Equality Requirements for Software Systems: A Survey Maryam Ali Al Hinai Ruzanna Chitchyan Information Technology Department Department of Computer Science Higher College of Technology University of Bristol Al-Khuwair, Oman Bristol, UK Email: maryam.alhinai@hct.edu.om ORCiD: 0000-0001-6293-3445 Email: r.chitchyan@bristol.ac.uk Abstract—In order to support social sustainability through assets by increasing their longevity [7]. Yet, at present very software systems, we must learn to engineer requirements that little work has been published on how to engineer social foster sustainability conditions within a society. Yet, the notion of sustainability through software [8]. Given that, as noted above social sustainability is very complex, encompassing equality, trust, cultural and religious diversity support, community participation [1], social sustainability encompasses a large set of complex and more. In this paper we present results of a survey-based characteristics (e.g., from cultural to religious and community study on the notion of equality within a range of software users. interactions, governance and trust) tackling this challenge Do diverse users converge to some common views on equality? would require addressing a single social sustainability char- Does the diversity of users itself influence these perceptions. And acteristic at time. Thus, in this paper we explore the equality can we formulate requirements statements for these perceptions? We explore these questions through analysis of data from 155 characteristic of social sustainability. (relatively well educated, English speaking, and technologically The key contributions of this paper are two-fold, as it literate) survey respondents. explores: Index Terms—Social sustainability; software; requirements engineering; equality, survey. • If there is an overall agreement across the wider software user community on what requirements are relevant for I. I NTRODUCTION engineering equality through software, and • If demographic characteristics of individuals are cor- A society is socially sustainable if it has a sound basic related with their perceptions of the relevance of the framework that supports its members cooperation at low trans- software equality requirements. action costs [1]. Shared values, equal rights, and community, religious and cultural interactions are necessary components To realise the above two contributions, the paper first of such a framework [1]. What this really means, is that discusses related work to equality and its support through in a socially sustainable society individuals and groups in- software in section II. Then presents an overview of how a teract (e.g., trade, borrow and lend, innovate, learn, govern, number of equality requirements statements (which a software and regulate, etc.) with the maximum personal and social system should enforce) have been identified from several benefit and minimum expenses, lost opportunities (e.g., due previously and independently specified software requirements to discrimination or corruption, etc.) or discomfort, i.e., with documents. Using these requirements statements, the present minimum transition costs. research has developed a survey instrument, which has been used for data collection and analysis that help us deliver the As software mediates more and more activities in the said contributions. The research methodology used in this modern societies, many software engineering researchers argue survey design and analysis is detailed in section III. The results that it also has a key role to play in supporting social are presented in section IV, with threats to validity discussed sustainability. Thus, some researchers [2], [3] argue that values in section V. The paper is concluded with section VI. drive decision making in software engineering. Consequently, these values would be embodied within software implemen- II. E QUALITY R EQUIREMENTS tations and would drive the way that software operates and structures social interactions around itself. Similarly, Becker Equality is defined as the right for all members in a et al. [4] note that sustainability values should be explicitly society to enjoy living and getting access to services and and intentionally engineered as requirements into software facilities without being discriminated because of their origin, requirements specifications, and subsequently implemented in believes, position, or (dis-)abilities [9]. It is an internationally the resultant software systems. advocated value that several standards and acts endeavour to Indeed, several pieces of research demonstrate how to instill. For instance: (i) the ISO 26000 [10] aims at helping integrate environmental sustainability concerns into software organisations to install such principles as gender equality system, e.g., by reducing resource consumption, and fostering and fair treatment; (ii) Social Accountability Standard 8000 reuse [5], [6], or improve technical sustainability of software [11] states guidelines on child labour, forced or compulsory labour, health and safety, freedom of association and collective III. R ESEARCH M ETHODOLOGY bargaining, discrimination and alike; (iii) Equality Act 2010 [12] aims at reducing socioeconomic inequalities, harmonising In order to elicit how the software users perceive equality equality laws and tackling discrimination and harassment due requirements for software systems, we chose to collect data to individuals’ personal characteristics. through a survey [22]. We followed survey design and analysis guidelines devised by Kitchenham and Pfleeger [23] and Van Equality is often considered from two seemingly contradic- Selem [22]. tory positions: The key objectives of this study are three-fold: • one (let’s call it equal equality) taking the viewpoint that • to discern a prioritised ranking of requirements related since all members of a society should be equal, they to equality, as perceived by the general software user must all receive the same treatment and equal access to community; • to investigate the effect of such demographic factors as resources [13]. If anyone is not treated equally, it raises unfairness complaints. users’ expertise, education, gender and religion on their • the other (which we’ll call unequal equality) interpre- perception of given equality requirements; • to observe if there is an overall agreement on what tation [13] suggests that since members of society differ, equality means providing different treatments to constitutes equality within the general software user com- accommodate the diversities of groups and individuals. munity. Here, “... failing to provide different treatment is itself unequal because of the unequal social and economic A. Eliciting Equality Requirements Statements position of different groups. . . . in this position, diversity To construct a survey, we first needed to identify equality enhances equality by ensuring that unequals are treated requirements statements relevant to software systems. For this differently and unequally.” we applied the equality requirements pattern and template pro- posed in [24] on software systems requirements specification A number of researchers take the equal equality view and documents1 which were developed previously by independent discuss equality in terms of equal resource allocation e.g., requirements engineers from across a variety of application for equal water [14] and housing [15] resource distribution domains. Most of these documents were selected as they have between members of society. They stand for removing race, been previously used as benchmarks in other requirements and age, religious [16] and gender [16], [17] discrimination, as software engineering-related studies by other researchers. The well as inequalities due to geographical location of society documents are sourced from reputable research or software members. The infrastructural access inequalities [18] could be practitioner organisations, are written in English and well reduced by enabling equal access to electronic resources [18], structured. as well as improving access to community (e.g. child care) A number of recurring equality-related requirements state- and social (e.g. cultural events) infrastructure [15]. ments were identified. These are general statements that are When ICT is concerned, Web Content Accessibility Guide- applicable to different software system. This was done to lines (WCAG) [19] aim to ensure that web content is ac- avoid users’ familiarity problem with a specific software cessible, supports equal access and opportunity for people system or domain. It is worth noting that we do not claim with disabilities [20], older users, people in rural areas and that the statements listed in Table I are the complete and developing countries. full requirements for the realisation of equality concern; this simply is a relevant sub-set obtained through application of the Those who advocate the unequal equality, argue that equal equality value patterns method [24] to a set of requirements distribution does not always entail fairness and it is fair, rather documents. than equal distribution and access to resources that is essential These requirements statements were combined with a few for social sustainability [14]. Thus, those with greatest need additional statements, that do not directly relate to equality should be provided with more resources (e.g., babies should (as presented in Table I and discussed in section III-B) and get more milk than adults), those who invest more should get were used for a survey-based study to observe the perceptions more back [21] (e.g., if one individual works twice as hard as of the software users on equality and relevance of equality the other, she should get twice as much pay), etc. requirements. While each of the above equality-related viewpoints have valid philosophical grounds, we, as software engineering re- B. Survey Design searchers, are interested in establishing if there are common re- A cross-sectional survey instrument was designed for the quirements that could be used for engineering equality through data collection purposes [23]. As we were interested in general software systems. Thus, we adopted a pragmatic approach aimed at identifying equality requirements statements relevant 1 These documents are: [14], [16], [25], [26], [17], [27], [28], [29], [30], to software systems, as detailed in the following section. [31], [15] software users’ perspectives, with varying demographics fac- on the priorities and perceptions of equality. The survey was tors (including religion), we chose to utilise an online survey concluded with a brief thank you note. [22] format (to fill it the respondents will clearly need to The first version of the survey was piloted with a small use software), which was distributed widely through different set of participants. The pilot helped to identify and rectify a communities and lists. number of concerns. Here the wording of the information sheet The online survey comprised an information sheet (with a was amended, several grammatical mistakes were corrected, participation consent form to be confirmed by the respon- and the demographic information section was moved from the dents), followed by three self-administered data collection beginning to the end of the survey. Thereupon, the final version sections. of the survey was published through a web survey tool2 . Section 1 of the survey consisted of 21 previously selected C. Population and Respondent Sampling requirements statements (as discussed in section III-A and Since we are interested in the views of the general software summarised in Table I). Here the respondents were asked to users (i.e., anyone who uses any kind of software), the targeted evaluate the importance of each of the given statements for population of this study is potentially counted in more than its’ relevance as a software system’s equality requirement. millions3 , with a similarly large numbers of online software The relevance was indicated as irrelevant (NAI) (see Table I, users [34]. This lead us to choosing sampling methods that for statements that do not relate to equality), neither relevant reach a diverse range of respondents and communities. Thus, nor irrelevant (Table I, NINU) to indicate the respondent’s the unrestricted sampling [22] and convenience sampling [23] indecision