Contents Introduction
Human values express what is essential for an individual or a society [1]. They are the principles that guide human actions and behaviour in daily life [2]. In recent years, human value research has gained attention in software engineering [3–6]. The existing studies [7, 8] have shown the importance of considering human values, e.g. they may significantly influence one’s intention to install and use the software. Also, violating or ignoring human values can cause adverse outcomes such as damage to a company’s reputation and loss of profits [9]. For example, the Facebook Cambridge Analytica scandal disclosed a data breach in which users’ private data was used for political reasons without consent. With this, Facebook violated user values, such as Security and Trust. As a result, users’ trust in Facebook dropped by 66%, suffered substantial reputational damage, faced harsh public scrutiny, and experienced a significant decline in profits [10]. The existing research mainly focuses on the identification of human values from different sources such as issue trackers [3], user stories [11], and app reviews [12]. However, little attention has been paid to other requirements engineering (RE) artefacts, such as goal models and particularly motivational goal models [13–17], which constitute an important artefact for representing early requirements [18, 19].
The methods of conceptual modelling using goal models and value models to develop digital healthcare solutions carry significant limitations. For example, the value models described in [20] mainly focus on increasing the business value of the service to patients instead of focusing on patients’ emotions and human values. Similarly, the authors in [21] developed a three-level goal hierarchy constituting disease-specific goals, functional goals and fundamental goals, representing patients’ life priorities. The authors argue that explicating fundamental goals or values can help identify mismatches between medical standards and patient values, such as treating multi-morbidity. While helpful in conceptualising patient goals, the model ignores the consideration of goals and value-oriented requirements for any other category of stakeholders, such as caregivers or relatives. The work described in [22] provides a checklist that includes patient long-term goals that aid in developing healthcare technology to support patients’ recovery process in their home environment. Again, this work only caters for the goals of patients. Recent research on emotion-oriented requirements [23] that relates more closely to our work has attempted to link functional and emotional goals with system requirements for a wider group of stakeholders. However, the study does not apply any existing theory of human values to map the emotional goals to the values of multiple stakeholder groups.
Motivated by this, to the authors’ knowledge, our study is the first kind of study that explores the presence of human values in motivational goal models [13–17]. To address this research gap, we investigate the following research questions:
RQ1: Can we identify human values in motivational goal models?
RQ2: What are the trends of human values identified in motivational goal models?
The contribution of our work is establishing the link between the emotional goals of motivational goal models and human values based on the Schwartz theory of human values [1]. We demonstrated the identification of human values in the industrial research and development project Pharaon 1, which comprises 40 partners. The Pharaon project aims to develop an ecosystem of technological solutions to improve the quality of life of older adults and their caregivers. The project has five large-scale trial applications, henceforth termed as pilots in this paper. Overall, we identified 9 out of 10 value categories and 28 out of 58 value items based on the Schwartz theory of human values [1] from six motivational goal models developed in the Pharaon project. Almost 45% of the identified values in our project belonged to the value categories of Security and Self-direction. Our finding is a significant insight that highlights the dominant stakeholder expectations beyond the main functional requirements for the offerings of digital healthcare and well-being services for older adults and the related stakeholder groups. Our study also indicates the possibilities to identify human values in software development artefacts open up several directions for future research.
The rest of this paper is structured as follows. In Section 2, the notion of human values and goal-based requirements modelling are explained. Section 3 elaborates on our research methodology by providing detailed descriptions of the case studies and explaining the methods employed for data collection and analysis. The findings obtained from our study are presented in Section 4. Subsequently, in Section 5, we discuss the results and address potential threats to validity. The related work is overviewed in Section 6. Finally, conclusions are drawn in Section 7.
