PaperlayerView exact revision on arXiv

Source-linked AI summary

Socio-Technical Grounded Theory for Software Engineering

Rashina Hoda

arXiv:2103.14235v3cs.SE

TL;DR

Software engineering researchers increasingly use Grounded Theory, but traditional sociological guidance is difficult to apply in modern socio-technical contexts. The paper presents Socio-Technical Grounded Theory, which contextualizes and expands the method with clearer application and evaluation guidance. It also distinguishes limited data-analysis use from full STGT, which is required for claims of mature theory.

  • Problem

    Grounded Theory is increasingly used in software engineering, but researchers face difficulty applying traditional sociological guidelines to socio-technical research contexts.

  • Method

    The paper presents STGT as an iterative and incremental method combining traditional and modern research techniques, with contextualized philosophical, methodological, application, reporting, and evaluation guidelines.

  • Results

    STGT defines the socio-technical research context and provides clearer, more flexible guidance for applications, outcomes, reporting, and evaluation.

  • Takeaways & Limitations

    STGT supports full grounded-theory studies and limited data-analysis applications, while enabling mature-theory claims only when the remaining STGT procedures are applied.

  • Takeaways & Limitations

    STGT for data analysis cannot be claimed as producing a mature theory without the remaining advanced procedures and full-method elements.

Abstract

from arXiv · show

Grounded Theory (GT), a sociological research method designed to study social phenomena, is increasingly being used to investigate the human and social aspects of software engineering (SE). However, being written by and for sociologists, GT is often challenging for a majority of SE researchers to understand and apply. Additionally, SE researchers attempting ad hoc adaptations of traditional GT guidelines for modern socio-technical (ST) contexts often struggle in the absence of clear and relevant guidelines to do so, resulting in poor quality studies. To overcome these research community challenges and leverage modern research opportunities, this paper presents Socio-Technical Grounded Theory (STGT) designed to ease application and achieve quality outcomes. It defines what exactly is meant by an ST research context and presents the STGT guidelines that expand GT's philosophical foundations, provide increased clarity and flexibility in its methodological steps and procedures, define possible scope and contexts of application, encourage frequent reporting of a variety of interim, preliminary, and mature outcomes, and introduce nuanced evaluation guidelines for different outcomes. It is hoped that the SE research community and related ST disciplines such as computer science, data science, artificial intelligence, information systems, human computer/robot/AI interaction, human-centered emerging technologies (and increasingly other disciplines being transformed by rapid digitalisation and AI-based augmentation), will benefit from applying STGT to conduct quality research studies and systematically produce rich findings and mature theories with confidence.

1 INTRODUCTION

Grounded Theory is increasingly used in software engineering, but traditional sociological guidelines are difficult to apply to socio-technical contexts and modern research practices. The paper introduces Socio-Technical Grounded Theory to provide clearer, more flexible guidance for conducting and evaluating such research.

  • Grounded Theory supports systematic, evidence-based theory development and has evolved into Classic, Strauss-Corbinian, and Constructivist versions.
  • Software engineering increasingly uses Grounded Theory to investigate human and social phenomena, including process improvement, self-organization, architecture, design, and self-assignment.
  • Traditional Grounded Theory guidelines were written for sociologists and are difficult for many software engineering researchers to understand, select, adapt, and apply.
  • Software engineering phenomena are socio-technical, so studying social or technical aspects in isolation can produce incomplete investigations and understanding.
  • Modern software engineering data sources, collection tools, and analysis techniques create opportunities that traditional Grounded Theory guidelines do not explicitly address.
  • Socio-Technical Grounded Theory defines the socio-technical context and provides contextualized guidelines for conducting and evaluating research.
  • STGT expands philosophical foundations, supports full-method or data-analysis applications, accommodates modern techniques, and permits emergent or structured theory development.
  • STGT encourages reporting interim, preliminary, and mature outcomes and provides nuanced evaluation guidelines for different applications and outcomes.

2 GROUNDED THEORY – SOCIAL TRADITIONS

Grounded Theory is a rigorous, evidence-based method for developing mid-range theories through iterative qualitative inquiry. Its three traditions differ in how theory emerges, is structured, and is interpreted.