or lack of knowledge on the relevance of a given methods were used by publicising the survey questionnaire statement, and relevant, where relevance was subcategories through LinkedIn, Instagram, Facebook, Twitter, ResearchGate into slightly relevant (SI), relevant (IE), and very relevant and WhatsApp. (VIE) options. Additionally, invitation emails were sent to the students and Because this is an importance scale, we did not aim to staff list at the University of Leicester, UK, and, through aca- produce balanced levels. Knowing that a statement is not demic colleagues, at other counties (such as USA, Brazil, Ger- important to equality is more relevant than knowing the degree many, and Oman). To encourage varied religious backgrounds or depth of unimportance. In the case of unimportance, the representation, we also explicitly reached out to colleges from direction is what we are looking for and not the depth. On Omani universities through randomly selected staff members the other hand, knowing the degree of importance can help emails available online, asking them to distribute the partici- requirements engineers to prioritise equality requirements and pation request locally. As we requested that the colleagues and decide which should be included in the first release and respondents forward the participation request to all who they which to be kept for later releases taking into consideration considered may have been willing to respond to the survey, resource availability (e.g. time, money, skills, etc.). In this we also incorporated elements of snowball sampling (which case, “discrimination . . . between the positive scale positions” is also an unrestricted and convenience sampling method) [23] is important [32]. into the data collection. To identify how well the respondents distinguish the notions D. Data validation and analysis of equality form other requirements relevant to social sustain- ability, we mixed into the survey a set of statements that did We consider a response valid if all the set questions not relate to equality. These are statements 1, 4, 5 and 17 in have been completed. Clearly, any respondent would have Table I where: completed the survey through use of software, and this (i.e., software use experience) was the only fundamental qualifying • statement 1 is a security requirement; constraint for participation in this survey. Thus all completed • statement 4 is a performance requirement; responses would be valid. • statement 5 is a robustness requirement; To analyse the data, we used frequency analysis to describe • statement 17 is an availability requirement. the importance of the equality requirements statements as Section 2 of the survey consisted of two questions aimed well as inferential statistical analysis methods to discern the to explicitly elicit the respondents’ notion of equality in terms influence of demographic factors on the equality perceptions. of priorities they give to software profitability, usability, func- IV. D ISCUSSION OF FINDINGS tionality, and equal goal support for various user groups. Here data was collected in multiple choice options with nominal Below we first describe the set of our respondents in terms scale. of their demographic characteristics, then address the above Section 3 of the survey comprised of 6 questions for stated study objectives on equality requirements prioritisation, collection of demographic information. This section was partly perceptions, and influence of the elicited demographic factors adopted from the survey by Osho et al. [33]. Here the questions upon these. aimed to both help characterise the sample of the participating 2 https://www.onlinesurveys.ac.uk respondents, and to elicit what effect the demographics have 3 http://chrissniderdesign.com/blog/resources/social-media-statistics/ TABLE I. Requirements Statements (* marks non-equality related entries) Requirements Statements S1*: User authentication is important for supporting equality. S2: Usability of software to users from different age ranges is important to support equality. S3: Suitability of software to users from different age ranges is important to support equality. S4*: Short response time to user enquiry is important for supporting equality. S5*: Short recovery time after system failure is important for supporting equality. S6: Suitability of software to users from different genders is important to support equality. S7: Considering direct stakeholders’ goals behind using a software is important to support equality. S8: Fairly selecting which goals will be implemented in the software is important for supporting equality. S9: Ability to accommodate new types of users is important for supporting equality. S10: Multilingual interface is important for supporting equality. S11: Different information presentation formats (e.g., audio, video, text) is important for supporting equality. S12: Compatibility of software application with different operating systems is important for supporting equality. S13: Compatibility of software application with different hardware devices is important for supporting equality. S14: Availability of software’s usage guidance (e.g., help, tutorials, and tips) considering users with no/little prior knowledge of this software is important for supporting equality. S15: Availability of software’s shortcuts to accomplish tasks for experts and fast learners is important for supporting equality. S16: Availability of software application on different web and mobile platforms is important for supporting equality. S17*: Availability for use 24 hours per day, 365 days per year is important for supporting equality. S18: Allowing stakeholders to equally access software services to achieve their goals is important for supporting equality. S19: Suitability of software for users from different religious beliefs is important for supporting equality. S20: Accepting information from different media (e.g., voice, text, braille) is important for supporting equality. S21: Considering indirect stakeholder goals that are affected by the software is important for supporting equality. A. Respondents profile PhD and 36.1% Masters degrees (see Table II). There were