Background
2.1 Human Values
The concept of “human values” has been thoroughly researched in psychology and sociology. Human values can be defined as principles that guide human actions and behaviour in daily personal and social life [2] and modes of conduct that a person likes or chooses among different situations [24]. Notable value theories are the value survey by Rokeach [2], culture dimensions by Hofstede and Bond [25], theory of basic human values by Schwartz [1, 26], and functional theory of human values by Gouveia and others [27]. In this study, we use the theory of basic human values by Schwartz [1] to link emotional goals with the corresponding value categories and value items included in the Schwartz theory of human values. We chose the Schwartz theory over the other theories of human values for three reasons. First, the Schwartz theory has been widely used in software engineering [3, 6, 11, 28] and requirements engineering [5, 12, 29, 30]. Second, the theory itself has been extensively validated in various domains and findings from across different cultures [31] and in 83 countries [32]. Third, according to [33], the Schwartz theory subsumes other theories of human values.
The Schwartz theory [1] distinguishes ten categories of human values and divides them into the main categories, which are the values with a personal focus and social focus. Values with a personal focus are linked to either self-enhancement or a person’s openness to change. The values with a social focus concern how one relates socially to others and affects their interests. A complete list of these ten categories along with their definitions is provided in Table. I. Additionally, the table includes all 52 individual value items identified by Schwartz. In the Schwartz theory, human values are arranged in a circular manner so that supporting values are adjacent to each other while values that are conflicting are in the opposite direction [1].
2.2 Motivational Goal Modelling
Humans are goal-directed creatures. Human motivation energizes, directs, and sustains their goal-directed activities [34]. Motivational modelling [35–37] is a method that allows ethnographers and requirements engineers to elicit and represent emotional requirements [38] for sociotechnical systems [39] related to the goals to be achieved. In motivational modelling, three kinds of goals– do, be, and feel goals – are elicited from project stakeholders. Do goals or functional goals describe what the system to be designed should do or achieve, be goals or quality goals describe how the system should be, or more specifically the quality characteristics of the things to be done or achieved, and feel goals or emotional goals describe how a user performing a particular role should feel when using the system or, in other words, what emotions should be constructed in the brain of the user when he or she uses the system. The difference between do and feel goals is that while functional goals characterize the system to be designed and created, emotional goals describe what emotions using the system should construct in the minds of its users, based on the theory of constructed emotion [16, 40]. The meaning of a functional goal as compared with activity or task in this context is that a goal is not directly executable, i.e., it leaves open how it should be done or achieved.
Motivational goal modelling represents user requirements by hierarchical goal models that decompose a high-level purpose of a system to be designed into lower-level goals, each of which represents a particular aspect of achieving its parent goal [13–17]. The notation for representing motivational goal models is shown in Fig. 1. The skeleton of a goal model is a hierarchy of functional goals drawn as a tree in a top-down manner, starting with the highest-level goal, which represents the purpose of the system. The hierarchical structure is used to indicate that achieving a sub-goal represents an aspect of achieving its parent goal. The roles, quality goals and emotional goals are attached to relevant functional goals at an appropriate level in the goal hierarchy. They also apply to the sub-goals of each relevant functional goal. An emotional goal is also attached to one or more roles, indicating how the performers of the respective roles should feel.
![Table I- Value Categories, Descriptions, and Value Items [1]](/mediastore/IEEE/content/media/10555577/10556397/10556435/iqbal.t1-053300a024-large.gif)
An advantage of motivational goal modelling over other goal modelling methods is that it enables to represent emotional requirements as first-class citizens in the form of emotional goals [16]. Motivational goal modelling also supports well the communication between technical and non-technical stakeholders [16, 17, 40–45], which is very relevant in our case study
According to Schwartz [1, 32], human values can be defined in terms of one’s motivational goals, where motivational differences between values can be seen as continuous rather than as discrete [46]. In other words, human values are not discrete entities but rather correspond to a continuum of motivational goals. However Schwartz’s findings indicate that people in most cultures respond to ten types of values as distinct entities [1].
