Model Analyses, Simulation
1. Introduction
Modern societies, increasingly intertwined with fast-changing cyber-physical infrastructures
(i.e. hence leading to socio-technical systems), experience rapid developments in their regulation
requirements. In this context, traditional policy-making processes are rather slow relative to
the present pace of societal changes, often leading to outdated policies – i.e. ‘institutional lag’.
Evolving such regulations involves several processes, including: collecting sufficient relevant
information, assessing the latest changes, determining unsuitable or lacking policies, proposing
suitable updates and going through various debating, negotiation and approval processes, before
implementing actual amendments or extensions.
AMPM’21: First Workshop in Agent-based Modelling & Policy-Making, December 8, 2021, Vilnius, Lithuania
Envelope-Open ada.diaconescu@telecom-paris.fr (A. Diaconescu); dominique.blouin@telecom-paris.fr (D. Blouin);
alice.ludvig@boku.ac.at (A. Ludvig)
GLOBE https://adadiaconescu.there-you-are.com/ (A. Diaconescu); https://www.telecom-paris.fr/dominique-blouin
(D. Blouin); https://boku.ac.at/personen/person/7B8ED36D42BDAF0C (A. Ludvig)
Orcid 0000-0002-2279-0846 (A. Diaconescu); 0000-0001-7606-0251 (D. Blouin); 0000-0003-1323-2276 (A. Ludvig)
© 2021 Copyright for this paper by its authors. Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0).
CEUR
Workshop
Proceedings
http://ceur-ws.org
ISSN 1613-0073
CEUR Workshop Proceedings (CEUR-WS.org)
As Socio-Technical Systems (STSs) are becoming increasingly complex, so are the policy-
making processes that aim to regulate them. This, in turn, leads to situations where policy-
makers are overloaded and unable to keep-up with latest developments. Notable difficulties
stem from the lack of resources to sustain long-term policies, the preference towards short-term
planning within the electoral process, and, more essential with respect to this work, the lack of
standardised frameworks and tools for monitoring and assessing policy impacts [1].
Modern Cyber-Physical Systems (CPSs) experience similar challenges, chiefly related to the
complexity of their adaptation management processes in a context of rapid and unpredictable
changes. Research solutions have been developed to deal with such challenges – spanning
from initial data collection, system modelling and analysis (e.g. data-mediation frameworks [2],
digital twins [3], Multi-Paradigm Modelling (MPM) [4]) and all the way to runtime decisions
and system adaptations (e.g. self-aware [5] and self-managing systems [6], decentralised and
multi-scale controllers [7], models at runtime [8], self-integrating systems [9]).
In this vision paper, based on our expertise with some of the above-mentioned topics, and
on our experience on policy-making for STSs of the forestry management domain [10, 11, 12],
we survey existing systems engineering and self-adaptation solutions for complex CPSs that
can potentially be adapted to better develop policies for modern STSs. Our objective is not to
propose a new policy making process, but rather to allow improving existing ones using multi-
paradigm modelling to support their specification and application, inspired from successful
solutions issued from systems engineering and self-adaptive goal-oriented, multi-scale control
CPSs, which exhibit characteristics similar to STSs. As a survey of approaches, we do not
present any specific case studies but focus on similarities between the domains of STSs CPSs to
establish an initial research plan. Developing specific case studies will be part of future work.
We start by introducing minimal notions on STSs and existing approaches for policy making
in section 2. We then introduce CPSs and associated successful engineering methods in section
3, including the main approaches on which our framework will build (i.e. Goal-Oriented
Requirements Engineering (GORE), Design-Space Exploration (DSE) and Goal-Oriented Multi-
Scale Control Systems (GOMSS). In section 4, we introduce enablers for theses approaches (i.e.
Model-Based Engineering (MBE) and its Multi-Paradigm Modelling (MPM) extension). We then
present the blueprint of our approach based on these preliminary notions in section 5 – this
includes the existing modelling languages and tools it may reuse, as well as the languages and
tools that we aim to develop for policy making activities and domain-specif modelling in the
planned case studies. We conclude the paper in section 7.
2. Policy Making for Socio-Technical Systems
STSs involve wide-spread interactions amongst people and technologies, hence integrating
complex social and technical infrastructures with human behaviour. The term STS and its
theory was coined during World War II to characterise systems such as coal mines in England
[13]. The purpose of STS theory was to study how these systems’ management processes could
be improved so as to increase overall system performance and quality in people’s work lives.
STSs management involves governance, which relies on a set of policies developed by policy-
makers. Policy-making is a complex process that involves many stakeholders, often pursuing
different or incompatible interests. It may also span over long periods during which stakeholder
interests may vary. As policies are strongly context-depend, they should consider, e.g., national,
economic, political, cultural and social structures.
For instance, [14] presents an interesting overview of the conceptual and methodological
aspects of policy-making for the European Training Foundation (ETF). It introduces basic
notions related to policy-making and its different approaches developed over time. We will
briefly present some of these notions so that we can later-on emphasise on their similarities
with systems engineering methods to illustrate how our proposed framework can build upon
these to better support policy making with adequate modelling languages and tools.
Policy-making processes build upon a set of policy analysis activities, which mainly investigate
alternative policy options by gathering and integrating the advantages and inconveniences
of each option. It is a problem-solving activity that attempts to predict the consequences of
alternative courses of action. Numerous perspectives and frameworks exist for policy analysis.