no unschooled respondents. Thus, our sample is clearly biased The respondents’ sample was nearly evenly balanced for towards highly educated software users. This, in itself, is not gender, with a slightly higher female participation (by 1.2%), entirely surprising, as the topic of the survey (software and as illustrated in Table II. The average age of the participants equality) as well as method of data collection already presumes was 37, with the age characteristic also well distributed in 18 some minimum education and technological literacy levels. to 64 years old range, but only one over 65 year old participant. The vast majority of participants were in employment or B. Frequency of Equality Statements education (see Table II). Only 1.9% of the respondents were novice software users and the majority (45.8%) were of an The ordering of the responses on the priority of equality advance proficiency (as per Table II). statement for software systems is summarised in Table III and demonstrated in Figure 2 with a divergent stacked bar [35]. The largest group (just over a half) of the respondents report to be of Muslim background, which could have arisen due As Figure 2 shows, statements S10: Multilingual interface, to the previously mentioned more direct invitation of Omani S14: Software’s usage guidance (e.g., help, tutorials, and participants. Christians and Hindus are the next two larger tips), S11: Different information presentation formats (e.g., religious groups (see Figure 1). audio, video, text), S20: Different input support formats, and S3: Support for users across various ages, are the highest ranked overall relevant stamens (i.e., highest combined slightly relevant, relevant, and very relevant). All these statements also have consistently low irrelevance and indecision ranking. Which clearly indicates that there is an agreement (though not unanimous) across the respondents that these statements are closely related to equality. Similarly, respondents have consistently marked out the deliberately introduced unrelated statements (i.e. S1, S4, S5 and S17) as both the least relevant, most irrelevant, and the most doubted (i.e., neither relevant, nor irrelevant) subset. However, here too, the respondents were not unanimous in their ratings, as some respondents have ranked S17 as having Fig. 1. Respondents Religion, N = 154 something to do with equality. S17 is a statement on software availability and may have been perceived as tangentially rele- The highest education level of the respondents is rather vant to equality by providing users with access at time of their biased towards the highly qualified end, with 34.2% holding convenience, without restrictions. Despite these discrepancies, TABLE II. Respondents Profile Background No (Valid %) Male 76 ( 49) Gender Female 78 ( 50.3) Undisclosed 1 ( 0.65) 18 to 24 years 21 (13.5) 25 to 34 years 40 (25.8) 35 to 44 years 57 (36.8) Age 45 to 54 years 29 (18.7) 55 to 64 years 7 (4.5) Age 65 or older 1 (0.6) Employed 115 (74.2) Student 36 (23.2) Employment Unemployed 4 (2.6) Retired 0 (0) PhD (or equivalent) 53 (34.2) Masters Degree (or equivalent) 56 (36.1) Highest level of education Undergraduate (or equivalent) 24 (15.5) A college degree (diploma and equivalent) 21 (13.5) High school degree or less 1 (0.6) Novice 3 (1.9) Intermediate 28 (18.1) Level of software use proficiency Advanced 71 (45.8) Expert 53 (34.2) Fig. 2. Statements ranks the overall correct and consistent ranking of the least relevant Considering the topic ranking, for the genuinely equality- statements, indicates that there is a general agreement among related statements list (i.e., with exclusion of S1, S4, S5 and the survey respondents on the statements that do not relate (or S17) we note that the highest importance is attributed to weakly relate) to equality. statements supporting interaction of the users with the software TABLE III. Statements Frequencies On the other hand, statement S15 might have been rated Statement NAI NINU SI IE VIE least relevant because the extra service of supporting experi- S1* (sec) 18 26 18 45 48 (11.6%) (16.8%) (11.6%) (29.0%) (31.0%) enced users was thought of as more related to usability than S2 8 (5.2%) 10 22 58 57 equality. It could also be that the extra functionality is viewed (6.5%) (14.2%) (37.4%) (36.8%) S3-1 4 (2.6%) 11 25 53 61 as a privilege given only to expert users. However, having (7.1%) (16.2%) (34.4%) (39.6%) extra features to for experts does not hinder novice users from S4* (per) 6 29 32 41 37 (10.3%) (18.7%) (20.6%) (26.5%) (23.9%) accessing same features in simpler but potentially more time S5-1* (rob) 20 21 26 39 48 consuming ways. (13.0%) (13.6%) (16.9%) (25.3%) (31.2%) S6-1 9 (5.8%) 11 20 53 61 (7.1%) (13.0%) (34.4%) (39.6%) C. Equality Goals S7 11 13 31 55 45 (7.1%) (8.4%) (20.0%) (35.5%) (29.0%) As noted before, the questions in section two were aimed S8-2 6 (3.9%) 14 22 64 47 (9.2%) (14.4%) (41.8%) (30.7%) at understanding what concerns drive the notion of equality S9 2 (1.3%) 15 18 61 59 for software systems in the respondents: from profit, to us- (9.7%) (11.6%) (39.4%) (38.1%) S10-2 3 (2.0%) 5 (3.3%) 19 55 71 ability, functionality, and user priorities. While Question2 4 (12.4%) (35.9%) (46.4%) asked to make a choice for the key equality-conducive goals, S11 4 (2.6%) 7 (4.5%) 19 53 72 Question35 asked which user groups should be supported. (12.3%) (34.2%) (46.5%) S12 6 (3.9%) 12 25 53 59 The results indicate that our respondents are relatively (7.7%) (16.1%) (34.2%) (38.1%) S13 6 (3.9%) 13 23 56 57 equally distributed in prioritising support for specific group (8.4%) (14.8%) (36.1%) (36.8%) goals, provision of same but equal functionality to all users S14 6 (3.9%) 5 (3.2%) 18 53 73 (11.6%) (34.2%) (47.1%) (without distinction), and simple usability of software as the S15 9 (5.8%) 16 33 60 37 key drivers of enabling equality through software. (10.3%) (21.3%) (38.7%) (23.9%) S16 6 (3.9%) 10 26 53 60 For Q2 the largest group of respondents (37.7%) said (6.5%) (16.8%) (34.2%) (38.7%) that to support equality, the most prioritised goals for each S17* (av.) 12 18 32 47 46 (7.7%) (11.6%) (20.6%) (30.3%) (29.7%) group should be integrated