=Paper=
{{Paper
|id=Vol-2541/paper1
|storemode=property
|title=Towards a Systematic Process for the Elicitation of Sustainability Requirements
|pdfUrl=https://ceur-ws.org/Vol-2541/paper1.pdf
|volume=Vol-2541
|authors=Barbara Paech,Ana Moreira,Joao Araujo,Peter Kaiser
|dblpUrl=https://dblp.org/rec/conf/re/Paech00K19
}}
==Towards a Systematic Process for the Elicitation of Sustainability Requirements==
https://ceur-ws.org/Vol-2541/paper1.pdf
Copyright © 2019 for this paper by its authors. Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0)
Towards a Systematic Process for the Elicitation of
Sustainability Requirements
Barbara Paech Ana Moreira, João Araujo
Institute for Computer Science NOVA LINCS/ Departamento de Informática
Heidelberg University FCT, Universidade NOVA de Lisboa
Heidelberg, Germany Caparica, Portugal
paech@informatik.uni-heidelberg.de {amm| joao.araujo}@fct.unl.pt
Peter Kaiser
Department of Computer Science
University of Applied Sciences Mannheim
Mannheim, Germany
p.kaiser@hs-mannheim.de
Abstract—In recent years, several approaches have been pro- duction of costs in software development and evolution influ-
posed in the area of sustainability requirements. They have clas- ences the economic dimension, and the software's ability to
sified sustainability into different dimensions and explored these cooperate with other systems impacts the technical dimension.
dimensions as well as the relationships of sustainability require- These complex interrelationships require new approaches for
ments to functional requirements and quality requirements, oth-
er than sustainability. Furthermore, case studies were reported.
requirements engineering to support appropriate decisions and
Only a few papers describe a systematic process for deriving sus- actions while maintaining a balance between the five dimen-
tainability requirements for a specific system. In this paper, we sions.
report on work in progress on such a process. The main idea is The goal of this work is to support the elicitation of sustain-
to provide a checklist of general and IT-specific details for the ability requirements. In recent years, several approaches have
sustainability dimensions and a checklist of general influences been proposed in the area of sustainability requirements, e.g.
between the dimensions. These checklists can be used to iterative- [2,7,8,11]. They have explored the dimensions as well as the
ly refine the requirements of a specific system with sustainability relationships of sustainability requirements to functional and
considerations which balance the different dimensions. We sketch
quality requirements other than sustainability. Furthermore,
this process and illustrate it with an example.
Index Terms — Sustainability dimensions, sustainability
case studies were reported. Only few papers describe a system-
requirements elicitation, sustainability checklists. atic process for deriving sustainability requirements for a spe-
cific system. In this paper we report on work in progress on
such a process. The main idea is to provide a checklist of gen-
I. INTRODUCTION eral and IT-specific details for the sustainability dimensions
Sustainable development is, according to [3], the “devel- and a checklist of general influences between the dimensions.
opment that meets the needs of the present without compromis- These checklists can be used to iteratively refine the require-
ing the ability of future generations to meet their own needs”. ments of a specific system with sustainability considerations
Attempts to comprehend this global challenge require the un- which balance the different dimensions.
derstanding of the interrelationships between social equity, The rest of the paper is structured as follows: in section II,
economic growth, and environmental degradation. Redclif [9] we present the process model and a metamodel of our concepts.
studies these interrelationships and proposes policy solutions to Section III and IV present the needs and effects checklists, re-
integrate these three dimensions of sustainability. Today, sus- spectively. Section V discusses an example and Section VI
tainability is investigated also from the individual and technical examines related work. Section VII concludes the paper and
dimensions, and it has become a key issue worldwide. Achiev- discusses directions for future work.
ing sustainable software development is a challenge that must
consider the above five distinct dimensions [7]: individual, II. METAMODEL AND PROCESS
social, economic, environmental and technical. Each dimen- In the following we first present the metamodel and then
sion addresses different needs (e.g., reduce costs, guarantee the process model. The metamodel is shown in Figure 1.