  • The method: GT systematically develops theory from evidence collected in real-world practice through data collection, analysis, and advanced theory-development procedures.Its distinctive focus is inductive theory development through interleaved rounds of data collection and analysis.
  • The method: GT is especially suited to understudied or complex phenomena requiring deep investigation of how and why, not only what happens.Typical procedures include interviews, observations, open coding, constant comparison, memoing, theoretical sampling, sorting, coding, and saturation.
  • Theory outcomes: GT theories are mid-range explanations limited to studied contexts but open to modification when further evidence becomes available.Their applicability is therefore bounded by the contexts investigated.
  • The evolution: The method evolved from Classic or Glaserian GT to Strauss-Corbinian GT and Constructivist GT.These versions differ in their approaches to emergence versus structured theory derivation and the researcher’s interpretive role.

3 GT IN SOFTWARE ENGINEERING

GT has become valuable for studying human and social aspects of software engineering, especially practice-based and theoretically underdeveloped phenomena. However, its traditional guidance remains difficult to access and apply consistently in SE, motivating contextualized alternatives.

  • The promise: GT is well suited to SE topics involving human and social aspects, theory development, practice-based settings, and complex how-and-why questions.Its empirical orientation can connect research rigor with practical relevance and support theory development for SE phenomena.
  • The progress: A review of 98 GT journal papers found that half claimed GT, while only a third documented their method application in detail.The review also found that most studies did not identify which GT version they used.
  • The challenges: SE researchers face recurring problems including unclear GT version selection, ad hoc adaptation, confusion between full GT and isolated analysis techniques, and inadequate evidence support.These problems are associated with limited methodological background, inaccessible traditional texts, and unclear boundaries around GT practice.
  • The challenges: Quality problems also include rushed or overstated GT applications, poor presentation, and reviews that are either overly trusting or overly harsh.The paper links these patterns to misunderstandings, apprehensions, weak reporting, and inappropriate evaluation criteria rather than always to intentional misuse.
  • The need for evolution: STGT is proposed as a modern, contextualized update that defines socio-technical research and provides clearer guidance for conducting it.The update is intended to address traditional GT’s limitations in SE’s socio-technical research context.

4 SOCIO-TECHNICAL GROUNDED THEORY

Socio-Technical Grounded Theory adapts GT for research contexts where human and technical elements are tightly interwoven. It combines iterative theory development with broader methodological guidance, application contexts, reporting practices, and evaluation considerations.

  • STGT method: STGT is an iterative and incremental method for generating novel, useful, parsimonious, and modifiable theories in socio-technical research.It uses traditional and modern techniques with primarily inductive, but also deductive and abductive, reasoning.
  • STGT method: STGT interleaves cycles of data collection and analysis, with each cycle advancing the project toward theoretical outcomes.Advanced procedures can follow either emergent or structured modes of theory development.
  • Guidelines: The STGT guidelines cover socio-technical context, philosophical foundations, methodological procedures, application and reporting, outcomes, and evaluation.The framework is presented as an organized update to traditional GT guidance.
  • Research context: The socio-technical research context is characterized across phenomenon, domain and actors, researcher, and data, tools, and techniques.These dimensions describe a research landscape different from the native contexts for which traditional GT was designed.
  • Research context: SE practices are socio-technical because human-human collaboration and human-technology interactions are tightly coupled.Studying either the social or technical aspect alone can produce an incomplete investigation and understanding.
  • Applications: STGT is positioned for contemporary SE, AI, and related domains involving human factors, values, social relations, technology-mediated activity, and inclusive responsible design.The paper also identifies ethical, credibility, relevance, bias, and interdisciplinary issues as part of the broader context.

Socio-Technical Domain and Actors

Socio-technical domains tightly couple social and technical aspects, with actors who create, maintain, and use technology. STGT therefore supports interdisciplinary researchers equipped with social-science, technical, theoretical, and domain knowledge.