The ETF outlines three main approaches – namely, the analycentric, the policy process and the
meta-policy – each dealing with problems at three different scales. The analycentric approach
typically focuses on individual, technical problems at the micro-scale and aims to identify the
most effective and efficient solution in technical and economic terms (e.g. the most efficient
resource allocation). The policy process approach usually focuses on political problems (e.g.
involved stakeholders) at the meso-scale, aiming to determine employed processes and means
and to explain the stakeholders’ role and influences. Problem solutions are identified by changing
the relative power and influence of certain groups (e.g., enhancing public participation and
consultation). Finally, the meta-policy approach focuses on the structural problems of the system
and its context, at the macro-scale (e.g. an economic system or political institution). It aims to
explain the contextual factors of the policy process– i.e., what are the political, economic and
socio-cultural factors influencing it.
Such policy analyses constitute the backbone activities of overall policy-making processes,
aiming to specify and validate, at least to a minimal level, the fact that a set of policies fits
the problem that must be solved within a targeted STS. Within this overall process, the ETF
proposes: first, to use elements of the policy process and meta-policy approaches to set policy
priorities; and secondly, to employ the analycentric approach when formulating the actual
policy option.
Other policy-making support solutions are also discussed in [14], such as the policy cycle
framework and the policy network perspective. The policy cycle aims to split complex policy-
making processes into manageable steps, breaking them down into sequential stages, examining
what happens within each individual stage, and assuming that each stage influences the fol-
lowing one. The policy network offers a different way to tackle policy-making complexities. It
concentrates on the (meta-)policy process and the relations amongst the actors (i.e. the network)
who participate in it, seeking to explain policy outcomes in relation to these characteristics.
The policy-making process proposed by the ETF combines the aforementioned policy cycle
with the policy networks. Here, policy networks are approached as a model of collective decision
making– an exchange process between actors operating within a market to gain control and
influence over resources. This leads to the overall-policy making process depicted in Fig. 1.
Fig. 2 shows the detailed steps decomposing the Policy Formulation stage of Fig. 1.
One key aspect as stressed by the ETF is that there is no single or best way to conduct policy
Figure 1: The public policy process proposed by the ETF [14].
Figure 2: Detailed steps for the Policy Formulation stage of Fig. 1 [14].
analyses, due to the multi-faceted nature of policy analyses. Such is also the case for systems
engineering where each organisation will typically need to customise standard development
processes in order to satisfy organisation and project specific needs. Hence, a policy making
framework must allow to seamlessly customise processes and tools by allowing to define and
integrate new activities and workflows as needed by the specific policy-making organisations,
projects and STSs under consideration.
Another interesting aspect mentioned by the ETF regarding policy analysis is the importance
of problem identification, as many policy-making failures are due to solving the wrong problem
(rather than to proposing wrong solutions to the right problem). This means that significant
effort should go into the formulation of the policy problem. This difficulty is also largely
observed in systems engineering, for which several requirements engineering approaches have
been developed, which as we will see later could be beneficial when applied to policy making.
Finally, the ETF note that under these circumstances, there is an increasing need to profes-
sionalise policy making and its activities to ensure effective governance of these processes and
the capacity to anticipate problems. This will help avoiding traps such as policy overload, which
happens when governments develop policy plans that are too complex or too vague, containing
too many priorities. This results in focus less fragmented priorities leading to endless stream of
ad-hoc initiatives. This was illustrated repeatedly during the COVID19 crisis, e.g., in France.
The lesson here is that policy plans must be actionable and clear, so as to ensure wide-spread
adoption by concerned stakeholders. Therefore, better tool support to policy making is required
to help avoid falling into these traps.
3. Engineering Cyber-Physical Systems
Compared to STSs, Cyber-Physical Systems (CPSs) are also complex systems, yet generally
excluding the more unpredictable human dimension. The CPS term emerged around 2006 at the
National Science Foundation in the United States [15] to identify systems that integrate multi-
physical processes (e.g. mechanical, electrical, biochemical) and computational processes (e.g.
control, signal-processing, logical inference, planning); and that typically run within uncertain
environments. These systems are part of many of our daily activities and drive innovation in
important domains (e.g. Automotive, Avionics, Civil Engineering, Industry 4.0, Robotics, smart
systems).
While CPS are generally engineered to be predictable – especially for safety-critical systems,
e.g. air planes – including humans in the loop brings about new dimensions of complexity
and unpredictability. Indeed, the latest trends in CPS development research is to better include
human factors – considering a system’s human actors and their socioeconomic context [16].
This brings about the new challenge of dealing with highly unpredictable entities. Therefore,
CPSs indirectly inherit several characteristics of STSs.
Similar to complex STSs, for which policy-making processes have been developed, well-
known engineering processes have been proposed for different kinds of engineered systems;
and for CPSs in particular. The most well-known is the Systems Engineering process promoted
by the INCOSE (International Council on Systems Engineering) 1 .
Systems Engineering is a trans-disciplinary integrative approach covering the entire system
life-cycle– starting from its development, its operation, maintenance and all the way to its
disposal. It consists of establishing stakeholders’ goals and required system functionalities,
including an appropriate system life-cycle model, process approach and governance structures;
while taking into account the system’s complexity, uncertainty, change and variety.