efficient energy consumption, avoid unemployment, evolve the Sustainability is divided into 5 sustainability dimensions
system easily) and impacts on other dimensions and their [1]. Each Dimension corresponds to a main Surrogate Stake-
stakeholders. For example, the use of software for general im- holder. This surrogate stakeholder has Needs which must be
provement of people’s lives affects individuals and society, satisfied by the system. (We introduce this surrogate stakehold-
memory and power efficiency impact on the environment, re- er, because it is more natural for a stakeholder to have needs
than for a dimension.) These needs describe the main sustaina- safety. They structure the elicitation according to standard is-
bility issues for each dimension. For example, the environmen- sues.
tal dimension has the surrogate stakeholder nature which has
the need little waste. Similarly, the economic dimension has the III. NEEDS CHECKLIST
surrogate stakeholder company with the need high revenue. When developing a system, it is useful to have at hand a
Needs can be of two kinds: they hold in general or they are checklist for the issues to be considered for each dimension.
specifically related to IT-Systems. IT-specific needs detail in We phrase these issues as needs of the surrogate stakeholder.
which way the system can influence a general need. For exam- We think this is helpful because many of these needs, especial-
ple, trust in the system is an IT-specific need of the surrogate ly for the environmental dimension and partly for the others,
stakeholder individual which details the general need agency. are not represented by real stakeholders. We have collected a
We emphasize IT-specific needs, because we propose a process first list of needs based on the definition of the dimensions and
to deal with IT-requirements. The needs influence each other related work (see Table 1). A general discussion of the contents
by means of Effects. An effect is a positive or negative or neu- in this table for each dimension is presented next.
tral relationship between needs. So, a need can trigger an effect
or be affected by the effect. For example, there is an effect be-
tween little waste and revenue. Little waste can lead to less
revenue, if it is costly to avoid waste. Thus, little waste triggers
this effect and revenue is affected by it. Note that these effects
are different from those in [1], where the software system has
systemic effects. Requirements should to satisfy the needs.
Effect
1..*
Requirement affects
triggers
0..*
Social
satisfies 1..* Surrogate
Need has Stakeholder
1
corresponds to Economic
1 Fig. 2. Process model
Dimension Environmental
General IT-specific The environmental dimension “covers the use of and
1..*
need need stewardship of natural resources. It includes questions ranging
Individual
from immediate waste production and energy consumption to
Sustainability the balance of local ecosystems and concerns of climate
Technical
change” [1]. This can be summarized as respect nature in the
Fig. 1. Sustainability metamodel, extended from [2] use of resources. Thus, the nature is the surrogate stakeholder.
Its needs are little waste, little pollution and little resource con-
The process model is shown in Figure 2. We envision an it- sumption, and positive influences to reduce climate change.
erative process which starts identifying some requirements Note that compared to the definition we have added “pollution”
(functional or non-functional, high-level or detailed) about the and generalized “energy” to “resource”.
core functionality of the system. To understand the sustainabil- The technical dimension “covers the ability to maintain
ity issues of the system, first the relevant needs from the di- and evolve artificial systems (such as software) over time. It
mensions are identified based on a checklist of needs. Then the refers to maintenance, evolution and resilience, as well the ease
effects between these needs are identified based on the effect of system transitions [1]. This can be summarized as respect
checklist. For each need affected by the effects on the list it is longevity of the system. Here, the surrogate stakeholder is the
checked whether the given requirements satisfy this need. All system itself. Its needs are easy maintenance, evolution and
requirements satisfy the involved needs as far as possible. It is resilience, as well the ease of system transitions.