  • Socio-technical domains have tightly coupled social and technical aspects, so studying either aspect alone produces an incomplete understanding.
  • Software engineers are distinctive actors because they both produce and use technology across physical, virtual, and combined workspaces.
  • Actors may include software engineers, domain professionals, software users, virtual avatars, or AI agents, depending on the phenomenon studied.
  • This repertoire can be provided by one researcher or an interdisciplinary team combining software engineering, sociological, and domain expertise.
  • Quality STGT research requires a repertoire spanning philosophical foundations, qualitative research, theoretical sensitivity, socio-technical sensitivity, and domain knowledge.

Socio-Technical Data Tools and Techniques

STGT accommodates traditional and modern data, tools, and techniques across socio-technical contexts. It also broadens the philosophical and reasoning foundations used to develop and evaluate grounded theories.

  • STGT studies can combine traditional interviews, observations, and field trips with modern data sources, tools, and analysis techniques.
  • Software repositories containing commits, source code, documentation, and wikis provide modern avenues for grounded-theory development.
  • Modern techniques such as sentiment analysis can complement open coding and constant comparison, while immersive technologies and automation may support future data collection and analysis.
  • STGT presents accessible foundations covering reasoning, ontology, epistemology, and research paradigms for researchers from software engineering and related fields.
  • Types of Reasoning: Abduction generates explanatory hypotheses from surprising findings and subjects them to further data collection, analysis, or examination.
  • Philosophical Foundations: STGT enables research across physical, virtual, combined, and simulated realities, with epistemological and paradigmatic choices reflecting the studied context.
  • Types of Reasoning: Although STGT remains predominantly inductive, it explicitly recognizes deduction in theoretical sampling and abduction in memoing, theoretical structuring, and theoretical integration.

Ontology

Ontology concerns what researchers believe exists or constitutes reality. In software engineering, socio-technical phenomena may unfold across physical, virtual, augmented, and combined realities.

  • Ontology specifies what researchers believe exists or perceive as reality within a research context.
  • Software engineering research often examines socio-technical phenomena unfolding beyond physical reality, including artificial, virtual, augmented, and combined realities.
  • A software development team may operate simultaneously through physical scrum boards, virtual project tools, and combined in-person interactions with online artefacts.

Epistemology

Epistemology concerns what counts as knowledge and how it is gained, while research paradigms combine ontological and epistemological stances. STGT links these choices to the studied reality and permits flexible paradigm selection.

  • Epistemology addresses what can be treated as knowledge and how knowledge about a research reality is gained.
  • STGT relates epistemological stance to the nature of the studied reality and researcher preferences, allowing subjective or objective approaches in suitable contexts.
  • A research paradigm combines ontological and epistemological stances to formalize the researcher’s worldview about reality and knowledge.
  • Constructivism treats reality and knowledge as socially constructed through researcher-participant interaction, whereas positivism treats reality as external and observable.
  • Researchers should declare their philosophical stance because it influences study conduct, presentation of findings, and evaluation criteria.
  • Researchers may select different paradigms for different studies when their choices fit the study’s nature and reflect mature philosophical understanding.

STGT Research Paradigms

STGT adopts a socio-technical worldview while allowing researchers to select ontological and epistemological stances suited to the research context. This flexibility requires explicit consideration of paradigms so findings can be interpreted and evaluated consistently.

  • STGT permits different research worldviews according to the study’s ontology, context, and researcher preference rather than enforcing one paradigm.This flexibility is not paradigm agnosticism; researchers must carefully consider the research context.
  • Physical and combined realities may suit subjective approaches, whereas finite and deterministic simulated or virtual realities may suit objectivist approaches.These are examples offered for novice researchers rather than universal prescriptions.
  • The same software engineering topic may support different worldviews depending on whether the focus is technical performance or human motivations and behavior.Test case selection illustrates this distinction: an objective approach can address tools and performance, while a constructivist approach can address testers’ rationales and behaviors.
  • Explicitly acknowledging paradigms helps researchers remain consistent and set expectations for how readers and reviewers should evaluate findings.

5 STGT METHOD – OVERVIEW AND BASIC STAGE

STGT combines adapted procedures from three traditional GT versions into a basic stage for accessible data collection and analysis, followed by flexible advanced theory development. It also distinguishes lean and targeted literature reviews and emphasizes socio-technical analysis that treats technical and social aspects together.