A critical activity in the systems engineering process – often called Design Space Exploration
(DSE) – is to consider alternative system designs and configurations, so as to find the best
design for the given requirements. Based on these results, system synthesis, verification and
validation from the system design can be performed. Fig. 3 illustrates the overall systems
engineering process. We can easily map several of its stages to the ones of the policy making
process of Fig. 1. For instance, each process starts with a stage including strong emphasis on
problem identification, as explicitly stated for policy making. This also takes place for system
engineering during the customer needs analysis stage. The Policy Formulation stage on the
policy making side, which consists of formulating adequate policies to solve the identified
1
INCOSE: https://www.incose.org/
Figure 3: The systems engineering process as proposed by the INCOSE.
problem, corresponds to System Design and Development on the systems engineering side,
where engineers Specify a solution to the identified customer problem. We note, from Fig. 2, that
each of these solution building stages include alternative solutions evaluation steps in order to
come up with the best solution to the specified problems. Policy Adoption and Implementation
stages correspond to the systems engineering Tests & Validation and Operation stages, and the
Policy Evaluation stage to the Review / Transition stage of systems engineering. Finally, we
note that each process circularly comes back to its initial problem identification stage with the
whole process being applied iteratively for both policy making and systems engineering cases.
3.1. Goal Oriented Requirements Engineering (GORE)
As we have seen previously, similar to policy-making, systems engineering also emphasises
the problem and solution domains. The problem phase has been given high attention by
the Requirements Engineering (RE) community. The notions of goals, or requirements, were
developed to specify the problem with the solution expressed via system architecture. However,
it is difficult to set a precise boundary between the problem and solution parts. Developing a
solution is iterative, and design choices often implying new more detailed requirements to be
added to the problem part, and so on.
Nevertheless, experience showed that most problems are introduced during the RE stage.
This triggered the development of Goal Oriented Requirements Engineering (GORE) approaches
[17, 18, 19, 20]. For instance, the KAOS approach defines four complementary and interrelated
views on the system and its environment:
• Goals representing owners, users, business managers, regulations, etc., which are analysed
taking into account conflicts and their resolution.
• Responsible agents, both human and machine from the system and its environment, which
captures system structure in terms of subsystems and their interactions.
• The specific problem domain represented by concepts and their relationships.
• Agent behaviours that agents must exhibit in order to achieve goals as well as possible.
The combination of these different views on the system allowed developing several activi-
ties to better analyse and validate requirements and significantly improve system design and
development.
3.2. Goal-Oriented Multi-Scale Control Systems (GOMSS)
While GORE traditionally focuses on the system’s offline specification and development, the
high unpredictability of execution environments requires to push these processes into the
run-time (i.e. during system execution). This is the case for most complex STSs, e.g. smart
homes and cities, power grids, autonomous vehicle networks, robotic swarms and integrated
Industry 4.0 systems-of-systems.
To deal with runtime changes, while requiring minimal human intervention and ideally no
service disruption, such systems must adapt themselves to dynamic changes in their internal
resources, execution environment and even targeted goals. Several relatively recent research
areas have been tackling this challenge, notably including Autonomic Computing (AC) [6],
Organic Computing (OC) [21], Self-Adaptive Systems (SAS) and Self-aware Computing (SeAC)
[5].
To enable self-adaptation, systems generally feature internal control feedback loops – for
system and context monitoring; problem detection and analysis; solution planning; and execution.
Such feedback loops rely on various types of knowledge (e.g. system, environment and goal
models), which can be updated and extended at runtime (e.g. learning).
In addition to specifying the design and algorithms of such control feedbacks (i.e. defining how
the system should self-adapt), more recent solutions have also considered explicitly specifying
the goals of such control feedbacks (i.e. what the self-adaptive system should achieve). Hence,
goal-oriented self-adaptive systems go beyond the dynamic selection of predefined adaptive
behaviours, allowing the system to learn and discover new adaptation behaviours, as suited for
dealing with problematic situations that couldn’t be predicted at design-time. E.g. [22] propose
to enable systems to self-integrate from an extensible set of control components, which can be
discovered at runtime, so as to achieve explicitly-specified goals when new problems occur.
Considering the very nature of most complex systems (i.e. typically composed of a large
number of interconnected entities), the scalability of self-adaptation solutions is always a major
issue. Various forms of hierarchical approaches have been proposed to handle this issue, either
via generic architectures (e.g. [23]) or domain-specific designs (e.g. traffic control [24] and
manufacturing [25]). Drawing inspiration from both natural and artificial domains, the Multi-
Scale Feedbacks (MSF) design pattern was proposed to offer a generic, reusable solution to
control scalability problems across various contexts [7]. Hence, goal-oriented, multi-scale self-
adaptation solutions (e.g. [26, 22]) combine goal-orientation – for enabling new self-adaptation
behaviours to be identified during runtime; with multi-scale designs – for ensuring the scalability
of self-adaptation processes.
4. Systems Engineering Enablers
Traditional systems engineering processes such as the one mentioned above used to be mostly
supported by natural language documents or at best structured documents such as spreadsheets.
Given the increasing complexity of systems, this turned out to be unmanageable to build
today’s systems at affordable costs. This triggered the development of the Model-Based System
Engineering paradigm, where natural language documents are replaced by models of Domain-
Specific Modelling Languages (DSML).
Models have been used for a long time to help us understand and analyse complex systems,
processes, or artefacts that we are studying, interacting with, managing or developing (e.g.
geographical maps, construction blueprints). They are abstractions of reality built for particular
purposes [27, 28]. Representing a simplified reality, a model is easier to process than the real
thing. However, the way in which a model abstracts reality must be carefully chosen according
to the model’s purpose. For example, a road map will abstract the land as a set of two dimensional
lines indicating road paths so that people can find their way to go from one place to another.