necessary to make tradeoff-decisions balancing the dimensions, The social dimension “covers relationships between indi-
meaning that not all needs can typically be fully satisfied at the viduals and groups. For example, structures of mutual trust and
same time. The sustainability requirements are new require- communication in a social system and the balance between
ments considering a need which has not been considered be- conflicting interests” [1]. This can be summarized as respect
fore. In addition, there can be changes of existing requirements society. Thus, the surrogate stakeholder is the society. Further
to make the core functionality more sustainable by satisfying needs were derived from the indicators gathered in [4], particu-
the involved needs. The requirements are integrated into the larly: high trust and communication between people, little con-
pool of requirements and the next iteration of the process can flict of interests (for society as a whole or subsets like munici-
start with the new or refined requirements and/or further needs. pality, state, region, community, citizen), good employment
Thus, needs and effects play a role similar to, e.g., threats for good health, equity, good education, good security, good ser-
vices and facilities, good resilience, high-level of human rights,
good social acceptance of technology, good social cohesion, Individual (Surrogate stakeholder Individuals such as users or workers in
good preservation of culture, good governmental laws and trust system production or operation)
• High freedom
of the people in them.
• High agency
The economic dimension “covers financial aspects and • High human dignity
business value. It includes capital growth and liquidity, ques- • High fulfillment
tions of investment, and financial operations” [1]. This can be • High trust of the user in the system (e.g., safety, privacy, transparency
summarized as respect involved companies and governmental of the lifecycle)
institutions. Thus, the surrogate stakeholders are the companies • Equal access of the system by the users
• Fair treatment of users by the system
involved in the production and operation of the system and the
responsible governmental institutions. To detail the benefits, This list is not complete but covers many important needs.
we looked at the balanced scorecard [5] which distinguishes the
financial perspective, the innovation and learning perspective, IV. EFFECT CHECKLIST
the customer perspective and the internal business process per- The needs are very diverse and influence each other. It is a
spective. The financial perspective can be refined into owner- lot of work to look at all needs and influences for each system.
ship, simple financial operations, little cost and high revenue. We think that the influences can be generalized to typical influ-
The individual dimension “covers individual freedom and ences between needs called effects. These effects can be de-
agency (the ability to act in an environment), human dignity rived from typical ways in which the stakeholders influence
and fulfillment. It includes the ability of individuals to thrive, each other. For example, nature is influenced by the behavior
exercise their rights and develop freely” [1]. This can be sum- of the system and of the users or workers. Thus, there are ef-
marized as respect individuals. Thus, the surrogate stakehold- fects between some environmental and technical or individual
ers are the individuals such as the system users or workers in needs. The companies have indirect influence on nature
the system production or operation. We added to these general through the system. Similarly, the society does not directly
needs IT-specific needs, which detail in which way a technical influence nature, but through influence on people or compa-
system can affect the general needs. One example is good trust nies. Companies clearly have influence on the system, and the
of the user in the system to support their agency (as they will system makes the difference for the well-being of the compa-
not work with the system, if they do not trust it). Similarly, nies. People and companies influence society, and also the oth-
equal access to the system by the users is needed for agency, er way around.
while fair treatment of users by the system will support human Based on the general influences between the stakeholders
dignity. we can reason about the effects. We want to provide a checklist
of typical effects for each need. Figure 3 gives an example for
TABLE I. INITIAL CHECKLIST FOR THE NEEDS FOR EACH DIMENSION such a checklist derived in a discussion session between the
Environmental (Surrogate stakeholder Nature) authors. It shows three effect groups. The green effect group
• Little resource consumption comprises the effects of high employment, the black group
• Little pollution (emission, noise, visual) comprises effects of good evolvability, and the blue group of
• Little waste
little waste. In the following we explain the groups.
• Positive influence on climate change
Technical (Surrogate stakeholder System)
• Easy maintenance of a system (e.g., quality/durability of material; skills Good Evolvability of the system:
of workers, easy disassembling: modular structure; predictive mainte- • supports that little waste and little pollution are pro-
nance) duced during the system lifecycle (production and op-
• Easy evolution of a system (e.g., modular structure; skills of workers;
adaptability; customization)
eration and demolition);
• Easy resilience • can decrease or increase the business value for the
• Easy of system transitions customer. It will increase the customer satisfaction
Social (Surrogate stakeholder Society or parts thereof such as municipali- and the stability of the processes, and therefore maybe
ty, state, region, community, citizen) also the revenue. However, if the mechanisms for
• Good trust and communication between people
• Little conflict of interests
evolvability are expensive, it will lower the revenue.
