Human Error Management in Requirements Engineering: Should We Fix the People, the Processes, or the Environment?

Requirements engineering (RE) is a foundational SDLC phase involving intensive human interaction to elicit, document, and validate requirements, making it susceptible to human errors such as misunderstandings, wrong assumptions, omissions, and typos. These errors can remain latent and surface as costly failures later in development. Organizations react by attempting to mitigate effects after errors occur or by changing practices to prevent recurrence, yet systematic studies of such strategies in RE are scarce. The paper addresses the need for an organized framework of prevention and mitigation approaches and for evidence about which types are most effective. The research goal is to construct a Human Error Management Taxonomy (HEMT) and classify error prevention and mitigation strategies across RE activities. Using data from two practitioner surveys, the authors analyze strategies previously extracted and labeled (prevention/mitigation) and extend that work by systematizing strategies into high- and low-level classes and mapping them to RE lifecycle activities. Contributions: (1) an analysis of industry-used error prevention and mitigation strategies, (2) an analysis of which activities these strategies target within the RE process, and (3) a taxonomy organizing these strategies by People, Process, and Environment changes.

The background situates human error as deviations in human cognition leading to unexpected outcomes (Reason). Prior work in domains like healthcare indicates that approaches focusing on changing people (e.g., reminders, exhortations) are often ineffective; more effective solutions modify technologies, tools, and organizational policies. A classic example from WWII aviation shows redesigning controls mitigated frequent post-completion errors better than training alone. In RE, Anu et al. created the Human Error Taxonomy (HET) extending Reason’s Slips, Lapses, and Mistakes to RE contexts, providing insight into appropriate management techniques. Other SE studies include Huang’s taxonomy and DPeHE framework emphasizing knowledge/regulation training (with limited generalizability), experiments on post-completion errors suggesting process/procedure redesign over people-focused solutions, Firesmith’s catalog of RE problems and mitigation guidelines (experience-based), Lopes et al.’s expert system linking problems to error types and literature-derived solutions, and Walia & Carver’s Requirements Error Taxonomy (RET) and error abstraction work. Hu et al. mapped industrial problems to HET error types and categorized solutions as prevention or mitigation. Gaps remain: limited cognitive underpinnings in some work, generalizability/validity issues, reliance on practitioners’ ability to select strategies, and lack of a structured organization of prevention/mitigation mechanisms in RE. This study addresses these gaps by developing a structured taxonomy grounded in practitioner-reported strategies.

The study builds on two industrial datasets and performs a deeper, systematic qualitative analysis to organize prevention and mitigation strategies into a taxonomy and map them to RE lifecycle activities. - Research questions: RQ1: What types of prevention/mitigation strategies do practitioners employ in RE and how can these be organized into a formal taxonomy? RQ2: Can the identified strategies be categorized into that taxonomy? - Data sources: (1) NaPiRE 2016 survey (226 respondents, 10 countries) where participants selected top RE problems and reported causes, effects, and management strategies. (2) CAPS (an RE-intensive organization) two-step survey of seven RE practitioners; Survey 1 elicited problems and management strategies; Survey 2 followed HET training and elicited additional errors and strategies. - Data cleaning: Many strategies lacked detail; where a response contained multiple actions (e.g., “introduction and use”), it was split into separate strategies, yielding 162 total strategies for analysis. - Analysis phases: 1) Identification of high-level categories: Each author independently reviewed strategies to propose high-level classes, then reconciled to four classes: People, Process, Environment, Ambiguous. Definitions clarified distinctions (e.g., People changes target individual behavior and may be lost with turnover; Process changes alter task flow/procedures and persist; Environment changes modify tools, technologies, roles, culture; Ambiguous for insufficient detail). 2) Definition of low-level categories: The team derived low-level subcategories within each high-level class (e.g., People: Internal Training Activities, Internal Team Encouragement, External Training/Coaching; Process: Investigation, Refinement and Redesign, Definition of Roles/Responsibilities, Resource Allocation; Environment: Addition of New Process/Tool/Approach, Addition of Personnel [internal/external], Addition of Structure, Addition of New Role). 3) Mapping to RE lifecycle: Low-level categories were allocated to RE activities (Elicitation, Analysis, Specification, Validation, Management, Unknown) following Kotonya and Sommerville. Authors independently coded, then resolved disagreements in iterative sessions per Kitchenham et al. - Procedure for classification: Strategies were batched and independently coded into high-level classes, then into low-level classes and lifecycle phases. Disagreements were discussed until full agreement; ambiguous or insufficient items were set aside. Counts adjusted as items were merged or removed for lack of detail.

