Information and Software Technology最新文献

{"title":"Psychological safety in software workplaces: A systematic literature review","authors":"Beatriz Santana ,&nbsp;Lidivânio Monte ,&nbsp;Bianca Santana de Araújo Silva ,&nbsp;Glauco Carneiro ,&nbsp;Sávio Freire ,&nbsp;José Amancio Macedo Santos ,&nbsp;Manoel Mendonça","doi":"10.1016/j.infsof.2025.107838","DOIUrl":"10.1016/j.infsof.2025.107838","url":null,"abstract":"<div><h3>Context:</h3><div>Psychological safety (PS) is an important factor influencing team well-being and performance, particularly in collaborative and dynamic domains such as software development. Despite its acknowledged significance, research on PS within the field of software engineering remains limited. The socio-technical complexities and fast-paced nature of software development present challenges to cultivating PS. To the best of our knowledge, no systematic secondary study has synthesized existing knowledge on PS in the context of software engineering.</div></div><div><h3>Objective:</h3><div>This study aims to systematically review and synthesize the existing body of knowledge on PS in software engineering. Specifically, it seeks to identify the potential antecedents and consequences associated with the presence or absence of PS among individuals involved in the software development process.</div></div><div><h3>Methods:</h3><div>A systematic literature review was conducted, encompassing studies retrieved from four digital libraries. The extracted data were subjected to both quantitative and qualitative analyses.</div></div><div><h3>Results:</h3><div>The findings indicate a growing academic interest in PS within software engineering, with the majority of studies grounded in Edmondson’s framework. Factors antecedents of PS were identified at the individual, team, and organizational levels, including team autonomy, agile methodologies, and leadership behaviors. The presence of PS was found to positively impact team dynamics, software quality, technical excellence, and job satisfaction. Furthermore, PS was observed to mediate relationships between constructs such as ethical leadership and innovative behavior.</div></div><div><h3>Conclusion:</h3><div>PS fosters innovation, learning, and team performance within software development. However, significant gaps persist in understanding the contextual factors influencing PS, its underlying mechanisms, and effective strategies for its enhancement. Future research should address these gaps by investigating the practical applications of PS within diverse organizational settings in the software engineering domain.</div></div>","PeriodicalId":54983,"journal":{"name":"Information and Software Technology","volume":"187 ","pages":"Article 107838"},"PeriodicalIF":3.8,"publicationDate":"2025-07-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144680589","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Detection of code smells in react with TypeScript applications","authors":"Maykon Nunes ,&nbsp;Carla Bezerra ,&nbsp;Fabio Ferreira ,&nbsp;Bruno Gois ,&nbsp;Marco Tulio Valente","doi":"10.1016/j.infsof.2025.107835","DOIUrl":"10.1016/j.infsof.2025.107835","url":null,"abstract":"<div><h3>Context:</h3><div>In the last few years, <span>React</span> has become a popular library for building web interfaces, and more recently, developers have started using it with TypeScript, a language that adds static typing to JavaScript. Although code smells have been explored in several works, there are few studies that focus on the quality problems specific to web interfaces built using <span>React</span> and this language.</div></div><div><h3>Objective:</h3><div>This work aims to identify common code smells in this context, assess how developers perceive their impact, propose a tool to detect them, evaluate the tool’s performance and usability, and analyze the prevalence of these smells in open-source projects.</div></div><div><h3>Method:</h3><div>We conducted a gray literature review to identify code smells, followed by interviews and a community survey to validate them. We developed the tool <span>SniffTSX</span> by extending <span>ReactSniffer</span> to detect the smells. To evaluate the tool, we built a dataset with labeled instances and measured its detection performance using standard metrics. We also carried out a TAM-based user study with undergraduate computing students.</div></div><div><h3>Results:</h3><div>We identified six relevant smells. According to developers’ perception, the most critical and frequent ones are <span>Any Type</span>, <span>Multiple Booleans for State</span>, and <span>Non-Null Assertions</span>. The tool achieved high performance (accuracy: 0.96, precision: 0.98, recall: 0.93, F1: 0.95). The user study showed positive perceptions of usefulness and ease of use. The empirical analysis confirmed the widespread presence of the smells, particularly <span>Any Type</span> and <span>Non-Null Assertions</span>, in real-world projects. Our evaluation results show that <span>SniffTSX</span> is useful.