We should also consider that human values by Schwartz are represented at the “abstract value level”, whereas values are manifested in specific behaviors at the “instantiation level” [47]. There is also inherent variability in the interpretation and application of a value [47]. For example, an Older Adult may perceive the value item “Capable” as referring to their mental capabilities rather than their physical capability to move around.
It’s noteworthy that motivational goals by Schwartz [1, 32] are more abstract than goals used in motivational goal modelling [35–37] discussed in our study. While Schwartz’s motivational goals [1, 32] represent situations desired by individuals, such as ‘Independent thought and action’s choosing’, the goals of a motivational goal model, represent what should be done or achieved and with what quality and how users should thereby feel, such as ‘Monitor health’. Therefore, in identifying human values, we should primarily focus on the emotional goals of motivational goal models, considering that ‘emotion is the dominant source of value importance’ [48]. Moreover, Thew and Sutcliffe [49] emphasize the importance of emotions by the stakeholders in understanding the values embedded in a socio-technical system. Maio [47] states that ‘emotions are the primary source of information for consensually important values’, and Schwartz notes that ‘when values are activated, they become infused with feeling’, leading people to feel aroused when their values are threatened [1]. In requirements engineering, emotions concerned with human values can be represented as emotional requirements in the form of emotional goals of motivational goal models. However, emotions alone are insufficient for identifying values from motivational goal models. We also need to know what a person performing a particular role should do or achieve. For example, the human values identified for the emotional goal Informed differ for the roles of Older Adult and Informal Caregiver because they need to feel informed about different aspects.
This paper addresses the identification of human values from motivational goal models via emotional goals, supplemented by contextual information provided by roles and functional goals associated with the emotional goals
Methodology
In this section, we first briefly discuss the Pharaon project and the data collection process. Furthermore, we describe the manual analysis of motivational goal models that we performed to identify human values in goal models. We also discuss the protocol and steps followed in our analysis.
3.1 Case Study
Pharaon is a project within the EU Horizon 2020 research and innovation funding program (EU Grant Agreement 857188). The purpose of the Pharaon project is to facilitate older adults to be independent and healthy and for their caregivers to have a comfortable environment with up-to-date information and communication technology solutions for taking good care of their relatives and patients. This is achieved by a set of integrated, highly customizable, and interoperable open platforms with advanced services, devices, and tools. Pharaon has 40 partners from 12 European countries, contributing to the development of open platforms and testing several digital solutions in six different pilots carried out in the following five European countries: Italy (the pilot sites of Tuscany and Apulia), Spain (the pilots of Murcia and Andalusia), the Netherlands, Slovenia, and Portugal (the pilot sites of Coimbra and Amadora). Pharon has 15 industrial partners, 4 research organisations, 7 universities 3 healthcare providers, 3 public authorities, 6 non-governmental organisations, and 2 standardisation bodies. The project lasts from the beginning of December 2019 until the end of November 2024.
A user-centric approach is applied in the Pharaon project to maximize the final usability and acceptance of the open Pharaon socio-technical ecosystem [50] being created by all stakeholders. The main end users of the Pharaon ecosystem will be older adults, whereas several other stakeholders, such as healthcare professionals, formal and informal caregivers, volunteers, governmental and non-governmental organizations, and others, have been identified and involved. The input gathered during the requirements elicitation process was utilized for representing user and pilot requirements for all six pilot sites from five countries involved in the project. The requirements were elicited by different means: face-to-face and virtual stakeholder workshops, online interviews and questionnaires, phone interviews, and review of the existing literature, including previous projects [45]. The project partners of the respective pilots represented early requirements as motivational goal models and transformed them later on into scenarios of motivational modelling [13] and into user stories [51–53]. The requirements served as a foundation to define the initial architecture of the Pharaon socio-technical ecosystem and will guide the project activities to achieve and deploy a successful final system. For the focused and detailed analysis presented in this paper, we have randomly chosen three pilots: Italy, the Netherlands, and Slovenia.
