User Personas Improve Social Sustainability by Encouraging Software Developers to Deprioritize Antisocial Features
Bimpe Ayoola, Miikka Kuutila, Rina R. Wehbe, Paul Ralph
TL;DR
The paper tackles the problem of measuring and improving social sustainability in software engineering by testing two common tools—stakeholder maps and persona models—in a controlled, lab-based backlog-prioritization task. Using a 2x2 factorial randomized design with 79 undergraduate CS participants and a facial-recognition case, the study analyzes prioritization of prosocial, neutral, and antisocial features via Cumulative Link Mixed Models. The key finding is that persona models significantly reduce antisocial feature prioritization ($p = 0.0088$, $OR \,\approx\,0.42$), while stakeholder maps show no reliable effect and no interaction is detected. The work provides empirical evidence supporting persona modeling as a practical intervention for social sustainability and offers a replicable methodology for evaluating similar interventions in software contexts, with implications for education and industry practice.
Abstract
Sustainable software development involves creating software in a manner that meets present goals without undermining our ability to meet future goals. In a software engineering context, sustainability has at least four dimensions: ecological, economic, social, and technical. No interventions for improving social sustainability in software engineering have been tested in rigorous lab-based experiments, and little evidence-based guidance is available. The purpose of this study is to evaluate the effectiveness of two interventions-stakeholder maps and persona models-for improving social sustainability through software feature prioritization. We conducted a randomized controlled factorial experiment with 79 undergraduate computer science students. Participants were randomly assigned to one of four groups and asked to prioritize a backlog of prosocial, neutral, and antisocial user stories for a shopping mall's digital screen display and facial recognition software. Participants received either persona models, a stakeholder map, both, or neither. We compared the differences in prioritization levels assigned to prosocial and antisocial user stories using Cumulative Link Mixed Model regression. Participants who received persona models gave significantly lower priorities to antisocial user stories but no significant difference was evident for prosocial user stories. The effects of the stakeholder map were not significant. The interaction effects were not significant. Providing aspiring software professionals with well-crafted persona models causes them to de-prioritize antisocial software features. The impact of persona modelling on sustainable software development therefore warrants further study with more experience professionals. Moreover, the novel methodological strategy of assessing social sustainability behavior through backlog prioritization appears feasible in lab-based settings.