To be usable, models must be understandable in some way. Hence, models must be expressed
in some modelling language (e.g. the map’s legend) that must be known to understand the
models. Such modelling languages can be specified using dedicated metalanguages such as
meta-models or grammars. Being expressed in terms of formally specified modelling languages,
models can be analysed with tools to automatically detect errors in the system early, which is
not possible when only natural language documents are used.
Figure 4: Goal-oriented multi-scale architecture for socio-technical systems
Applied to developing Cyber-Physical Systems (CPS) such Model-Based Engineering paradigm
was proven effective leading to better system quality and important cost savings [29]. Besides
providing model analysis capabilities, modelling languages also provide a common understand-
ing of the domain intended to be covered by the language. This is achieved through a set of
precise concepts and their relationships that formally specify the language and can be used to
constitute a common vocabulary for domain actors. It has been shown that expressing some-
thing in terms of a properly constructed Domain-Specific Modelling Language (DSML) greatly
helps in reducing risks of imprecision, miss-communication and inconsistencies, compared to
using natural language. As pointed out in [30], by definition a model is always precise, since its
vocabulary is formally defined through a modelling language, being at least precise enough to be
understood by computer programs independently of the level of abstraction of the represented
reality.
An issue when using such DSMLs is that while expressiveness increases with the specificity
of the language, what can be expressed using the language is limited to the covered domain
(scope). Therefore, several DSMLs must often be used conjointly to cover all required domains.
This observation leads to Multi-Paradigm Modelling (MPM) [31, 4, 32, 33, 34], which advocates
the use of a set of modelling languages, each one being most appropriate for the particular
subset of activities to be performed with models of the language, rather than trying to build a
single monolithic language to support all activities. In particular, Multi-Paradigm Modelling
builds upon the following principles:
• Model each part and aspect of a system explicitly, so as to capture all relevant information
in terms of formal languages that can be understood and processed by computers;
• Model at the appropriate level(s) of abstraction (or scales) using the most appropriate
formalism to avoid introducing accidental complexity.
While the main asset of MPM is to tackle the multi and heterogeneous aspects of systems,
it comes at the expense of needing to properly combine them. This is a research topic in its
own — model management — where the set of employed models, their modelling languages
and the relationships between the models are also formally specified with dedicated DSMLs [33,
35, 36]. Thanks to MPM, language extensibility is supported by allowing the integration of new
languages configured for project or organisation-specific needs.
5. Our Approach
Inspired by model-based approaches and paradigms that were successful for Cyber-Physical
Systems (as presented earlier) and considering the previously identified commonalities between
system engineering and policy making processes, and the commonalities between goal-oriented
multi-scale control systems and STSs, we propose to develop a conceptual framework that
better supports existing and new policy-making processes specification and application. Such
conceptual framework will rely on a set of combined DSMLs and associated tools, to help specify,
analyse and understand policy-driven STSs. This will help to detect missing, misapplied or
outdated policies; as well as simulate and predict specific effects of various policy updates. This
approach will be based on MPM– to benefit from its modelling assets and model integration
capabilities.
5.1. Multi-Paradigm Modelling for Socio-Technical Systems
Following MPM principles, we will reuse, adapt, develop and combine the most appropriate
modelling languages to cover all aspects of STSs including their policy making processes,
as relevant for the modelling activities to be performed at the appropriate abstraction levels.
This will provide a precise, shared vocabulary for the policy-making domain, which is already
an improvement (compared to using only natural language documents) as it helps to avoid
misunderstandings among project stakeholders. For instance, looking at the policy making
approach specified in the ETF document [14], we believe that expressing the presented notions
and their relationships formally with an appropriate modelling language (ontology) would
enforce making the underlying knowledge more explicit and help developing tools to better
support the ETF processes.
Reusing or adapting existing modelling languages instead of reinventing new ones can be
very beneficial to policy making by taking advantage of knowledge gained in other domains
that experienced similar problems. In particular, for the problem identification and policy
formulation phases of the policy making process of Fig. 1, policy makers can benefit from the
extensive work of the RE community and in particular its aforementioned GORE approaches.
For this, we intend to reuse as much as possible existing GORE languages and tools such as
the User Requirements Notation (URN)2 , which incorporates a large part of the i* language
and approach [17] to capture and analyse goals, stakeholders, their conflicts, resources and so
on. Besides, we will benefit from its use case maps sub-language, which allows specification
of use case scenario at a high level of abstraction for early policy making phases, and from its
goal-oriented decision making methodology based on Key Performance Indicators (KPI) [37].
Thanks to its jUCMNav tool3 supporting these features, several analyses to evaluate KPIs can
easily be developed and evaluated from goal models. In addition, use case scenarios can be
simulated with the built-in simulator to detect inconsistency in system behaviour early, thanks
to a simple action language. Several benefits of using this tool early have been shown in [30],
among others.
Furthermore, existing successful solutions for adaptation management in complex CPSs
will be extended and customised to bring some of the necessary support to modern policy-
driven STSs. A starting point will be to propose a DSML to represent, analyse and understand
policy-driven adaptations for the specific STS domain (e.g. forestry management, co-design of
shared living spaces). Coupled with the aforementioned goal models using MPM, the language
will support modelling goal-oriented, multi-scale architectures for controlling complex STSs
(e.g. smart homes and power grids [26, 38]), Fig. 4. Here, goals are first-class system entities,
representing anything that a system is willing to achieve or enforce – e.g. objectives, values,
constraints, rules, policies, norms, priorities, actions. Each goal is defined by: i) a specific
evaluation function, defining how to assess goal achievement; and ii) a spatio-temporal scope,
defining where and when to achieve the goal.
