onboarding

This study addresses the lack of generalizability in onboarding recommendations for newcomers to open-source projects and the conflation of their needs with those of long-term contributors. Through a systematic literature review (SLR), we distilled 15 actionable onboarding recommendations and conducted a large-scale, cross-platform (Gerrit/GitHub) empirical analysis across 1,160 real-world projects. We innovatively identified four types of recommendation effects: universally positive, context-dependent, universally negative, and newcomer-exclusive. Notably, we introduce the “timely abandonment of newcomer-specific strategies” paradigm, emphasizing stage-adaptivity in contributor guidance. Results show that four recommendations significantly increase first-patch acceptance rates, while three are beneficial only during the newcomer phase—continued application reduces long-term retention. This work provides both theoretical grounding and a practical framework for precision, lifecycle-aware onboarding of open-source contributors.

This study investigates effective mentoring practices in open-source software (OSS) communities to identify strategies, mentor attributes, and outcome metrics that enhance newcomer onboarding and competency development. Method: A two-phase mixed-methods survey involving 155 contributors was conducted, integrating Likert-scale quantitative analysis with qualitative thematic coding. Contribution/Results: The study proposes the first empirically grounded “Strategy–Quality–Outcome” mentoring framework for OSS, comprising 21 actionable, evidence-based mentoring strategies; 12 high-consensus core mentor qualities (e.g., patience, responsiveness); and 7 measurable developmental outcomes (e.g., capacity for independent contribution, sense of community belonging). It further introduces a novel Challenge–Strategy mapping matrix and a Quality–Outcome association model. Validated through rigorous empirical analysis, the framework has been formally adopted as a mentor training guideline by multiple international open-source foundations.

New software engineers face significant challenges due to rapid technological evolution and the high cost and obsolescence of conventional onboarding training. This paper proposes an integrated, multi-agent onboarding assistant embedded directly within the development environment, enabling natural-language-driven, real-time code insights, project navigation, and context-aware explanations—delivering low-intervention support. Our key contribution is the first deep integration of chain-of-thought (CoT) reasoning with retrieval-augmented generation (RAG), forming an adaptive, interpretable, and personalized guidance mechanism. The system is implemented using large language models (LLMs) within a collaborative multi-agent architecture. An empirical evaluation with eight participants yielded average ratings of 3.26/4 for helpfulness and 3.0/4 for usability. All source code, documentation, and demonstration videos are publicly released under an open-source license.

This study addresses the persistent challenges research computing centers face during new user onboarding, where despite extensive documentation and training, users often remain confused due to the complexity of infrastructure and software resources. To tackle this issue, the authors propose the first systematic and reusable onboarding framework that transforms generic resources into personalized support through user journey mapping, needs analysis, and service design. The framework was empirically validated within the research infrastructure services at Washington University in St. Louis, demonstrating significant improvements in users’ comprehension and efficient utilization of computing resources. It also increased service adoption rates and user satisfaction. By institutionalizing onboarding best practices for the first time, this work establishes a scalable and transferable optimization paradigm for research computing environments.

This study investigates linguistic inequality faced by non-native English-speaking developers in globally distributed, English-dominant software development environments—and its impact on career advancement and team status. Drawing on Bourdieu’s sociological theory of field, capital, and habitus, it pioneers a systematic sociological analysis of linguistic capital in software engineering. Through in-depth interviews, participant observation, and critical discourse analysis, the study identifies three latent linguistic capital barriers: (1) asymmetrical technical terminology translation, (2) turn-taking exclusion in meetings, and (3) inequitable construction of document authority. It proposes a “progressive habitus adaptation” practice model, explicating how non-native speakers strategically reconfigure linguistic capital to enhance discursive agency and professional voice. Validated and adopted by industry development teams, the model offers both a theoretically grounded framework and actionable interventions for equitable, cross-cultural software collaboration. (149 words)

This study investigates the temporal evolution of the effectiveness of the "Good First Issue" (GFI) mechanism in open-source projects. Through a longitudinal analysis of over 400,000 issues and 1,117 new contributors’ GFI-related pull requests across 37 popular GitHub repositories over four years, it reveals a significant decline in the availability of GFI-labeled issues since 2024. While the rate of new contributor participation has remained stable at approximately 27%, the merge rate of their pull requests has dropped markedly from 61.9% to 42.2%, indicating a weakening efficacy of the GFI mechanism. Furthermore, the study finds that characteristics of initial pull requests fail to predict their eventual merge outcomes, challenging prevailing assumptions and highlighting a growing disconnect between current GFI practices and the retention of new contributors.

This work addresses the challenge faced by new developers in efficiently comprehending complex codebases when documentation and expert support are scarce, while traditional mentoring is costly and difficult to scale. To overcome this, the authors propose LACY, a novel system that uniquely integrates AI-generated content with expert curation to produce reusable “code tours.” LACY introduces innovative interaction modalities such as Voice-to-Tour, supporting voice-recorded explanations, comprehension quizzes, podcast-style audio, and interactive visual dashboards. An evaluation in Beko’s production environment demonstrates that learners using expert-curated tours achieve significantly higher quiz scores (83%) compared to those using purely AI-generated tours (57%). Experts reported lower effort in creating curated tours than delivering live walkthroughs, and the system has been formally adopted, confirming its scalability and practical utility.

This study addresses the limited understanding within open-source communities regarding how characteristics of a newcomer’s first task influence their long-term retention, a gap that hinders the effectiveness of task recommendation and onboarding strategies. Through a large-scale empirical analysis integrating multidimensional task and community interaction data, the work employs machine learning prediction models, SHAP-based interpretability, and causal inference techniques to demonstrate that interactive features—such as submissions by moderately experienced users, moderate discussion intensity, active project member involvement, and neutral-to-slightly-negative comment sentiment—are more strongly associated with increased newcomer retention than intrinsic task attributes. The study further identifies key combinations of features predictive of high-retention tasks, offering actionable insights for designing effective community onboarding interventions.

This study addresses the challenge of cognitive overload experienced by newcomers to open-source software, often caused by disorganized structure, redundant content, and fragmented information in onboarding documentation. To mitigate this barrier to participation, the work proposes a generative AI–driven optimization framework that systematically applies Cognitive Theory of Multimedia Learning (CTML) to documentation redesign. The framework employs semantic segmentation to extract discrete task units, infers implicit workflows, removes redundancies, and generates multimodal explanations to enhance comprehensibility. Expert evaluations confirm the approach’s completeness and feasibility, while user studies demonstrate that it significantly reduces cognitive load for novices, leading to higher task success rates and improved perceived usability.

This work addresses the lack of structured guidance faced by health and life sciences teams when initiating federated learning (FL) projects, often hindered by fragmented frameworks, complex governance, and diverse stakeholder roles. To bridge this gap, the authors propose FLKit—an open, community-maintained onboarding toolkit that innovatively incorporates role-aware entry points tailored to clinicians, legal experts, governance officers, and technical staff. Grounded in ELIXIR’s research data management principles, FLKit integrates an interdisciplinary glossary, a curated tool catalog, and a full-cycle workflow methodology. It supports project planning and documentation through four lifecycle phases and FAIR-aligned FL Story templates. By December 2024, FLKit comprised 39 pages across eight chapters, featuring seven real-world case studies spanning multiple sclerosis, inflammatory bowel disease, and genomics.