3.2 Data Collection
We manually analyzed the goal models from three pilots2 of the Pharaon project to discover human values. In the analysis presented in this paper, we included six goal models from three pilots. Fig. 2 shows the goal model of the Italian pilot3. The notation for motivational goal models is shown in Fig. 1. According to the goal model represented in Fig. 2, the highest-level goal – or purpose of – the socio-technical system modelled in the figure is to Manage health. The main stakeholder roles are Older Adult (whose health is managed), Informal Caregiver and Professional Caregiver. Roles are attached to the functional goals for the achievement of which they are responsible. Functional goals can be associated with emotional goals that model how the stakeholders performing the corresponding roles should feel when achieving the functional goals. Examples of emotional goals when achieving the functional goal Manage health are Empowered for Older Adult, Informed for Informal Caregiver, and Aware for Formal Caregiver. Quality goals like Easy to use represent quality aspects of how the corresponding functional goals, like Manage health in the given case, should be achieved. The functional goal Manage health has been elaborated into four second-level functional goals. Each of these functional goals represents a particular aspect of achieving its parent goal, whereas the order of achieving the functional goals is not represented and will be defined only by scenarios of motivational modelling [13]. Some of these functional goals are elaborated further, following the same principle. Thus, roles, emotional goals and quality goals have been attached to the corresponding functional goals, when necessary, and they also apply to the sub-goals of each relevant functional goal.
3.3 Data Analysis
Our study involved five analysts (authors of the paper) ranging from PhD students to associate professors. To avoid biasedness, our analysts consisted of 2 females (40%), and 3 males (60%).
The overall analysis process of mapping human values from emotional goals in the context of the respective roles and functional goals is represented in Fig. 3. Before initiating the analysis process, we ran a trial session to learn if it was feasible to map emotional goals in the context of roles and functional goals to human values and if the outcome was reasonable. For this, we analyzed the goal model representing the requirements of the Andalusian pilot in Spain. Two analysts (the first two authors of this paper) who have software engineering background and sufficient knowledge about the Schwartz theory of human values [1] individually assigned human values to emotional goals, considering also their contextual roles and functional goals. Later, they discussed the results with the social scientist who is an expert in the Schwartz theory (illustrated in orange color in Fig. 3). We concluded from the trial session that the identification of human values from emotional goals was possible and the results were promising. Therefore, we planned the data analysis with five analysts as the next step. Please note that the results from the above-described trial session are not included in any further data analysis.
In the training session, the analysts attended a joint workshop where the Pharaon project was explained and also we demonstrated the goal model that was used in the trial session and discussed the mapping process. Each analyst was provided with supporting materials, including the project report on the requirements, goal models, and literature on human values. As a result, all analysts had a reasonable understanding of the project and human values before starting the analysis part.
As human values are subjective, we performed a mock session with all five analysts to bring everyone to the same level of interpretation and understanding of human values, Pharaon project, and motivational goal models. We divided the analysts into two groups, each consisting of two members. Each group individually analyzed one motivational goal model and assigned human values to emotional goals. The remaining analyst, whose background is in social science (illustrated in orange color in Fig. 3), moderated the discussion session as an expert in the Schwartz theory of human values. In the discussion session, each group shared their results with the expert and the other groups. All the results were discussed, and the expert analyst finalized human values and value items for emotional goals with a mutual agreement. This mock session mainly aimed to understand the process and criteria employed by the analysts to identify human values from emotional goals in the context of roles and functional goals.
After finishing the mock session, we provided each of the five analysts with the three pilots’ goal models to assign human values to emotional goals. At this point, each member was familiar with the process and confident to perform this mapping because of the earlier mock session. In this phase, each analyst analysed goal models for value items associated with human values by Schwartz. While performing the analysis, each analyst also considered project-related information in addition to the goal models. The analyst was allowed to propose any human value from the Schwartz value theory.