2
URN: https://www.itu.int/rec/T-REC-Z.151/en/
3
jUCMNav: http://softwareengineering.ca/jucmnav/
A system may pursue several goals simultaneously. Goals may be in conflict if they pursue
incompatible objectives over intersecting spatio-temporal scopes. A continuous control feedback
loop aims to achieve each system’s goal, based on internal resources and the external context. To
reach a system goal, such control loop may aim to achieve ‘lower-level’ goals first, and so forth,
recursively. This leads to a multi-scale organisation of control loops, which run simultaneously
and coordinate to achieve goals and manage conflicts.
Modelling complex systems in this manner helps preliminary analysis at design time – by
rendering goals, system controllers and relevant context aspects explicit and allowing to detect
conflicts and/or tune controllers accordingly. It also becomes essential for system adaptation at
run-time, when dynamic changes occur – e.g. detecting new conflicts and adapting controllers
to resolve them, when new goals or resources are added-on. The goal-oriented aspect contrasts
traditional control approaches based on predefined rules and static system models, which could
not adapt to unforeseen changes. The multi-scale design also helps manage system complexity,
by abstracting system concepts at different scales and hiding irrelevant details from the scales
below – hence applying a divide-and-conquer technique that limits the amount of resources
necessary at each scale [39].
We believe that applying these concepts and modelling technique to policy-driven STSs will
help formalise policy-management processes and render them more traceable and rational. It
will help, for instance: to identify all system goals (objectives, policies, values, constraints, etc);
to analyse and identify potential conflicts; to detect missing, misapplied or ineffective policies;
and, to simulate some of the effects of various policy changes. Providing a multi-scale design
also matches the current structure of most political institutions, where the aforementioned
processes may run simultaneously, while dealing with cross-scale conflicts and synchronisation
issues. Existing knowledge from cross-domain multi-scale control systems may also be useful
here, to better understand and manage such multi-scale policy-making processes.
In order to better guide users in following adopted policy-making processes such as the
one depicted by Figs. 1 and 2, our framework will also allow modelling these processes using
standard workflow models and tools, adapted to the needed modelling activities, similar to
what is proposed in [40, 34]. By guiding stakeholders in performing their modelling and
analysis activities thanks to these workflows, we can propose more advanced automatic tools for
problem detection, solution proposals and impact predictions. For instance, more sophisticated
simulations may be proposed via agent-based modelling (ABM) relying on existing platforms
– e.g., Repast4 (social sciences), GAMA5 (spatially-aware agents), SARL6 (holonic agents);
NetLogo7 (general-purpose entry-level ABM platform), JADE8 . In particular, if other simulating
tools are found beneficial, we will use MPM co-simulation approaches [41] to coordinate existing
simulators and improve simulation by integrating different paradigms.
4
Repast: https://repast.github.io
5
GAMA: https://github.com/gama-platform
6
SARL: http://www.sarl.io
7
NetLogo: https://ccl.northwestern.edu/netlogo
8
JADE: https://jade-project.gitlab.io/
5.2. Assumptions, Threats and Limitations
While promising, and despite the observed similarities between complex CPSs (including
human actors) and STSs, our approach relies on several assumptions that can only be verified
by pursuing the proposed research direction, developing associated tools and applying them to
policy making in various real-world STSs. We discuss these assumptions and how they can be
mitigated below.
The main question is whether what worked for CPS can also work, at least partially, for
STSs. Regarding DSMLs, developing formal ontologies for certain domains has proven rather
tedious a process, due to the difficulty in converging on a single set of commonly-accepted
domain concepts. For instance, for the CPS domain, modelling languages such as AADL9
have been developed for more than 15 years. However, mitigating this threat, we note that
this situation did not prevent approaches and tools from being developed early-on during the
language specification, allowing benefits to be obtained right from the incipient phases. This
is also the case for programming languages. For example, discussing the appropriateness of
one language over another often leads to long debates without resolution. What matters is that
DSMLs, even if imperfect, can already bring benefits.
Besides, while DSML development is a complicated, iterative process, approaches such as [42]
exist to take into account users and usage context at the very beginning of DSML development.
This will help ensure that the targeted stakeholders, such as policy makers (often with social
science backgrounds), will be able to use modelling processes and tools despite their differences
from typical model and tools users with engineering backgrounds.
We must also consider the fact that the proposed modelling and analysis processes would be
carried-out by human actors (rather than automatic agents) and within open socio-technical
contexts (rather than constrained environments that can be completely formalised and con-
trolled). Hence, we aim to make sure that the provided modelling platform offers sufficient
flexibility and expressiveness to suit complex unforeseen socio-technical situations; and avoid
over-constraining and pre-formatting all possible system representations. Indeed, our intention
is to provide a helpful tool, rather than a limiting control framework. Thus, we will ensure that
actors can easily extend or update the underlying languages and tools to express exceptional
cases and diverse view-points.
Another important aspect is that this proposal can only provide support for achieving well-
defined goals, in conformity with well-established societal values. While it can allow actors to
change and evolve their goals over time, it provides no magic solution against conflicting goals,
misaligned values or lack of political will for systemic change.
Finally, we hope that substantial gain from modelling can be achieved by providing a common
vocabulary, facilitating understanding and encouraging better precision in policy specifications;
and by providing better support for policy analyses, including simulation. Such benefits must be
sufficiently significant to overcome the overheads involved in learning new tools and modelling
processes.