</div></div><div><h3>Conclusion:</h3><div>We contribute a catalog of <span>React</span>-specific TypeScript code smells and a tool to detect them. The tool proved accurate, well-received by students, and effective in revealing real-world smell prevalence, supporting developers in improving code quality.</div></div>","PeriodicalId":54983,"journal":{"name":"Information and Software Technology","volume":"187 ","pages":"Article 107835"},"PeriodicalIF":4.3,"publicationDate":"2025-07-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144739224","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"GNPSum: A code summarization enhancement framework based on Graph Node Position","authors":"Haogang Cheng ,&nbsp;Ling Xu ,&nbsp;Luwen Huangfu ,&nbsp;Chao Liu ,&nbsp;Meng Yan ,&nbsp;Yan Lei","doi":"10.1016/j.infsof.2025.107837","DOIUrl":"10.1016/j.infsof.2025.107837","url":null,"abstract":"<div><div>Code summarization is essential for effectively communicating a code’s core functionality and logic, enhancing software development efficiency, collaboration, and code quality. Traditional work has focused on generating summaries from textual information extracted from the source code. However, these approaches often fail to capture the hierarchical structure critical for effective summarization. To effectively capture the hierarchical structure of the code, which is crucial for accurate summarization, researchers often integrate structural elements such as Syntax Trees (AST) into their models. However, conventional embedding methods struggle to accurately discern the semantic nuances within the code, particularly for nodes with similar content but distinct structural roles. The relative positional information of these nodes, which often conveys semantics absent from the source code itself, is frequently overlooked, limiting the model’s ability to fully exploit the hierarchical and contextual richness inherent in the code structure.</div><div>To overcome these limitations, we propose GNPSum, a code summarization enhancement framework based on the position of the graph node. GNPSum employs a structural combinatorial graph approach (SCG), which extends the AST edges with CFG and DFG to aggregate multimodal information. We introduce a novel positional embedding technique that leverages distances between nodes to reduce semantic ambiguity and guide effective summary generation. Evaluations on extensive Java and Python datasets demonstrate that GNPSum improves 3.30% and 1.82% in the BLEU score, compared to the highest performance baseline. Furthermore, our validation shows that GNPSum significantly enhances the structural comprehension for pre-trained models, resulting in a 1.93% performance boost over models fine-tuned without our framework.</div></div>","PeriodicalId":54983,"journal":{"name":"Information and Software Technology","volume":"187 ","pages":"Article 107837"},"PeriodicalIF":4.3,"publicationDate":"2025-07-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144772357","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Agile software development method cargo cult - Devising an analytical tool","authors":"Tanja Elina Havstorm,&nbsp;Fredrik Karlsson,&nbsp;Shang Gao","doi":"10.1016/j.infsof.2025.107851","DOIUrl":"10.1016/j.infsof.2025.107851","url":null,"abstract":"<div><h3>Context</h3><div>Despite the widespread adoption of agile software development methods (ASDMs) today, many organizations struggle with effective implementation. One reason for this is that some organizations claim to use an ASDM without fully understanding its core principles, or they adhere to old practices while professing to follow a contemporary software development method. This phenomenon is sometimes referred to by practitioners as “cargo cult” (CC) behavior. However, simply labeling something as CC lacks analytical depth.</div></div><div><h3>Objective</h3><div>This paper aims to conceptualize and validate an analytical tool for diagnosing CC and non-CC behavior in software development teams’ use of ASDMs.</div></div><div><h3>Method</h3><div>This study uses a longitudinal ethnographic approach to conceptualize and validate the analytical tool by analyzing four agile practices used by a global industrial manufacturing company.</div></div><div><h3>Results</h3><div>The analytical tool features eight stereotypes—three representing non-CC behaviors and five representing CC behaviors—designed to aid in the analysis of ASDM usage. The tool draws on Social Action Theory and Work Motivation Theory to capture and interpret the CC phenomenon in ASDM use. Using the stereotypes, 36 actions were categorized as CC behavior deviating from documented ASDM practices, and 23 actions as non-CC behavior because they aligned with the documented ASDM and reflected agile goals and values. The tool thus can help both researchers and practitioners gain a deeper understanding of ASDM use in organizations.</div></div><div><h3>Conclusion</h3><div>This study advances understanding of ASDM use by moving beyond the simplistic use of the term “cargo cult”. The developed tool enables structured identification and classification of CC behaviors. The stereotypes provide a way of classifying recurring software development actions against the intended ASDM, allowing the identification of specific types of CC behaviors. The analytical tool enables managers to gain deeper insights into the underlying reasons for deviations, thereby supporting more grounded and effective agile practices within organizations.