• Good social indicators: employment (full-time work, women employ- High Employment rate:
ment, working time arrangements, job opportunities, wages); health;
equity; education; security; human rights, social cohesion, preservation • supports dignity and freedom of individuals;
of culture, governmental laws and trust of people in them • can decrease/increase financial aspects of the compa-
Economic (Surrogate stakeholders Companies and Governmental Insti- ny. If the company employs more humans instead of
tutions) machines, the cost might be higher and thus the reve-
• Good distribution of ownership of parts
• Simple financial operations
nue lower. But on the other hand, people have more
• Little cost money to buy new products. It supports stable pro-
• High revenue (for capital growth, liquidity, financial investment) cesses within the company.
• High customer satisfaction
• High level of innovation and learning
• Stable business processes
Little waste (or No waste): graphed (to initiate a legal procedure to fine the owner of the
• supports the health of society. It reduces, however, the vehicle later).
freedom of individuals, as they have to take care of We demonstrate one cycle of the iterative process. We as-
waste. This can be expensive for the production, re- sume the functional and quality requirements are defined from
sulting in lowering the revenue, or, in contrast, it can the point of view of the toll gate system stakeholders. One of
increase the revenue, if less material is needed. Also, these requirements is “a car shall be identified by a gizmo”. We
the stability of the processes is negatively affected, flip through the checklist to find needs, which are heavily in-
because processes need to be changed. A high level of fluenced by this requirement: we take little waste. It is of im-
innovation (shown in bracket) is not result of little portance because of the huge number of registered vehicles,
waste, but good innovation needs to be present to each one with one device (or gizmo) on its windscreen. There
achieve little waste. will be about 1.5 million if one third of all cars in Portugal are
equipped with it. (And the system is already in use in other
countries.)
Looking at the effect group “little waste” we see that the
most related needs according to our effect checklist are: “health
of society, high freedom of the user, and revenue and stable
process of the company”. Therefore, which sustainable re-
quirements can we propose to handle the selected need and
what is their influence on the effects group needs (i.e., do they
confirm the identified impacts)? We propose to address the
“little waste” need by adding sustainability requirements like
“expendable materials shall be replaceable by the driver”, “the
gizmo shall have a modular/easy-to-repair structure”, “the giz-
mo shall have a hardened case”. In this way, the device and its
parts have a long lifetime and hence we reduce waste. We can
mark the need “little waste” as changed to the better. So, the
requirements seem to satisfy this need.
Now, we have to check whether the new sustainability re-
quirements have a negative effect on the needs of the remaining
dimensions. From the “little waste” effects group, “health” is in
Fig. 3. Screenshot from a meeting between the authors. Effect groups for general positively influenced by “little waste” and we think this
employment (in green), evolvability (in black) and little waste (in blue).
(The identifiers of the nodes in the graph are composed of the dimension also holds true in our example, particularly regarding the re-
initial together with the name of the stakeholder, e.g., Tsytem means the quirements “the gizmo shall have a modular, easy-to-repair
stakeholder system of the Technical dimension. The economic structure” and “the gizmo shall have a hardened case”. We
dimension uses the initial F, as E was already taken). decide that in this case the requirements can also be satisfied by
not so expensive materials. Therefore, we think that the reve-
For a specific system these general effects must be detailed
nue will be higher due to less material needed. Now we have to
so that the value (positive, negative, neutral) of the effect can
think about the two negative influences: little waste may lower
be determined.