into software. This suggests that S18 4 (2.6%) 12 27 51 61 most respondents consider that equality is supported through (7.7%) (17.4%) (32.9%) (39.4%) S19-3 5 (3.3%) 13 18 50 66 looking at the different stakeholders groups and finding the (8.6%) (11.8%) (32.9%) (43.4%) best way of allowing them to achieve their goals through the S20 5 (3.2% 10 26 57 57 ) (6.5%) (16.8%) (36.8%) (36.8%) software system. In Q3 72.4% of this respondent group had S21 5 (3.2%) 20 39 65 26 consistently noted that goals of different stakeholder groups (12.9%) (25.2%) (41.9%) (16.8%) Note: * marks statements that are not directly related to equality; S1 relate should be implemented to support equality. Thus, this group of to security (sec), S4 relate to performance (per), S5 related to robustness respondents considers equality as equitable support of various (rob), and S17 on availability (av) goals of the different user groups for a given software system. Another large group of respondents (31.8%) for Q2 said (S10, 14, 11, 20), followed by the mixed order of topics that to them the best way to enable equality through software related to user variability support (e.g., age, technical platform is by providing same functionality at same level to all user support, etc.). The statements that support stakeholder goals groups. For Q3, the majority (69.4%) of this respondent group (S8, 7, 21) come at the bottom part of the list. 4 Statement of Q2: To support equality in a given software application, we Out of 17 genuine equality-related statements, the least should implement the goals that • Deliver the highest profit importantly ranked was S21 (Considering indirect stakeholder • Provide best usable interface goals) and S15 (availability of softwares shortcuts). This, could • Provide same level of functionality to all user groups be explained with the fact that the key focus of software is nor- • Implement the most prioritized goals for each user group. mally placed on direct software users, with indirect stakehold- 5 Statement of Question 3: An online shopping software has several groups ers considered thereafter. Clearly, the long-term cumulative of users. Normal users are regular users who use the software to perform effects of a software system could dramatically affect indirect basic shopping through the website. Gold users are those who buy expensive stakeholders. For example, the long-term use of Amazon.com products that generate high income for the business. Special users are those who use the application with special request of adjustment to their disabilities by large number of individuals has gradually undermined (colour blindness, hearing loss, etc.). Each group of users have different goals many physical retail shops, and their located environments to be implemented by the online shopping application. To support equality and communities (e.g., empty shops in town centres leading to through the online shopping application, which goals will you select to be implemented in the application. unattractive social spaces). Yet, normally software users first of • The normal users goals The gold users goals all focus on the direct effect of their immediate interaction with • The special users goals None of them the software system (e.g., ability to obtain the desired book at • The normal and special users goals lower price from an online bookstore). This tension between • The normal and gold users goals • The special and gold users goals priorities of direct and indirect stakeholders is, indeed, one of • All of them the characteristics of social sustainability concerns. also chose that goals of different stakeholder groups should be percentage (more than 20%) of cells with under 5 count. supported. Thus, this group considers equality as delivery of Consequently, the Fisher’s exact test [37] is preferred to the the same service at same level to all user groups for a given chi-square. Additionally, since the compared variables are at software system. more than 2 categorical levels (e.g., employment status has The third largest group of respondents (27.9%) for Q2 said 4 options: employed, student, unemployed and retired), the that provision of the best usable interface is the key driver Fisher-Freeman-Halton Test6 was undertaken. for equality. In response to Q3, 50.0% of respondents in this Where the relationships between age factor and Q3’s user sub-group indicated that goals of all user groups should be group selection was studied, we utilised the Monte Carlo implemented. It could have been expected that those concerned method as an alternative to Fisher’s test (to resolve the issues with the usability of the software would care more about the of high memory intensity posed by the exact tests [38]). users with disabilities (who could face more difficulties with The test results are detailed in Table IV, from which the software use). However, only 7.1% from this sub-group we observe that the probability of dependence between the indicated the specialist (disadvantaged) user groups as those background variables and goal priorities is negligible for who should be especially supported via the software system all demographic characteristics but that of software usage to enable equality. proficiency. Thus, the above stated H0 hypothesis has to Furthermore, a small group of respondents (2.6%) consid- be accepted for the majority of demographic characteristics ered profit maximisation as a key direction to enable equality (where p-value is over 0.05 as per Table IV), but rejected for in Q2, and for Q3 half of this sub-group chose prioritising the software usage proficiency factor. requirements of the “gold user” group (i.e., those who buy Thus, we observe that in our sample of respondents, the expensive packages of service for a given software) as the key decision on which groups’ goals should be implemented (as in achieving equality, while the other half chose to support all per Q3) in a software system to support equality is related to user groups. respondents proficiency in software usage: In summary, while (as found from Q3 responses) the vast • Half of the novice respondents said that equality would be majority of respondents (64.7%) think that goals of all user achieved by supporting goals of all types of user groups, groups should be supported, the perception of the way that but the other half though that support of the “gold users” equality should be delivered with software system is varied: goals would lead to better equality. just over one third of survey respondents noted the need to • The majority of the intermediate, advanced and expert support goals that each user groups prioritises; another third users considered that goals of all types of groups should focused on equal service level to all - whichever service that be supported (37.0%, 70.4% and 71.7% respectively). may be and to whomever it is delivered; and the other (slightly less than) third portion of respondents underlined the usability This could be explained by more experienced users having had of software as key to its support for equality. more opportunities to participate in software system use as part of different groups (e.g., service provider or consumer (e.g., D. Impact of Background Factors seller and buyer), viewer, support team member, developer, or To study what impact background factors may have on an administrator, etc.). We expect that such an experience of the perception of equality, we consider if and how the de- broader exposure would, naturally, underline the relevance of mographic factors affect both equality goal priorities (i.e., multiple groups of system users (though presently we do not responses to Q2 and Q3) and ranking of equality-related have sufficient data to substantiated this causality opinion). statements. 2) Impact on Statement Rating: Inference analysis for the 1) Impact on Equality Goal Priority: We start our study of role of the background factors on the rating of the statements the demographic factors impact by formulating a null hypoth- not-related to equality (i.e., S1, 4, 5, 15, 17) shows that the esis that the background variables (Q4-9) and equality goals rating of the statements 1, 17 and 15 is not affected by the (Q2 and Q3) variables are independent, with the alternative background factors of the respondents. However, we observe hypothesis that they are indeed dependant: a relationship between religion of the respondents and their rating of the statements for S4: Response time and S 5: • H0 : Background variables and equality goals variables Recovery time (at p-value = 0.047 which is under 0.05 for are not related. S4 and p-value = 0.008, which is under 0.01 for S5). • Ha : Background variables and equality goals variables are related. Many Muslims participants (31.0%) rated statement 4 as very important to equality. Many Christian (24.1% ) and Given that all variables for respondents background (ques- Hindu (44.4%) participants found it important to equality, tions 4 to 9) and equality goals (questions Q2-3) are cat- and a large number of those with other beliefs (38.5%) rated egorical (i.e., nominal and ordinal), chi-square test would this statement as slightly important to software equality. normally be performed [36] to examine the relationships Finally, many of the respondents who did not disclose their between these variables. However, when a cross-tabulation of these two variables is created, the data contains a high 6 http://www-01.ibm.com/support/docview.wss?uid=swg21479647 TABLE IV. Equality Goals and Background Factors Background factor Equality Goal Results Interpretation Q2 FET=1.577, P=0.677 No relation Gender Q3 FET=5.451, P=0.651 No relation Q2 FET=14.992, P=0.479 No relation Age Q3 FET=41.617*, P=0.280 No relation Q2 FET=15.314, P=0.435 No relation Religion Q3 FET=29.912, P=0.951 No relation Q2 FET=10.566, P=0.655 No relation Education Q3 FET=37.014, P=0.157 No relation Q2 FET=4.290, P=0.666 No relation Employment Q3 FET=14.511, P=0.431 No relation Software Q2 EFT=10.474, P=0.286 No relation proficiency Q3 FET=31.563, P=0.048 There is relation *Monte Calro estimates using 10000 sampled tables was used religious beliefs (44.4%) rated the statement as neither im- via paper and online dictionaries, and translators), we believe portant nor unimportant. this was a reasonable choice to make. In addition, we carried For statement 5, many Christian (34.5%) and Muslim out pilot for the data collection to improve both questions and (36.1%) participants found it very important, and most Hindu statements wording and clarity, as well as structure. participants (38.9%) rated it as important to equality . B. External Validity Participants with other religious backgrounds scored a tie between neither important nor unimportant and slightly A threat to external validity can arise if the respondents are important to software equality with 23.1%. While 33.3% not representative of the population [41]. This could have been of those who preferred not to disclose their religion rated influenced by the used sampling methods [41] (see section statement 5 as not important. III-C). To ensure that our respondents were representative of We observe that both of these statements are related to different countries, religions, and ages, we posted the survey time, and previous research has demonstrated that there is a request internationally and across various age-groups. correlation between time valuation and cultural factors [39] Yet, since we used distribution methods (e.g., LinkedIn, [40]. Thus, it is likely that this relationship is a demonstration Research Gate, etc.) accessible to us, it is likely that some of such a cultural, time-related influence. populations with very different views and software use meth- ods would not have been reached. Indeed, we have reported V. T HREATS TO VALIDITY that the respondents are rather over-educated compared to To discuss validity of findings, we consider internal and the expected average set of software users, as the request for external validity factors. participation was widely posted through university lists and personal requests to university academics. A. Internal Validity Moreover, due to the used distribution and data collection The study design has a large role to play in ensuring that format, it was not possible to calculate the response rate. the results correctly convey the