</div></div>","PeriodicalId":54983,"journal":{"name":"Information and Software Technology","volume":"187 ","pages":"Article 107851"},"PeriodicalIF":3.8,"publicationDate":"2025-07-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144704782","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"An exploratory study of sustaining organisational agility","authors":"Mali Senapathi ,&nbsp;Diane E. Strode","doi":"10.1016/j.infsof.2025.107842","DOIUrl":"10.1016/j.infsof.2025.107842","url":null,"abstract":"<div><h3>Context</h3><div>How an organisation sustains agility over time, and the actions it takes when encountering challenges to its agility, are not well established.</div></div><div><h3>Objective</h3><div>This study explores how an organisation sustained its agility beyond the initial adoption stage when its agility was founded on the values, principles, and practices of agile software development.</div></div><div><h3>Method</h3><div>A qualitative case study of the Retail organisation within a company was carried out. The Retail organisation had sustained agility for six years.</div></div><div><h3>Results</h3><div>Nine main challenges are: disparate levels of maturity, resistance to change, competing priorities, lack of team empowerment, dependencies between teams, onboarding people to the agile way of working, lack of role clarity, staff turnover, and incompatible funding model. Based on the solutions to these challenges, eight general actions applied to sustain agility are: a clear reason for implementing agile principles, continuously raising agile awareness, continuously adjusting the business operating model, embedding specialist agile roles, changing the funding model, making work visible, supporting agile teamwork, and monitor, measure, feedback, and act.</div></div><div><h3>Conclusions</h3><div>The contributions of this article are practical. The challenges, detailed solutions, and general actions identified could be helpful for organisations that need to sustain agility in contexts similar to the case organisation. The findings should enable synthesis with future case studies and potential theory-building.</div></div>","PeriodicalId":54983,"journal":{"name":"Information and Software Technology","volume":"187 ","pages":"Article 107842"},"PeriodicalIF":4.3,"publicationDate":"2025-07-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144722110","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"A joint optimization approach for bug triage leveraging individual ability and collective responsibility of developers","authors":"Haojie Li ,&nbsp;Wei Wei ,&nbsp;Jianfei Sui ,&nbsp;Guanfeng Liu ,&nbsp;Yuying Li ,&nbsp;Yang Feng ,&nbsp;Junwei Du","doi":"10.1016/j.infsof.2025.107833","DOIUrl":"10.1016/j.infsof.2025.107833","url":null,"abstract":"<div><h3>Context:</h3><div>Efficient bug triage is essential to ensuring the quality of open-source software. Studies have shown that incorporating bug tossing information into bug triage has improved its performance substantially.</div></div><div><h3>Objectives:</h3><div>However, the current bug triage approach that integrates tossing relationships only considers the statistical characteristics of the tossing relationship network or the attribute information of bugs, failing to characterize developers’ capabilities based on tossing network comprehensively. Consequently, various factors, such as developers’ availability, roles within the development organization, and subjective intentions, are overlooked, leading to assigning bugs to developers who are not suited for resolving them. We are attempting to propose an approach to address the above problem.</div></div><div><h3>Methods:</h3><div>We introduces an approach that integrates developers’ <strong>I</strong>ndividual <strong>A</strong>bility with the <strong>C</strong>ollective <strong>R</strong>esponsibility(<strong>IACR</strong>) derived from the bug tossing network. Firstly, we use the text and metadata of the bug reports (including product details, hardware, and components) to define the developer’s individual ability feature. The network structure information learned from tossing network relations also represents the collective responsibility feature. Then we design a joint optimization approach to achieve a better fusion of individual ability and collective responsibility in the form of vector representations.</div></div><div><h3>Results:</h3><div>Experimental results from five datasets indicate that IACR improves the average recommendation accuracy and MRR by 6.94% and 4.72%, respectively, compared to the optimal baselines.</div></div><div><h3>Conclusion:</h3><div>IACR achieves an accurate representation of developers’ capabilities and effectively improves the accuracy of bug triage.