9
AADL: http://www.aadl.info/
6. Related Work on Policy-Making Tools for Socio-Technical
Systems
To our knowledge, there has been limited efforts to provide policy-making support tools. For
instance, we note the GRACeFUL project (Global systems Rapid Assessment tools through
Constraint Functional Languages)10 , which aims to develop smart tools that guide decision
makers during the implementation of urban projects. The main objective of this project was
to develop a simulation platform for modelling the possible responses of citizens and other
stakeholders involved, so that the tool can generate predictions from multiple view-points about
various policy implementations. Similar to our approach, one concern of the project was to
make this kind of modelling accessible to nonspecialists. This was achieved apparently via a
graphic platform that facilitated the development of custom software tools.
We plan to achieve similar advantages with our approach, while adopting MPM at the heart of
our language engineering, tool building and model integration processes. Still, while the focus of
GRACeFUL was on functional languages, where policy making is seen as a constraints-solving
problem, our contribution is considerably wider. It aims to develop not only the decision making
support tools but also to establish the necessary foundations and conceptual framework for the
policy-making domain; and enable these to be customised and extended for various application
domains and project requirements. Moreover, we believe that the policy-making domain will
be better addressed by a combination of paradigms rather than by a single one (constraints
solving).
Regarding process modelling, [43] investigated the feasibility of using workflow tools to
support policy-making processes. It presents lessons learned from cases in four European
countries and concludes that the processes show several commonalities (common tasks and
sequence of steps) despite that they belong to different policy domains. The authors also identify
a lack of any supporting information technology, which prevents transparency in policy making
processes. These are shortcomings that our proposal aims to address.
7. Conclusion and Future Work
This position paper identified the need for more formal support in policy-making processes, to
help policy-makers deal with inherent complexities: identifying and considering the multiple
stakeholders, objectives and viewpoints relevant to the specified policies; taking into account
the diverse facets to be managed in the targeted application domain (via the specified policies);
understanding the potential effects of policy updates and interpreting the observable effects of
existing policies upon their application domains. Drawing inspiration from successful solutions
to similar problems in technical systems (e.g. CPS), we propose to develop a conceptual modelling
framework and support tools to help such policy-making processes. We rely on our previous
experience with: i) domain-specific modelling languages and multi-paradigm modelling; and ii)
goal-oriented multi-scale architectures for complex self-adaptive systems. We aim to extend
and adapt relevant solutions available from these areas to address the particular challenges
of the policy-making domain, within the socio-technical systems context. We believe that
10
GRACeFUL: http://www.fetfx.eu/project/graceful/
such modelling framework, even if necessarily imperfect and continuously evolving, would
provide essential support for improving the efficacy, efficiency, comprehension and traceability
of policy-making processes.
References
[1] J. M. Barrutia, C. Echebarria, M. R. Paredes, P. Hartmann, V. Apaolaza, From rio to rio+20:
twenty years of participatory, long term oriented and monitored local planning?, Journal
of Cleaner Production 106 (2015) 594–607.
[2] P. Lalanda, C. Hamon, An autonomic mediation framework for complex physical
environments, in: 2014 IEEE International Congress on Big Data, Anchorage, AK,
USA, June 27 - July 2, 2014, IEEE Computer Society, 2014, pp. 506–513. URL: https:
//doi.org/10.1109/BigData.Congress.2014.80. doi:10.1109/BigData.Congress.2014.80 .
[3] I. Errandonea, S. Beltrán, S. Arrizabalaga, Digital twin for maintenance: A literature
review, Comput. Ind. 123 (2020) 103316. URL: https://doi.org/10.1016/j.compind.2020.
103316. doi:10.1016/j.compind.2020.103316 .
[4] M. Amrani, D. Blouin, R. Heinrich, A. Rensink, H. Vangheluwe, A. Wortmann, Multi-
paradigm modelling for cyber–physical systems: a descriptive framework, Software and
Systems Modeling 20 (2021). doi:10.1007/s10270- 021- 00876- z .
[5] S. Kounev, P. R. Lewis, K. L. Bellman, N. Bencomo, J. Cámara, A. Diaconescu, L. Esterle,
K. Geihs, H. Giese, S. Götz, P. Inverardi, J. O. Kephart, A. Zisman, The notion of self-
aware computing, in: S. Kounev, J. O. Kephart, A. Milenkoski, X. Zhu (Eds.), Self-Aware
Computing Systems, Springer International Publishing, 2017, pp. 3–16.
[6] P. Lalanda, J. A. McCann, A. Diaconescu, Autonomic Computing - Principles, Design
and Implementation, Undergraduate Topics in Computer Science, Springer, 2013. URL:
https://doi.org/10.1007/978-1-4471-5007-7. doi:10.1007/978- 1- 4471- 5007- 7 .
[7] A. Diaconescu, L. J. D. Felice, P. Mellodge, Exogenous coordination in multi-scale systems:
How information flows and timing affect system properties, Future Gener. Comput.
Syst. 114 (2021) 403–426. URL: https://doi.org/10.1016/j.future.2020.07.034. doi:10.1016/j.
future.2020.07.034 .
[8] R. B. France, B. Rumpe, Model-driven development of complex software: A research
roadmap, CoRR abs/1409.6620 (2014).
[9] K. L. B. et al., Self-improving system integration: Mastering continuous change, Future
Gener. Comput. Syst. 117 (2021) 29–46. URL: https://doi.org/10.1016/j.future.2020.11.019.
doi:10.1016/j.future.2020.11.019 .
