Human-Centred Requirements Engineering for Critical Systems: Insights from Disaster Early Warning Applications

Technical Summary: Human-Centred Requirements Engineering for Critical Systems

Problem Statement
Critical systems (e.g., healthcare, defence, disaster management) traditionally prioritize technical assurance, safety, and regulatory compliance in their Requirements Engineering (RE) processes. However, this focus often neglects the human and social contexts in which these systems operate, particularly the needs of vulnerable or marginalised user groups. The paper argues that neglecting these needs constitutes a requirements failure rather than a mere usability oversight. In safety-critical domains like disaster early warning systems (EWS), such oversights can lead to life-threatening consequences if alerts are inaccessible, incomprehensible, or unactionable for specific populations (e.g., older adults, rural users, those with low digital literacy, or colour-blind users). Current RE practices in safety-critical domains remain largely compliance-focused, with limited mechanisms for engaging diverse stakeholders or tracing human-centric needs from elicitation to validation.

Methodology
The authors propose and evaluate a four-stage human-centred RE process, applied to the domain of disaster early warning systems:

1. Requirements Elicitation: A targeted review of academic and grey literature (HCI and software engineering) was conducted to extract 62 inclusive design guidelines. These guidelines were categorized into four vulnerable communities: older adults, low-digital-literacy users, rural users, and colour-blind users. The analysis identified both group-specific guidelines and shared guidelines across multiple groups.
2. Requirements Specification: The extracted guidelines were systematically translated into a traceable catalogue of 67 software requirements (17 functional and 50 non-functional). Each requirement was assigned an identifier linking it to its source guideline and classified by type (e.g., Usability, Accessibility, Performance, Reliability).
3. Requirements Prototyping: An adaptive mobile Early Warning System (EWS) prototype was developed to operationalise the requirements. Rather than creating separate versions for each group, the prototype featured a self-customisation interface allowing users to adjust settings (e.g., font size, colour scales, navigation modes, content layout) based on their individual needs. The prototype included core EWS functions: real-time notifications, fire danger maps, and preparation guides.
4. Requirements Validation: The prototype was evaluated through two complementary methods:
   • Semi-structured interviews: Conducted with six community members (two older adults, four rural residents).
   • Cognitive walkthroughs: Conducted by eight participants using personas representing all four vulnerable groups.
   • Data was analysed thematically to assess how well the prototype addressed the specified requirements.

Key Contributions

1. A Validated Catalogue of Inclusive Requirements: The paper presents a catalogue of 67 functional and non-functional requirements specifically for inclusive disaster early warning systems. These are derived from 62 evidence-based design guidelines and cover four vulnerable communities.
2. A Traceable Human-Centred RE Process: The authors propose a replicable process that connects evidence-based elicitation, structured specification, prototype operationalisation, and participatory validation. This process makes vulnerable-user needs explicit and traceable from design guidelines to validation evidence, moving inclusivity from a late-stage concern to a core RE dimension.
3. Empirical Insights into Validation: The study provides empirical evidence on the validation of human-centred requirements. It demonstrates that 84% of the 67 requirements were addressed during validation, with 70% receiving positive validation evidence. The study also highlights trade-offs regarding adaptivity, colour customisation, and connectivity constraints.

Results

• Validation Coverage: Across the 67 requirements, 56 (84%) were addressed in the prototype evaluation, and 47 (70%) received positive validation evidence.
• Group-Specific Outcomes:
  • Elderly and Rural Users: Requirements for these groups achieved full validation coverage (100% and 80% respectively) with high positive validation rates (85% and 80%).
  • Low-Digital-Literacy Users: 81% of requirements were addressed, with 75% positively validated.
  • Colour-Blind Users: 90% of requirements were addressed, but only 50% were positively validated. The lower positive rate was attributed to usability issues with the colour customisation features, which participants found non-intuitive.
  • Shared Requirements: Requirements shared across multiple groups (e.g., Elderly + Rural + Low-Digital-Literacy) achieved 100% coverage, though some shared requirements (e.g., Rural + Low-Digital-Literacy) had lower addressed rates (20%) due to their focus on performance and connectivity, which could not be fully tested in a Figma prototype.
• Qualitative Feedback: Participants generally found the prototype's content clear and the interface simple. However, they identified specific areas for improvement, including the need for clearer map legends, more responsive feedback mechanisms for adaptive settings, and better explanations of what adaptive options do. Some users struggled with external links and inconsistent text, while others requested more dynamic, location-specific adaptivity.

Significance and Claims
The paper positions human-centredness not as an ethical add-on but as a traceable quality concern essential for the safety and reliability of critical systems. The authors claim that:

• Inclusivity drives innovation: Addressing the needs of vulnerable groups (e.g., through simplified navigation or high-contrast design) often results in broader usability improvements for all users.
• Early integration is critical: Deferring inclusivity to late-stage usability testing risks producing systems that fail the most vulnerable users in safety-critical situations.
• Traceability is achievable: The study demonstrates that inclusive design guidelines can be systematically translated into requirements and validated through participatory methods, even in safety-critical domains.
• Adaptivity requires clarity: While adaptive interfaces are valuable, they must be explainable and controllable; otherwise, they may introduce confusion or hesitation during emergencies.

The authors conclude that their proposed RE process offers a pathway for developing safe, equitable, and socially responsible critical systems, urging the field to treat human-centricity as a fundamental component of safety rather than a peripheral concern.