After finishing the individual analysis, we conducted a joint workshop session for each pilot with all five analysts. The workshop was moderated by the social scientist, and followed a negotiated agreement method [54, 55] to resolve any disagreements and conflicts. Using the negotiated agreement method, all analysts collaboratively agreed on the label of an item under review. In our analysis, the label was a human value attached to an emotional goal. According to the study [56], the negotiated agreement approach is particularly useful for addressing reliability issues of codes when there are multiple categories as opposed to two categories where a Cohen’s Kappa measure would suffice. We resolved conflicts through discussion and finalized the human value categories and value items with mutual agreement. As an output of the analysis, we created tables containing emotional goals mapped to the corresponding human value categories and value items. The tables also included functional goals and roles associated with emotional goals.
Qualitative Analysis of Goal Models. The overall qualitative analysis of all emotional goals from the project is illustrated in Fig 4, which shows the frequency of each emotional goal. We retrieved 46 emotional goals from six goal models that were developed by three pilots: Italy, Slovenia, and the Netherlands. Out of 46 emotional goals, 23 were distinct emotional goals. The emotional goals with the highest occurrence frequency were Informed, Empowered, and Safe. Emotional goals with the second highest occurrence frequency were In touch, Reassured, and Aware. The maximum and minimum number of occurrences for emotional goals were 5 and 1, respectively. As is shown in Fig. 4, the Italian pilot shared more similar emotional goals with the Slovenian pilot (7 goals) compared with the Dutch pilot (3 goals). This observation can be attributed to certain similarities between the Italian and Slovenian pilots. The highest number of distinct emotional goals originated in the Dutch (8 goals) and Italian (5 goals) pilots. Differently, the Slovenian pilot only had three distinct emotional goals.
Results
In the analysed three pilots of the Pharaon project, we were able to assign human values to emotional goals in the context of the respective roles and functional goals. The steps that we took for the mapping were as follows: (i) running a trial session with two analysts to establish the feasibility of mapping human values from motivational goal models; (ii) running a mock session of mapping emotional goals to human values initially with the five analysts in two groups, each consisting of two members, and thereafter in a joint discussion session, which was moderated by a social scientist who is an expert in the Schwartz theory of human values; (iii) assigning by each of the five analysts individually human values to emotional goals included by the motivational goal models of the three pilots; (iv) conducting for each pilot a joint workshop session with all of the five analysts, moderated by the expert in the Schwartz theory, and discussing the obtained results, where conflicts were resolved through discussion and the human values and value items mapped from emotional goals were finalized with a mutual agreement. Therefore, we answer RQ1 by stating that it is possible to identify human values in motivational goal models.
Fig. 5 answers RQ2, which is dealing with the trends of human values identified in the motivational goal models, by showing the total count of human values present in the three pilots. The figure contains the main value categories Self-direction, Stimulation, Hedonism, Achievement, Power, Security, Conformity, Benevolence, and Universalism. We identified 9 out of 10 human values by Schwartz in the pilots, with a total occurrence count of 79 times. Our analysis did not find any occurrence of the human value Tradition. The majority of the emotional goals refer to the human values Security and Self-direction, with the respective occurrence frequencies 20 and 15. The least prominent values appeared to be Universalism and Conformity, with the respective occurrence frequencies 2 and 4. Furthermore, we identified 28 out of 52 value items from the Schwartz theory of human values, as is shown in Fig. 7. The value items Sense of belonging and Healthy from the value category Security rank as the top value items identified in the project. The value item Sense of belonging was mapped from a maximum number of emotional goals, including Informed, In touch, Supported, Cared for, Belonging, Included, and Involved.
The presence of human values for each pilot is depicted in Fig. 6. Out of 9 human values, 7 were shared by all three pilots. The value categories Universalism and Conformity could not be found in the goal models of the Italian and Dutch pilots, respectively. The most prominent human value category for the Slovenian and Dutch pilots is Security. In contrast, the top value categories for the Italian pilot are Self-direction and Stimulation.