[10] J. Wilkes-Allemann, A. Ludvig, S. Gobs, E. Lieberherr, K. Hogl, A. Selter, Getting
a grip on negotiation processes: Addressing trade-offs in mountain biking in aus-
tria, germany and switzerland, Forest Policy and Economics 136 (2022) 102683. URL:
https://www.sciencedirect.com/science/article/pii/S1389934121002896. doi:https://doi.
org/10.1016/j.forpol.2021.102683 .
[11] G. Weiss, E. Hansen, A. Ludvig, E. Nybakk, A. Toppinen, Innovation governance in the
forest sector: Reviewing concepts, trends and gaps, Forest Policy and Economics 130
(2021) 102506. URL: https://www.sciencedirect.com/science/article/pii/S138993412100112X.
doi:https://doi.org/10.1016/j.forpol.2021.102506 .
[12] A. Ludvig, M. Braun, F. Hesser, L. Ranacher, D. Fritz, T. Gschwantner, R. Jandl, G. Kinder-
mann, T. Ledermann, W. Pölz, K. Schadauer, B. F. Schmid, C. Schmid, P. Schwarzbauer,
G. Weiss, B. Wolfslehner, P. Weiss, Comparing policy options for carbon efficiency
in the wood value chain: Evidence from austria, Journal of Cleaner Production 292
(2021) 125985. URL: https://www.sciencedirect.com/science/article/pii/S0959652621002055.
doi:https://doi.org/10.1016/j.jclepro.2021.125985 .
[13] S. Long, Socioanalytic Methods: Discovering the Hidden in Organisations and Social Sys-
tems, Karnac Books Limited, 2013. URL: https://books.google.fr/books?id=LWL7krX7jgAC.
[14] Madlen Serban, Supporting Public Policy Making through Policy Analysis and Policy
Learning , Technical Report, European Training Foundation, 2015. doi:10.2816/721877 .
[15] E. Lee, The past, present and future of cyber-physical systems: A focus on models, Sensors
15 (2015) 4837–4869. URL: http://dx.doi.org/10.3390/s150304837. doi:10.3390/s150304837 .
[16] D. Nunes, J. S. Silva, F. Boavida, A Practical Introduction to Human-in-the-Loop
Cyber-Physical Systems, Wiley, 2018. URL: https://www.wiley.com/en-us/A+Practical+
Introduction+to+Human+in+the+Loop+Cyber+Physical+Systems-p-9781119377788.
[17] E. Yu, N. Giorgini, Paola anb Maiden, J. Mylopoulos, Social Modeling for Requirements
Engineering, MIT Press, 2011.
[18] A. Dardenne, A. van Lamsweerde, S. Fickas, Goal-directed requirements acquisition,
Science of Computer Programming 20 (1993) 3–50. URL: https://www.sciencedirect.com/
science/article/pii/016764239390021G. doi:https://doi.org/10.1016/0167- 6423(93)
90021- G .
[19] A. I. Antón, C. Potts, The use of goals to surface requirements for evolving systems, in:
Proceedings of the 20th International Conference on Software Engineering, ICSE ’98, IEEE
Computer Society, USA, 1998, p. 157–166.
[20] J. Horkoff, F. Aydemir, E. Cardoso, T. Li, A. Maté, E. Paja, M. Salnitri, L. Piras, J. Mylopoulos,
P. Giorgini, Goal-oriented requirements engineering: an extended systematic mapping
study, Requirements Engineering 24 (2019). doi:10.1007/s00766- 017- 0280- z .
[21] R. P. Würtz (Ed.), Organic Computing, Understanding Complex Systems, Springer, 2008.
[22] A. Diaconescu, S. Frey, C. Müller-Schloer, J. Pitt, S. Tomforde, Goal-oriented holonics
for complex system (self-)integration: Concepts and case studies, in: G. Cabri, G. Picard,
N. Suri (Eds.), 10th IEEE International Conference on Self-Adaptive and Self-Organizing
Systems, SASO 2016, Augsburg, Germany, September 12-16, 2016, IEEE Computer Society,
2016, pp. 100–109.
[23] H. A. Simon, The architecture of complexity, in: Facets of systems science, Springer, 1991,
pp. 457–476.
[24] K. Fischer, Holonic multiagent systems – theory and applications, in: P. Barahona,
J. J. Alferes (Eds.), Progress in Artificial Intelligence, Springer Berlin Heidelberg, Berlin,
Heidelberg, 1999, pp. 34–48.
[25] M. Ulieru, R. A. Este, The holonic enterprise and theory emergence: On emergent features
of self-organization in distributed virtual agents, Cybern. Hum. Knowing 11 (2004) 79–98.
[26] S. Frey, A. Diaconescu, D. Menga, I. M. Demeure, A generic holonic control architecture for
heterogeneous multiscale and multiobjective smart microgrids, ACM Trans. Auton. Adapt.
Syst. 10 (2015) 9:1–9:21. URL: https://doi.org/10.1145/2700326. doi:10.1145/2700326 .
[27] J. Rothenberg, The Nature of Modeling, John Wiley & Sons, Inc., USA, 1989, p. 75–92.
[28] B. Combemale, R. France, J.-M. Jezequel, B. Rumpe, J. Steel, D. Vojtisek, Engineering
Modeling Languages - Turning Domain Knowledge into Tools, Chapman and Hall/CRC,
2017.