  • STGT method overview: STGT adapts Glaserian, Strauss-Corbinian, and Constructivist GT procedures into an accessible basic stage and flexible advanced theory-development stage.The advanced stage offers emergent and structured modes after researchers gain experience with basic data collection and analysis.
  • Basic stage: The basic stage combines lean literature review, study preparation and piloting, and iterative basic data collection and analysis, producing initial concepts, categories, and preliminary hypotheses.
  • Advanced stage: The advanced stage uses targeted literature review and either emergent or structured theory development, with both modes proceeding toward theoretical saturation and mature theory.Emergent development is iterative and organic; structured development follows a guiding theoretical structure.
  • Literature review: STGT separates early lean literature reviews for identifying gaps from periodic targeted reviews that compare emerging findings with existing work and situate them in the research landscape.The approach also supports theoretical sensitivity, socio-technical domain knowledge, and refinement of existing theory.
  • Literature review: Systematic reviews are not recommended for novice GT researchers, and their positivist stance must be reconciled with the STGT stance and subsequent findings when used.
  • Basic data analysis: Socio-technical coding and constant comparison analyze technical and social aspects together, rather than treating software engineering as isolated social and technical components.This approach aims to uncover the full socio-technical nature of activities such as requirements discussions and architecture debates.

6 STGT METHOD – ADVANCED THEORY DEVEL-

STGT postpones commitment to a traditional GT version until researchers have experience with basic analysis. Its advanced stage then offers emergent and structured theory-development modes sharing targeted review, theoretical sampling, advanced memoing, and saturation.

  • STGT delays choosing among traditional GT versions until novice researchers have experienced basic data-analysis procedures.This addresses the difficulty of understanding differences among Glaserian, Strauss-Corbinian, and Constructivist GT approaches upfront.
  • Emergent and structured modes both include targeted literature review, theoretical sampling, advanced memoing, and theoretical saturation.They differ in whether theory development proceeds organically or through a structured theoretical framework.
  • Advanced memoing revisits, reassesses, refines, and compares memos to strengthen supported relationships, reject unsupported ones, and establish new links between categories.
  • Comparing memos helps generate possible links between concepts and categories, whose development uses abductive reasoning and is strengthened by additional evidence.
  • STGT treats hypothesis development and structuring as iterative and incremental rather than a one-time theoretical-sorting activity.

1. Advanced memoing

STGT’s advanced procedures guide researchers from theoretical sampling and targeted or structured analysis toward saturation, then toward optional structuring or integration. Emergent mode suits unclear theoretical structures, while structured mode suits topics with a clear guiding structure.

  • Advanced memoing: Theoretical saturation occurs when further data collection adds no new or significant concepts, categories, or insights, with final collection attempts serving mainly to verify the theory.
  • Emergent mode: Emergent mode iteratively targets data collection and analysis until saturation, then uses optional theoretical structuring to derive mature theory.The approach is especially suited when strong categories exist but a clear theoretical structure has not yet emerged.
  • Emergent mode: Targeted coding limits advanced analysis to the most significant concepts and categories from the basic stage while allowing some new categories to emerge.
  • Emergent mode: Theoretical structuring identifies a suitable theory genre and may map the emergent structure onto a pre-existing template; it is optional but highly recommended.
  • Structured mode: Structured mode uses a guiding theoretical structure to organize further data collection and analysis, progressively fitting, strengthening, and saturating existing categories.It suits relatively narrow topics whose basic stage produces a comparatively clear theoretical structure.
  • Structured mode: Theoretical integration finalizes the theory’s storyline by identifying the phenomenon that its categories and relationships collectively capture and explain.Integration may also produce an additional theoretical layer, such as a classification system.

7 STGT - APPLICATION, OUTCOMES, REPORTING

STGT supports both full theory development and limited data analysis across standalone, mixed-methods, and R&D contexts. The section distinguishes the outcomes and claims appropriate to each application level while emphasizing ethical data practices and reporting at different stages.