the freedom of the users. However, in this case the waste of the
V. EXAMPLE production is reduced, and the individuals do not have to
change their behavior. Regarding the requirement “materials
Let us now apply the metamodel and the process exemplari- shall be replaceable by the driver”, we think that, in fact, allow-
ly to a Toll Gate System, a simplified version of the Via Verde ing the vehicle owner to change, for example, the batteries of
toll collection system in use since 1991 on the Portuguese the gizmo will have a positive impact on her behavior as she
highways. In this system, drivers of authorized vehicles are does not need to drive to the nearest Via Verde shop for some-
charged at toll gates automatically, when passing in special thing so simple (and this with extra positive effect for the envi-
lanes, known as green lanes. A gizmo device must be installed ronment and the vehicle owner finances). Thus, the potentially
at the windscreen of the vehicle. After registration, the gizmo is negative influence is not important and does not require an
sent to the client to be activated using an ATM (this associates additional tradeoff. However, to accommodate the new sustain-
the gizmo identifier with the car owner bank account number ability requirements, the company’s production has to be
for direct debits). A gizmo is read by the toll gate sensors and adapted, leading to unstable processes. This negative impact
the information is stored by the system and used to debit the car requires a tradeoff. However, we think this will only be a short
owner account. The amount paid depends on the class of the instability, and thus we prioritize little waste and accept this
vehicle, and on motorways it also depends on the distance trav- instability. Altogether, we have made sure in our process that
elled. If an exception is detected (e.g., the vehicle is non- all needs of this effect group are satisfied by the given and the
authorized, the gizmo identifier is invalid, or the vehicle’s class new requirements.
does not correspond to the registered one) a yellow light is
turned on, an alarm sounds, and the plate number is photo-
This process needs to be repeated and new iterations per- (LPs), which are “locations within a system where a small
formed for all the needs considered relevant for the problem change in one aspect can result in significant system-wide
under analysis. The result is an extended set of requirements, as changes”. LPs provide an analysis tool to help software engi-
well as a set of priorities and respective tradeoffs that need to neers to face sustainability challenges through insights on
be taken into consideration in the development phases that fol- transformation mechanisms or strategies to find alternatives.
low. They do not give guidance on the dimensions.
When thinking about the need and effects, it can be that Oyedeji et al [6] propose a sustainability design catalogue
new needs or effects are discovered. A new effect can be a to- to assist software developers and managers in eliciting sustain-
tally new relationship or an added positive or negative value. ability requirements. It is based on the Karlskrona manifesto
This means that the checklists should be updated. principles and the indicators (related to the sustainability di-
Knowing the effects on all needs, the team must decide mensions and their order of impacts) of sustainability associat-
whether the specification is now acceptable. If not, further iter- ed with each criterion. The orders of impact cover the positive
ation is necessary resulting in different requirements (sustaina- and negative effects of software on the environment, including
bility or others). Obviously, traceability between needs, effects immediate effects, enabling effects and structural effects. They
and requirements is essential. do not cover the effects between all dimensions.
Of course, this exercise serves for illustrative purposes on-
ly, for a system that exists and works well for almost 30 years VII. CONCLUSION AND FUTURE WORK
now. Given the current advances in plate number recognition We have outlined a process and two checklists to support
technologies, for example, a more sustainable solution for the the systematic elicitation of sustainability requirements. We
part of the problem discussed above, would be to equip the toll introduced needs and effects as main concepts to structure the
gates with recognition technology. This would avoid the need elicitation. Clearly, a lot of work is necessary to apply this fully
for the gizmo, even though it would, of course, still require the to a real example. The work should continue by extending the
registration of the vehicle in the system to allow automatic need checklist. On one hand we should look for detailed indica-
bank debits. However, our point is: overall, the process helps a tors, similarly to the societal dimension for all other dimensions
team to design a sustainable system by providing them with a to derive more detailed lists. On the other hand, we should look
systematic way of considering all relevant information. into each dimension for IT-specific needs. Similarly, the effect
checklist should be extended. For each need we should derive
VI. RELATED WORK an effect group. If new needs emerge, the groups have to be
In the following we discuss related work which also gives updated. Furthermore, the iterative process has to be tested.