</div></div>","PeriodicalId":54983,"journal":{"name":"Information and Software Technology","volume":"187 ","pages":"Article 107833"},"PeriodicalIF":3.8,"publicationDate":"2025-07-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144662579","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Enhancing autonomous driving simulations: A hybrid metamorphic testing framework with metamorphic relations generated by GPT","authors":"Yifan Zhang ,&nbsp;Tsong Yueh Chen ,&nbsp;Matthew Pike ,&nbsp;Dave Towey ,&nbsp;Zhihao Ying ,&nbsp;Zhi Quan Zhou","doi":"10.1016/j.infsof.2025.107828","DOIUrl":"10.1016/j.infsof.2025.107828","url":null,"abstract":"<div><h3>Context:</h3><div>Autonomous Driving Systems (ADSs) have rapidly developed over the past decade. Given the complexity and cost of testing ADSs, advanced simulation tools like the CARLA simulator are essential for efficient algorithm development and validation. However, the intricacies of autonomous driving (AD) simulations pose challenges for software testing, particularly the oracle problem, which relates to the difficulty in determining the correctness of outputs within reasonable timeframes. While many studies validate ADS algorithms using simulations, few address the validity of the simulated data, a fundamental premise for ADS testing.</div></div><div><h3>Objective:</h3><div>This study addresses the oracle problem in AD simulations by employing Metamorphic Testing (MT) and Metamorphic Relations (MRs) to detect software defects in the CARLA simulator. Additionally, we explore AI-driven approaches, specifically integrating ChatGPT’s customizable features to enhance MR generation and refinement.</div></div><div><h3>Method:</h3><div>We propose a human-AI hybrid MT framework that combines human inputs with AI-driven automation to generate and refine MRs. The framework uses the GPT-MR generator, a customized large language model (LLM) based on Metamorphic Relation Patterns (MRPs) and ChatGPT, to produce MRs according to user specifications. These MRs are then refined by MT experts and fed into a test harness, automating test-case creation and execution while supporting diverse parameter inputs.</div></div><div><h3>Results:</h3><div>The GPT-MR generator produced effective MRs, leading to the discovery of four significant defects in the CARLA simulator, demonstrating their effectiveness in identifying software flaws. The test harness enabled efficient, automated testing across multiple modules and vehicle-control approaches, which enhanced the robustness and efficiency of our methods.</div></div><div><h3>Conclusions:</h3><div>Our study highlights the effectiveness of MT and MRPs in addressing the oracle problem for AD simulations, enhancing software reliability, and ensuring robust validation processes. The combination of AI-driven tools and human knowledge offers a structured methodology for validating simulated data and ADS performance, contributing to more reliable ADS development and testing.</div></div>","PeriodicalId":54983,"journal":{"name":"Information and Software Technology","volume":"187 ","pages":"Article 107828"},"PeriodicalIF":3.8,"publicationDate":"2025-07-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144623599","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Beyond hard vs. soft: A mixed-methods approach to developing a competency model for E-commerce software engineers","authors":"Shuai Chen ,&nbsp;Yang Zhao","doi":"10.1016/j.infsof.2025.107836","DOIUrl":"10.1016/j.infsof.2025.107836","url":null,"abstract":"<div><h3>Context</h3><div>The competency model for software engineers has emerged as a crucial tool for assessing and enhancing professional skills. However, the conventional binary classification of soft and hard skills fails to adequately capture the nuanced competency requirements across different industries, particularly in e-commerce.</div></div><div><h3>Objectives</h3><div>This study aims to develop a refined competency model specifically tailored to the e-commerce industry, identifying the key competencies essential for software engineers in this domain.</div></div><div><h3>Methods</h3><div>First, a competency framework for e-commerce software engineers was constructed based on a literature review, incorporating four dimensions: social, cognitive, professional, and meta-competencies. Subsequently, machine learning algorithms, topic modeling, and text clustering techniques were applied to job postings from China’s labor market to extract key competency factors. A questionnaire survey was then conducted among 58 software engineers to assess and rank the importance of these competencies.</div></div><div><h3>Results</h3><div>This study developed a competency model comprising four dimensions and 15 competency factors. The survey findings reveal several key insights: cognitive and functional competencies are perceived as the most critical for technical development. Moreover, the perceived importance of competencies varies significantly across different career stages. In the early career phase (1–3 years), cognitive and functional competencies are prioritized. As professionals advance (3–5 years), social competencies gain prominence. However, at the senior level, cognitive competencies once again become central.