  • Application modes: STGT can be applied as a full method producing mature theories or in a limited capacity producing descriptive findings, categories, hypotheses, and propositions.Limited application uses basic data-analysis techniques within or alongside other methods.
  • Application modes: Full STGT applications may be standalone or integrated into mixed-methods studies, larger projects, or research-and-development programs.Standalone studies range from small investigations over six to eighteen months to large studies spanning several years.
  • Outcome boundaries: Limited STGT applications should not be presented as producing mature grounded theories without the advanced analysis, theory-development, theoretical-sampling, and iterative procedures of full STGT.Combining basic and advanced procedures with other methods can make grounded-theory claims difficult to justify.
  • Ethics: STGT studies should conduct early ethical assessment, obtain approval where applicable, and protect confidentiality and anonymity when collecting or reporting data.Public digital data can be especially challenging because consent and de-identification may be difficult, particularly when reporting could cause harm.
  • Outcomes and reporting: Full STGT studies can report mature theories, while partial findings and preliminary theories may be reported earlier through workshops, conferences, and journals to obtain practitioner and research feedback.Mature theories are suited to formats such as journals, theses, or books because of their size and density.

8 STGT – EVALUATION GUIDELINES

STGT evaluation separates the quality of method application from the quality of outcomes. Criteria are deliberately nuanced because partial findings, preliminary theories, and mature theories require different standards and evidence.

  • Evaluation framework: STGT manuscripts should be evaluated for both method application and outcomes rather than with a single undifferentiated checklist.The evaluation framework distinguishes criteria for different outcome types.
  • Method application: Credibility depends on showing how STGT guidelines were practically applied, including recruitment, sampling, iterative analysis, memoing, theoretical sampling, and saturation.Manuscripts should also report the theory-development mode and relevant ontological and epistemological stances.
  • Reporting: Methodological space constraints do not remove the need to report application details, which can instead be supplemented with additional materials.The methodology section should be treated as a first-class part of the manuscript.
  • Method application: Rigour requires evidence of raw data, analysis procedures, research artefacts, and, for mature theories, the advanced theory-development process.Sanitised quotations, coding examples, and evidence of theory development strengthen the evidentiary basis while respecting privacy.
  • Outcome evaluation: Partial outcomes and preliminary theories should demonstrate originality, relevance, and density, whereas mature theories additionally require novelty, usefulness, parsimony, and modifiability.Parsimony concerns compact conceptual density that explains complex phenomena through key categories and their relationships.
  • Outcome evaluation: Mature theories are expected to offer unprecedented or fresh insights, actionable implications, compact explanations, and the capacity to be modified.These criteria go beyond originality, relevance, density, and practical resonance required for less mature outcomes.

9 VISION AND NEW FRONTIERS

The paper positions STGT as a framework for socio-technical research in software engineering and related fields. It identifies extensions beyond conventional grounded theory, including varied applications, broader contexts, modern data, and potential AI augmentation.

  • Core contribution: STGT defines a socio-technical research context and provides philosophical, methodological, application, reporting, and evaluation guidelines for researchers developing theory.The framework is intended to clarify and ease grounded-theory use for researchers who struggle with traditional GT approaches.
  • Boundaries and future directions: STGT may not fit studies focused on social phenomena without technical aspects, for which traditional sociological grounded-theory methods may be more applicable.The paper also identifies potential for technology and AI to supplement, augment, or automate parts of qualitative analysis.
  • Beyond software engineering: The framework is positioned as relevant beyond software engineering to socio-technical disciplines where primarily qualitative data can illuminate the phenomenon.Examples include information systems, computer science, human-computer interaction, artificial intelligence, and user experience research.
  • Beyond grounded theory: The socio-technical research framework may also inform methods beyond STGT, including socio-technical case studies, survey research, and ethnography.Its relevance is presented as extending to research methods generally in increasingly socio-technical settings.
  • Beyond theory development: STGT may be used for full theory development or for data analysis alone, either standalone, with other methods, or within R&D studies and large programs.This range extends STGT beyond a single study design or level of methodological commitment.
  • Beyond local theories: Modern digital data sources and collection and analysis approaches may improve the relevance and generalisability of grounded theories across contexts without eliminating contextual nuance.The paper does not treat universality as an aim or quality criterion of STGT.
Loading 2103.14235v3…