- Strategy corpus and filtering: 162 strategies identified after cleaning; following classification discussions, 2 merged and 8 marked Ambiguous, leaving 153; after further analysis 2 removed for insufficient information, producing 151 strategies mapped to lifecycle activities. - Distribution by high-level class: Approximately 51% Process changes, 23% Environment changes, 21% People changes; about 5% initially Ambiguous. - Distribution by RE activity (n=151): 51% Management, 15% Validation, 13% Elicitation, 13% Analysis, 5% Specification, 3% Unknown. - Dominant low-level categories: - Process: Refinement and Redesign (RR) was largest with 70 strategies, encompassing communication, documentation, planning/organization, and review/cross-check process changes. - People: Internal Training Activities (ITA) had 26 strategies; People strategies were relatively few and mostly training-related. - Environment: Addition of New Process/Tool/Approach (ANP) had 18 strategies; Environment strategies were more evenly distributed across their subcategories than People strategies. - Frequent concrete mechanisms: Planning and execution of regular communication events/meetings and the use of formal reviews were the most common Process mechanisms; checklists and structured validation were noted as effective. - Emphasis pattern: Strategies overwhelmingly emphasized changing Processes (and secondarily Environment) rather than People, aligning with human factors evidence from other domains that people-focused solutions are less effective for error management. - Notable absence: No explicit strategies focused on removing process steps/roles; all involved introduction/definition/refinement of processes, roles, or responsibilities.

The taxonomy (HEMT) systematically organizes practitioner-reported prevention and mitigation strategies into People, Process, and Environment classes with actionable low-level categories and aligns them with RE lifecycle activities. This structure directly addresses RQ1 and RQ2 by revealing that practitioners predominantly manage human errors through Process redesign/refinement and Environment changes rather than People-focused interventions. The dominance of Management and Validation activities indicates that oversight, communication, change management, and formal reviews are central to error management in RE. The prominence of the RR category suggests practitioners find communication improvements, documentation standards, planning/organization, and review/cross-checks particularly effective. The scarcity and training-centric nature of People strategies, together with external literature, imply that relying on changing individual behavior is less effective and less durable (especially under turnover) than modifying processes and environments. The observed lack of explicit “removal” strategies suggests opportunities to explore elimination of unnecessary steps or sub-tasks as a potential error-reduction technique. Overall, the findings guide practitioners toward prioritizing process and environment interventions to prevent and mitigate human errors.

The study introduces the Human Error Management Taxonomy (HEMT) for requirements engineering, organizing prevention and mitigation strategies into People, Process, and Environment classes with detailed subcategories and mappings to RE lifecycle activities. Analysis of practitioner-reported strategies shows that process-oriented changes dominate and may be broadly applicable across error types, while people-focused approaches are less prevalent and mainly training-oriented. This aligns with evidence from other domains that process and environment modifications are generally more effective than attempts to change individual behavior. The taxonomy provides a structured foundation for future empirical evaluations of strategy effectiveness and completeness, and it can help practitioners select strategies appropriate to specific RE activities. Future work will evaluate HEMT’s usefulness and effectiveness in industrial settings, compare the impact of different strategy types, and investigate underexplored areas such as removal of process steps or roles as an error mitigation technique.

- Subjectivity of survey data: Strategies are self-reported by practitioners; actual effectiveness was not validated. Some responses lacked detail; ambiguous or non-English responses were excluded to reduce misinterpretation. - Construct validity: NaPiRE was not explicitly framed around human error; mapping problems to human errors and strategies may introduce interpretive bias. CAPS participants received HET training, but misunderstanding is still possible. - Sample limitations: NaPiRE offers broad cross-industry, international coverage, improving external validity; CAPS reflects one organization with seven participants, limiting generalizability. Nevertheless, the taxonomy is grounded in real practitioner strategies; further studies are needed to evaluate efficacy and completeness. - Coding challenges: Overlaps between Environment and Process changes and limited detail in responses made precise classification difficult; some items remained ambiguous or were removed.