Although all pilots are parts of the same Pharaon project in a similar healthcare and well-being domain, results still vary for each pilot, which can likely be attributed to different functional goals, roles and emotional goals. As Italian and Slovenian pilots shared similar goals, which can be seen in Fig. 4, human values discovered for these two pilots are comparable. We can conclude that in addition to emotional goals, roles and functional goals also impact the presence of human values. Therefore, the presence of human value in different pilots can not be generalized.
Discussion
Overall, we can conclude from our study that it is possible to identify human values in motivational goal models. We identified human values according to the theory of human values by Schwartz [1] in motivational goal models that were created for representing early user requirements in an EU project on healthcare and well-being for older adults and their caregivers. In summary, we answer the research questions (RQs) that were posed in Section 1 as follows:
RQ1: It is possible to identify human values in motivational goal models. For this, we propose mapping emotional goals from motivational goal models to the human values by Schwartz, considering also the context provided by roles and functional goals. In our industrial case study, we identified in six goal models from three pilots of the Pharaon project human values belonging to 9 out of 10 human value categories and 28 out of 58 value items based on the Schwartz theory of human values.
RQ2: We found Security and Self-direction are the most prominent human values in the analysed three pilots, with the corresponding occurrence frequencies 20 and 15. The common prevalent human value category for all three pilots is Security. All three pilots share 7 out of 9 identified value categories, but these values are not evenly distributed, which can likely be attributed to different goal models.
In the following, we will discuss the answers to the RQs in more detail.
With respect to the research question RQ1, on the possibility of identification of human values in motivational goal models, we collected data from an ongoing EU project. We considered three pilots – Italy, Slovenia, and the Netherlands – and analysed their motivational goal models. Based on our results, we propose mapping emotional goals from goal models to human values based on the theory of human values by Schwartz, in the context of the corresponding functional goals and roles. We performed qualitative analysis described in Section 3.3 and demonstrated that 9 out of 10 human values could be mapped from emotional goals in the context of roles and functional goals. Our results did not find the human value Universalism for any of the analysed pilots. This can be due to the nature of the Pharaon project.
We observed during the analysis that different human values could be identified for the same emotional goal. A possible reason for this could be functional goals or roles in the models providing different contexts for the emotional goal. For example, in Table 2, the emotional goal Informed elicited for the respective roles Older Adult and Informal Caregiver is associated with the identical functional goal Manage health. However, the human values identified for Informed differ in these two cases because the roles associated with each case are different: Older Adult vs Informal Caregiver (i.e., relative or friend). According to the corresponding scenario, an older adult wants to feel informed about his/her health management. Therefore, the identified human values are Self-direction and Security. Differently, when an informal caregiver expects to be informed about managing the health of an older adult she/he is taking care of, the identified human values are Power and Benevolence. The explanation is that if a caregiver is informed about the health of the older adult, he/she can be helpful and responsible, but for giving good care, he/she also needs to have power over people and resources. Similarly, the emotional goal Empowered is related to the same role Older Adult, but has been mapped to partially different human values due to the different functional goals Manage health and Socialize. Therefore, we can conclude that emotional goals solely do not determine human values, and functional goals and roles influence the results, defining the context for the emotional goals. Other two reasons for variability in the identification of human values were stated in Section 2. First, human values are not discrete and rather correspond to a continuum of motivational goals. Second, some variability in the identification of human values can be attributed to the instantiation of a value in a particular context.
With respect to the research question RQ2, we analyzed human values by value categories and value items shown in Table I. The results of our analysis demonstrate that almost 45% out of 79 value occurrences from 9 distinct value categories belong to the value categories Security and Self-direction with the respective occurrence frequencies 20 and 15. The least apparent value categories in our results are Universalism and Conformity, with the respective occurrence frequencies 2 and 4. We found that the value items Sense of belonging and Healthy were the most frequent value items. Since our industrial case study is concerned with improving the health and quality of life of older adults, the most frequent role appearing in motivational goal models is Older Adult. The most frequent value items and role indirectly validate our results, because improving sense of belonging to a community and health of older adults are among the main objectives of the Pharaon project.