guidance for the elicitation of sustainability requirements. How many needs are typically relevant for a system? Should
Work proposing a metamodel for sustainability (e.g., [7,2, we first look at the needs of one dimension and then iterate
10]) mainly defines meta-concepts and applies them directly to through the other dimensions? Or is it better to iterate through
a case. In this paper we make an effort to standardize the the effects’ groups? How much work is it to get to specific re-
needs and effects of the dimensions. quirements from the general needs and effects? If we succeed
Penzenstadler and Femmer [7] introduce a reference meta- in providing comprehensive checklists, elicitation of sustaina-
model used to instantiate generic models for sustainability, bility requirements will be adopted more easily in practice.
decomposing it into the five dimensions. The aim of the model
is to serve as a reference model for both process and require- ACKNOWLEDGEMENTS
ments engineer who instantiates the model for a software de- The first and last author contributed to this work during
velopment company or for a specific system under develop- their sabbatical at NOVA LINCS. We would like to thank
ment, respectively. Detailed guidance for the dimensions is not NOVA LINCS (current funding ref. UID/CEC/04516/2019)
given. and our universities for supporting this research work.
Brito et al. [2] extend Penzenstadler and Femmer’s meta-
model to accommodate concern responsibilities and tradeoff REFERENCES
management. The authors treat sustainability as concern and [1] C. Becker, S. Betz, R. Chitchyan, L. Duboc, S. Easterbrook, B.
specify the five dimensions as concerns based on a template. Penzenstadler, C. Venters: Requirements: The key to
This template is based on their extended metamodel. Again, sustainability. IEEE Software, 33(1), 2016, 56-65.
detailed guidance for the dimensions is not given. We take their [2] I. Brito, J. Conejero, A. Moreira, J. Araújo: A concern-oriented
idea of effects. sustainability approach. RCIS 2018: 1-12.
Saputri and Lee [11] proposed a GQM-based approach to [3] Brundtland Commission. Our Common Future. Oxford Univer-
define sustainability requirements from stakeholders needs. sity Press: Oxford.,1987.
Sustainability property analysis is performed to evaluate impact [4] M. Hinai, R. Chitchyan: Social Sustainability Indicators for
and trade-off analysis of those requirements. A metamodel for Software: Initial Review. RE4SuSy@RE 2014: 21-27.
sustainability is provided, but without considering effects at a [5] R.S. Kaplan, D. Norton: The Balanced Scorecard: Measures
more fine grained level. That Drive Performance, Harvard Business Review, Reprint
Penzenstadler et al. [8] propose an approach to identify from 1992, July/August 2005, pp.172-181.
successful sustainability interventions using leverage points
[6] S. Oyedeji, A, Seffah, & B. Penzenstadler: A catalogue sup- [9] M. Redclift: Sustainable Development Sust. Dev. 13, 2005,
porting software sustainability design. Sustainability, 10(7), 212–227 (Wiley InterScience www.interscience.wiley.com).
2296, 2018. [10] K. Roher and D. J. Richardson: Sustainability requirement
[7] B. Penzenstadler and H. Femmer: A generic model for sustain- patterns, in Third IEEE International Workshop on Require-
ability with process- and product-specific instances, in Work- ments Patterns, RePa 2013, Rio de Janeiro, Brazil, July 16,
shop on Green in/by software engineering – GIBSE. New York, 2013, 2013, pp. 8–11.
New York, USA: ACM Press, 2013, pp. 3–8. [11] T.R.D. Saputri, S.W. Lee: Incorporating sustainability design in
[8] B. Penzenstadler, L. Duboc, C. Venters, S. Betz, N. Seyff, K. requirements engineering process: a preliminary study, in: S.
Wnuk, R. Chitchyan, S. M. Easterbrook, C. Becker: Software Lee, T. Nakatani (Eds.), Requirements Engineering Toward
Engineering for Sustainability: Find the Leverage Points! IEEE Sustainable World. APRES 2016, Springer, Singapore, 2016,
Software 35(4): 22-33 2018. pp. 53–67 671 CCIS.