information contained in the Our respondents are those who volunteered to participant due study data. With this respect, internal validity of this study to some personal interest; clearly not all who got/read the could be threatened if the statements in Q1 and goals in participation requested have completed the survey. Q2 are poorly related to the equality concern. Though this Thus, we must note that the generality of results presented threat cannot be fully eliminated we have mitigated it as in this paper relate to the sub-section of well educated, English we have ensured that the equality statements and goals are speaking, and technologically literate software users. representative and are closely related to concerns expressed in requirements specifications from several independently defined VI. D ISCUSSION AND C ONCLUDING R EMARKS software system requirements documents. In this paper we presented a survey-based study on percep- Although we discussed in Section III-B the reasons of tions of software users on equality. We investigated if the wider having an unbalanced scale, we note that this could pose a community of software users has a set of “generally agreed threat to validity if it were to mislead respondents. upon” equality goals and equality requirements ratings, that We have also used English language survey, which has been should be supported via software. We also looked at whether filled in by participants from other countries, who are likely to users’ background characteristics affect they perceptions of be non-native English speakers. Thus, it is possible that some equality. respondents may interpreted some of the statements differently The sample of the respondents to our survey is some- that they were intended. Yet, as English is the most widely what skewed, as the respondents are not equally distributed used language for academic research and publications, and the throughout the world regions, and are mainly well educated, participants would have had access to transition support (e.g., English speaking, and technologically literate software users. The findings here are thus related to this particular sample. Our [3] M. A. Ferrario, W. Simm, S. Forshaw, A. Gradinar, M. T. Smith, and respondents demonstrated a clear and nearly equal split in their I. Smith, “Values-first se: Research principles in practice,” in Proceed- ings of the 38th International Conference on Software Engineering perception of equality goals: roughly a third of them considers Companion, ser. ICSE ’16. New York, NY, USA: ACM, 2016, pp. equality in terms of equal distribution (what we called equal 553–562. equality in previous discussion), another third perceives it in [4] C. Becker, S. Betz, R. Chitchyan, L. Duboc, S. M. Easterbrook, B. Penzenstadler, N. Seyff, and C. C. Venters, “Requirements: The key terms of unequal equality whereby each group needs to be to sustainability,” IEEE Software, vol. 33, no. 1, pp. 56–65, Jan 2016. treated differently by supporting their own priorities and needs [5] M. Mahaux, P. Heymans, and G. Saval, “Discovering sustainability to achieve their own goals. Finally, the last third perceives requirements: An experience report,” in Requirements Engineering: Foundation for Software Quality: 17th International Working Confer- software equality in terms of more narrowly scoped set of ence, REFSQ 2011, Essen, Germany, March 28-30, 2011. Proceedings, accessibility requirements. While all these topics form part of D. Berry and X. Franch, Eds. Springer, 2011, pp. 19–33. the present equality discourse, this nearly equal split of pri- [6] J. Cabot, S. Easterbrook, J. Horkoff, L. Lessard, S. Liaskos, and J.- N. Mazón, “Integrating sustainability in decision-making processes: A orities was somewhat unexpected. Furthermore, the majority modelling strategy,” in Software Engineering-Companion Volume, 2009. of respondents also indicated that, in order to be conducive ICSE-Companion 2009. 31st International Conference on. IEEE, 2009, for equality, a software system should support the whole wide pp. 207–210. [7] C. Becker, “Sustainability and longevity: Two sides of the same quality?” variety of its user groups. in Proceedings of the Third International Workshop on Requirements For the software engineering professionals this means that Engineering for Sustainable Systems, RE4SuSy 2014, co-located with 22nd International Conference on Requirements Engineering (RE 2014), in order to engineer a software system that is perceived to Karlskrona, Sweden, August 26, 2014., 2014, pp. 1–6. be conducive to equality characteristic of social sustainability, [8] T. Johann and W. Maalej, “Position paper: The social dimension of that system must: sustainability in requirements engineering,” in Proceedings of the 2nd International Workshop on Requirements Engineering for Sustainable • not only provide accessibility support to all its user Systems. Citeseer, 2013. [9] “Understanding equality,” accessed: 06-08-2017. [Online]. Available: groups, http://www.equalityhumanrights.com/private-and-public-sector- • but also support unequal equality, i.e., to ensure that more guidance/education-providers/secondary-education-resources/useful- support is provided to those who need it more, more information/understanding-equality [10] “ Guidance on social responsibility ISO 26000,” Switzer- reward is provided to those who contribute more (i.e., land, 2010, accessed: 04-08-2017. [Online]. Available: groups are positively differentiated with respect to the http://www.cnis.gov.cn/wzgg/201405/P020140512224950899020.pdf goals they want to achieve and support that they need), [11] “Social accountability 8000 international standard,” New York, June 2014, accessed: 04-08-2017. • yet the equal equality is also observed, whereby de- [12] “Equality act 2010 c.15,” 2010, accessed: 04-08-2017. spite the differentiated stakeholder goals, all access to [13] M. OB́rien, “Equality and fairness: Linking social justice and social resources and services are perceived to be equal and fair. work practice,” Journal of Social Work, vol. 11, no. 2, pp. 143–158, 2011. This, clearly is not an easy task to accomplish, but if either [14] M. G. Bos, D. H. Murray-Rust, D. J. Merrey, H. G. Johnson, and W. B. Snellen, “Methodologies for assessing performance of irrigation and is not upheld, at least one third of the potential user community drainage management,” Irrigation and Drainage Systems, vol. 7, no. 4, is likely to be disappointed. pp. 231–261, Dec 1993. [15] C. Landorf, “Evaluating social sustainability in historic urban environ- Furthermore, we observe that though there is no unanimous ments,” International Journal of Heritage Studies, vol. 17, no. 5, pp. agreement on which requirement statements are most impor- 463–477, 2011. tant for equality, there is a general convergence of views that [16] A. Chib and A. Komathi, “Extending the technology-community- management model to disaster recovery: Assessing vulnerability in rural accessibility notions are paramount, closely followed by user asia,” in Information and Communication Technologies and Develop- diversity support requirements, and then differentiated goal ment (ICTD), 2009 International Conference on, April 2009, pp. 328– support requirements. 336. [17] P. R. Walsh, “Creating a “values” chain for sustainable development in There is also a generic convergence amongst the software developing nations: where maslow meets porter,” Environment, Devel- users around the notions that are not relevant to equality. opment and Sustainability, vol. 13, no. 4, pp. 789–805, 2011. [18] M. Galinium, I. P. Defindal, and I. Melissa, “E-learning system intro- Though demographic factors (such a religion and user’s expe- duction: equality in education for teachers in rural area of indonesia,” in rience) affect these perceptions significantly. 2012 Ninth International Conference on Computer Science and Software Engineering (JCSSE), May 2012, pp. 201–206. [19] W3C, “Web content accessibility guidelines (wcag) 2.0,” ACKNOWLEDGEMENTS 2017, working draft, Accessed: 03-08-2017. [Online]. Available: http://www.w3.org/TR/WCAG21/ This research is partially funded by the UK EPSRC Refac- [20] ——, “Accessibility,” 2016, accessed: 03-08-2017. [Online]. Available: toring Energy Systems fellowship (EP/R007373/1). http://www.w3.org/standards/webdesign/accessibility [21] D. R. Forsyth, “Conflict,” Group Dynamics, pp. 388–389, 2006. [22] M. Van Selm and N. W. Jankowski, “Conducting online surveys,” R EFERENCES Quality and Quantity, vol. 40, no. 3, pp. 435–456, 2006. [23] B. A. Kitchenham and S. L. Pfleeger, “Personal opinion surveys,” in [1] R. Goodland, “Sustainability: Human, social, economic and environmen- Guide to Advanced Empirical Software Engineering. Springer, 2008, tal,” Encyclopedia of Global Environmental Change, 2002. pp. 63–92. [2] B. Barn, R. Barn, and F. Raimondi, “On the role of value sensitive [24] M. Al Hinai and R. Chitchyan, “Engineering requirements for social concerns in software engineering practice,” in Proceedings of the 37th sustainability,” in ICT for Sustainability 2016. Atlantis Press, 2016. International Conference on Software Engineering - Volume 2, ser. ICSE [25] L. Zhen, S. Cao, Y. Wei, O. Dilly, X. Liu, F. Li, H. Koenig, K. Tsch- ’15. Piscataway, NJ, USA: IEEE Press, 2015, pp. 497–500. erning, and K. Helming, “Comparison of sustainability issues in two sensitive areas of china,” Environmental Science and Policy, vol. 12, (CYBER-Abuja), 2015 International Conference on. IEEE, 2015, pp. no. 8, pp. 1153 – 1167, 2009. 202–211. [26] R. Bijl, “Never waste a good crisis: Towards social sustainable develop- [34] A. Lenhart, K. Purcell, A. Smith, and K. Zickuhr, “Social media & ment,” Social Indicators Research, vol. 102, no. 1, pp. 157–168, 2010. mobile internet use among teens and young adults. millennials.” Pew [27] D. Sharma and S. Shardendu, “Assessing farm-level agricultural sustain- internet & American life project, 2010. ability over a 60-year period in rural eastern india,” The Environmental- [35] R. M. Heiberger, N. B. Robbins et al., “Design of diverging stacked ist, vol. 31, no. 3, pp. 325–337, 2011. bar charts for likert scales and other applications,” Journal of Statistical [28] K. Murphy, “The social pillar of sustainable development: a literature Software, vol. 57, no. 5, pp. 1–32, 2014. review and framework for policy analysis,” Sustainability: Science, [36] R. A. Hanneman, A. J. Kposowa, and M. D. Riddle, Basic Statistics for Practice, & Policy, vol. 8, no. 1, 2012. Social Research. John Wiley and Sons, 2012. [29] R. Jitpakdee and G. B. Thapa, “Sustainability analysis of ecotourism [37] G. H. Freeman and J. H. Halton, “Note on an exact treatment of contin- on yao noi island, thailand,” Asia Pacific Journal of Tourism Research, gency, goodness of fit and other problems of significance,” Biometrika, vol. 17, no. 3, pp. 301–325, 2012. vol. 38, no. 1/2, pp. 141–149, 1951. [30] S. A. Hosseinijou, S. Mansour, and M. A. Shirazi, “Social life cycle [38] C. R. Mehta and N. R. Patel, “Ibm spss exact tests,” IBM Corporation, assessment for material selection: a case study of building materials,” Cambridge, MA, 2012. The International Journal of Life Cycle Assessment, vol. 19, no. 3, pp. [39] K. Ishii, C. Eisen, and H. Hitokoto, “The effects of social 620–645, 2013. status and culture on delay discounting,” Japanese Psychological [31] K. R. Jones and L. Mucha, “Sustainability assessment and reporting for Research, vol. 59, no. 3, pp. 230–237, 2017. [Online]. Available: nonprofit organizations: Accountability “for the public good”,” VOLUN- http://dx.doi.org/10.1111/jpr.12154 TAS: International Journal of Voluntary and Nonprofit Organizations, [40] W. Du, L. Green, and J. Myerson, “Cross-cultural comparisons of dis- vol. 25, no. 6, pp. 1465–1482, Dec 2014. counting delayed and probabilistic rewards,” The Psychological Record, [32] I. Brace, Questionnaire design: How to plan, structure and write survey vol. 52, no. 4, pp. 479–492, 2002. material for effective market research. Kogan Page Publishers, 2008. [41] C. Wohlin, P. Runeson, M. Höst, M. C. Ohlsson, B. Regnell, and [33] O. Osho, V. L. Yisa, and O. J. Jebutu, “E-voting in nigeria: A survey A. Wesslén, Experimentation in software engineering. Springer Science of voters’ perception of security and other trust factors,” in Cyberspace & Business Media, 2012.