The granularity of value items included by the Schwartz theory of human values can be utilized to better inform the pilots and practitioners of the Pharaon project. For example, Security is the most prevalent human value category identified in the analysed pilots, where the most common value items are Sense of belonging and Healthy. This information can be helpful for developers and practitioners when making design and development decisions. For example, they can consider human values when creating user stories to improve user experience and acceptance by main end users of the Pharaon ecosystem – older adults – and their caregivers. The identified and embedded human values should also be validated with end users, which we were not able to do yet because of the current stage and large scope of the Pharaon project. Overall, all three pilots share 7 out of 9 human values, but these values are not evenly distributed due to different goal models which have been created by different pilots. Therefore, the presence of human values across different pilots can not be generalized.
Moreover, our findings can and should be extended to developing other similar platforms of digital healthcare and well-being services for older adults and the related stakeholder groups. Our findings indicate that the usage of motivational goal modelling for eliciting and representing early requirements enables to consider from the very start of the software development process human values important for older adults and other stakeholders, such as Security and Self-direction.
Although our results demonstrate that it is possible to identify human values in goal-based requirements, we could establish this only for motivational goal models [13–15, 17]. Identifying human values in other kinds of goal models, such as i* [57, 58] or KAOS [19, 59] goal models need more research.
The possibilities to identify human values in software development artefacts open up several directions for future work. For example, it opens up the possibility to compare the level of attention given to human values in different projects. It also enables an opportunity for automation of identifying human values that would allow for larger-scale studies. Another direction can also be investigating the presence of human values as a part of the quality assurance of software.
5.1 Threats to Validity
A threat to construct validity arises from the decision to use the Schwartz theory of human values [1]. There are also other theories of human values [2, 25, 27] that could be used in place of the Schwartz theory. However, none of these theories has been applied in software engineering, while the Schwartz theory has already been widely used in software engineering [3, 6, 11, 28] and requirements engineering [5, 12, 29, 30]. Another possible threat to construct validity is the definition of human values; they might have been vague for the analysts. We mitigated this threat by referring to the literature on the Schwartz theory of human values and its applications in software engineering and requirements engineering during each analysis session. This helped to create a basic understanding of the human values’ theory. Moreover, we undertook detailed discussion sessions that were moderated by a social scientist who is an expert in the Schwartz theory of human values, thus helping in cases that were initially missing the social science perspective.
Another threat to internal validity involves the analysis process, as mapping emotional goals to human values can be subjective and error-prone. For mitigation, we performed an iterative analysis process, consisting of the trial session, mock session, and individual analysis sessions. Furthermore, we undertook a detailed discussion session between the analysts and moderator, who is an expert in the Schwartz theory of human values, to resolve disagreements and conflicts, and finalise the human values only after reaching a mutual agreement. This process reduced the number of errors due to misinterpretations.
A threat to external validity is the generalizability of our results due to the Pharaon project that we chose for our study. Our project is focused on the domains of healthcare and well-being and is limited by the size of the data available to us. We mitigated this threat by analyzing three pilots from three different countries However, we acknowledge that it may not be possible to generalize our findings to other domains or even to a similar domain. More research needs to be conducted to explore this. We have provided the dataset4 that includes the goal models for the study’s replicability. However, human value is subjective, which can cause different results. However, if other researchers follow the process we provided in Fig 3, they should get comparable results.
Related Work
The foundation for representing human values through emotional requirements is discussed in [13], where emotional requirements are represented as quality goals and further elaborated in [14], where the agent paradigm was introduced for requirements elicitation and representation. Emotional requirements were explicitly introduced and exemplified by the case study of emergency alarm systems for older adults in [15]. The same case study was extended by elaborating emotion-oriented requirements engineering in [23], based on [13] and [15]. The relevance of emotional requirements in healthcare was demonstrated by two case studies in [16]. However, none of these studies applies any existing theory of human values to map emotional goals to human values by multiple stakeholder groups.