[29] J. Hansson, S. Helton, P. Feiler, ROI Analysis of the System Architecture Virtual Inte-
gration Initiative, Technical Report CMU/SEI-2018-TR-002, Software Engineering Insti-
tute, Carnegie Mellon University, Pittsburgh, PA, 2018. URL: http://resources.sei.cmu.edu/
library/asset-view.cfm?AssetID=517157.
[30] D. Blouin, H. Giese, Combining requirements, use case maps and aadl models for safety-
critical systems design, in: 2016 42th Euromicro Conference on Software Engineering and
Advanced Applications (SEAA), 2016, pp. 266–274. doi:10.1109/SEAA.2016.15 .
[31] H. Vangheluwe, J. de Lara, P. Mosterman, An introduction to multiparadigm modelling
and simulation, in: Proceedings of the AIS’2002 Conference 9–20 (2002). 35, Press, 2002.
[32] D. Blouin, R. Al-Ali, M. Iacono, B. Tekinerdogan, H. Giese, Chapter 2 - an ontological
foundation for multi-paradigm modelling for cyber-physical systems, in: B. Tekinerdogan,
D. Blouin, H. Vangheluwe, M. Goulão, P. Carreira, V. Amaral (Eds.), Multi-Paradigm
Modelling Approaches for Cyber-Physical Systems, Academic Press, 2021, pp. 9–43. URL:
https://www.sciencedirect.com/science/article/pii/B9780128191057000076. doi:https://
doi.org/10.1016/B978- 0- 12- 819105- 7.00007- 6 .
[33] H. Giese, D. Blouin, R. Al-Ali, H. Mkaouar, S. Bandyopadhyay, M. Iacono, M. Amrani,
S. Klikovits, F. Erata, Chapter 4 - an ontology for multi-paradigm modelling, in: B. Tekin-
erdogan, D. Blouin, H. Vangheluwe, M. Goulão, P. Carreira, V. Amaral (Eds.), Multi-
Paradigm Modelling Approaches for Cyber-Physical Systems, Academic Press, 2021, pp.
67–122. URL: https://www.sciencedirect.com/science/article/pii/B978012819105700009X.
doi:https://doi.org/10.1016/B978- 0- 12- 819105- 7.00009- X .
[34] D. Blouin, R. Al-Ali, H. Giese, S. Klikovits, S. Bandyopadhyay, A. Barišić,
F. Erata, Chapter 5 - an integrated ontology for multi-paradigm modelling
for cyber-physical systems, in: B. Tekinerdogan, D. Blouin, H. Vangheluwe,
M. Goulão, P. Carreira, V. Amaral (Eds.), Multi-Paradigm Modelling Approaches
for Cyber-Physical Systems, Academic Press, 2021, pp. 123–145. URL: https://www.
sciencedirect.com/science/article/pii/B9780128191057000106. doi:https://doi.org/10.
1016/B978- 0- 12- 819105- 7.00010- 6 .
[35] D. Blouin, Y. Eustache, J.-P. Diguet, Extensible global model management with meta-model
subsets and model synchronization, in: GEMOC@MoDELS, 2014.
[36] T. Beyhl, R. Hebig, H. Giese, A model management framework for maintaining traceability
links, in: Software Engineering, 2013.
[37] A. Pourshahid, I. Johari, G. Richards, D. Amyot, O. Akhigbe, A goal-oriented, business
intelligence-supported decision-making methodology, Decision Analytics 1 (2014). doi:10.
1186/s40165- 014- 0009- 8 .
[38] A. Diaconescu, J. Pitt, Holonic institutions for multi-scale polycentric self-governance, in:
A. Ghose, N. Oren, P. R. Telang, J. Thangarajah (Eds.), Coordination, Organizations, Institu-
tions, and Norms in Agent Systems X - COIN 2014 International Workshops, COIN@AA-
MAS, Paris, France, May 6, 2014, COIN@PRICAI, Gold Coast, QLD, Australia, December 4,
2014, Revised Selected Papers, volume 9372 of Lecture Notes in Computer Science, Springer,
2014, pp. 19–35.
[39] A. Simon, The architecture of complexity, in: Proceedings of the American Philosophical
Society 106, 1962, pp. 467–482.
[40] L. Lúcio, S. Mustafiz, J. Denil, H. Vangheluwe, M. Jukss, FTG+PM: An Integrated Framework
For Investigating Model Transformation Chains, in: International SDL Forum, 2013, pp.
182–202.
[41] C. Gomes, C. Thule, D. Broman, P. G. Larsen, H. Vangheluwe, Co-simulation: A survey,
ACM Computing Surveys 51 (2018) 49.
[42] A. Barišić, D. Blouin, V. Amaral, M. Goulão, A requirements engineering approach for
usability-driven dsl development, in: Proceedings of the 10th ACM SIGPLAN International
Conference on Software Language Engineering, SLE 2017, Association for Computing
Machinery, New York, NY, USA, 2017, p. 115–128. URL: https://doi.org/10.1145/3136014.
3136027. doi:10.1145/3136014.3136027 .
[43] V. Weerakkody, H. Lee, A. Ko, T. D. Medeni, R. El-Haddadeh, A. Tsohou, K. Al-Yafi, Z. Irani,
L. M. Campos, Supporting public policy making processes with workflow technology:
Lessons learned from cases in four european countries, Int. J. Electron. Gov. Res. 8 (2012)
63–77. URL: https://doi.org/10.4018/jegr.2012070104. doi:10.4018/jegr.2012070104 .