The study [60] provides a framework for capturing consumer preferences in terms of human values and mapping them to softgoals of the Tropos methodology [58]. The article [61] proposes to extend soft-goals of the Tropos methodology with contexts representing “human-centric aspects of end-users” that are associated with tasks contributing to achieving the soft-goals. The work [5] presents a method for reasoning about human values in software systems by means of goal models of the Tropos methodology where goals represent human values and tasks – features of software systems. The paper [30] puts forward a method and notation for representing personal values and relating them to motivations modelled as quality goals and emotions captured as emotional goals of motivational goal models. The work [62] proposes to link organisational values like gender-inclusiveness with personal human values through emotional goals of motivational goal models, such as Valued, Supported, and Inclusive. As compared to the studies mentioned above, our approach described in this paper stands out because (i) it is simple as human values are mapped from emotional goals in the context of roles and functional goals; (ii) is has a strong theoretical foundation, as the approach is based on the theory of constructed emotion [16, 40], in addition to the Schwartz theory of human values [1]; (iii) it lends itself more easily into practical applications because motivational goal models can be transformed in a straightforward manner into detailed requirements in the form of user stories [52, 53], as has been demonstrated in the Pharaon project.
Conclusions
Our study demonstrates that human values can be identified in motivational goal models, which is one of the artefacts in goal-based requirements engineering. From six goal models of three pilots of the Pharaon project – Italy, the Netherlands, and Slovenia – we identified human values belonging to 9 out of 10 value categories and 28 out of 58 value items based on the Schwartz theory of human values. All in all, we identified 79 value occurrences from 9 distinct value categories by Schwartz. We found that almost 45% of all the value occurrences belonged to the value categories Security and Self-direction. All three pilots shared 7 out of 9 identified value categories. We also identified the value items Sense of belonging and Healthy as the most frequent value items, both of which belong to the value category Security. We also found that analysing the emotional goals solely is not sufficient for identifying human values, as also the broader context in the form of functional goals and roles should be used.
The relevance of our work does not lie in identifying human values in motivational goal models per se but rather paves the way for including human values in goal-based requirements so that they could be elaborated into detailed requirements and software systems preserving human values embedded in them. The corresponding recommendation for practitioners arising from our work is to use motivational modelling to represent early user requirements, because, as we have shown in our paper, emotional goals included in motivational goal models can be mapped to human values, considering the context provided by the roles and functional goals. Our results indicate that any requirements engineer working on the project at hand who has familiarized herself with the Schwartz theory of human values is able to perform this kind of mapping. Moreover, to make the inclusion of human values in early requirements explicit, we plan to complement motivational goal models by dedicated “value goals”, whose enumerated types would correspond to the conceptual definitions of values in terms of motivational goals [32]. Such “value goals” would be attached to the relevant functional goals of a hierarchical goal tree, similar to roles, emotional goals, and quality goals. Early requirements represented by motivational goal models can then be elaborated into scenarios and user stories, preserving the human values captured by motivational goal models. We will demonstrate a methodology for doing this as an important part of our future work. Another important area of our future work will be to include quality (i.e., non-functional) goals in the analysis of human values. We intend to expand our exploration beyond ten value categories by Schwartz by incorporating identifed software-related values. This approach aims to broaden our perspective and enhance the relevance of our work in the field of software engineering. Additionally, we aim to include in the analysis the remaining three pilots of the Pharaon project and explore similar projects within our scope.
ACKNOWLEDGMENT
The research work presented in this paper has received funding from the Pilots for Healthy and Active Ageing (Pharaon) project of the European Union’s Horizon 2020 research and innovation programme under the grant agreement no. 857188 and from the European Social Fund via the IT Academy programme.