</div></div><div><h3>Conclusion</h3><div>These findings offer valuable insights for e-commerce companies in designing employee training programs and career development strategies. Additionally, educational institutions can refine their curricula to better align with the evolving competency demands of the industry.</div></div>","PeriodicalId":54983,"journal":{"name":"Information and Software Technology","volume":"187 ","pages":"Article 107836"},"PeriodicalIF":3.8,"publicationDate":"2025-07-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144633618","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"On the use of Test-Driven Development for Embedded Systems","authors":"Pietro Cassieri ,&nbsp;Michelangelo Esposito ,&nbsp;Simone Romano ,&nbsp;Giuseppe Scanniello","doi":"10.1016/j.infsof.2025.107779","DOIUrl":"10.1016/j.infsof.2025.107779","url":null,"abstract":"<div><h3>Context:</h3><div><em>Test-Driven Development</em> (<em>TDD</em> ) is an agile development approach where the writing of unit tests precedes the writing of production code. Some software engineers have been advocating the use of TDD to develop <em>Embedded Systems</em> (<em>ESs</em>) despite the lack of empirical evidence on TDD applied to ES development.</div></div><div><h3>Objectives:</h3><div>Our overarching objective is to start building knowledge on TDD applied to ES development. Specifically, we investigated: <em>(i)</em> the claimed benefits of TDD (<em>i.e.,</em> increased external quality of ESs and developers’ productivity) and <em>(ii)</em> how developers apply TDD when developing ESs.</div></div><div><h3>Method:</h3><div>We conducted two empirical investigations. The former investigation, comprising two experiments, aimed to study the claimed benefits of TDD. To that end, we asked Computer Science (CS) Master’s students to fulfill ES development tasks by using TDD or a non-TDD approach. The latter investigation consisted of an ethnographically-informed study to gain insights into how developers apply TDD when developing ESs. In this study, we involved CS Master’s students, who were observed while fulfilling an ES development task. The participants in the two empirical investigations were different.</div></div><div><h3>Results:</h3><div>We found that the external quality of the developed ESs increased when using TDD as compared to a non-TDD approach, while there was not a substantial difference in developers’ productivity. However, TDD is perceived as more difficult to apply, and the development task is deemed more challenging with TDD. Several patterns emerged about how developers approach TDD when developing an ES (<em>e.g.,</em> the participants did not always develop in conformity with TDD).</div></div><div><h3>Conclusion:</h3><div>TDD seems to be a promising approach to the development of ESs, even though further research is needed. In this regard, we believe that our results can pave the way for future research with the involvement of software companies and professionals.</div></div>","PeriodicalId":54983,"journal":{"name":"Information and Software Technology","volume":"187 ","pages":"Article 107779"},"PeriodicalIF":3.8,"publicationDate":"2025-07-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144596496","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Assisting early-stage software startups with LLMs: Effective prompt engineering and system instruction design","authors":"Thea Lovise Ahlgren ,&nbsp;Helene Fønstelien Sunde ,&nbsp;Kai-Kristian Kemell ,&nbsp;Anh Nguyen-Duc","doi":"10.1016/j.infsof.2025.107832","DOIUrl":"10.1016/j.infsof.2025.107832","url":null,"abstract":"<div><h3>Context:</h3><div>Early-stage software startups, despite their strong innovative potential, experience high failure rates due to factors such as inexperience, limited resources, and market uncertainty. Generative AI technologies, particularly Large Language Models (LLMs), offer promising support opportunities; however, effective strategies for their integration into startup practices remain underexplored.</div></div><div><h3>Objective:</h3><div>This study investigates how prompt engineering and system instruction design can enhance the utility of LLMs in addressing the specific needs and challenges faced by early-stage software startups.</div></div><div><h3>Methods:</h3><div>A Design Science Research (DSR) methodology was adopted, structured into three iterative cycles. In the first cycle, use cases for LLM adoption within the startup context were identified. The second cycle experimented with various prompt patterns to optimize LLM responses for the defined use cases. The third cycle developed “StartupGPT”, an LLM-based assistant tailored for startups, exploring system instruction designs. The solution was evaluated with 25 startup practitioners through a combination of qualitative feedback and quantitative metrics.</div></div><div><h3>Results:</h3><div>The findings show that tailored prompt patterns and system instructions significantly enhance user perceptions of LLM support in real-world startup scenarios. StartupGPT received strong evaluation scores across key dimensions: satisfaction (93.33%), effectiveness (80%), efficiency (80%), and reliability (86.67%). Nonetheless, areas for improvement were identified, particularly in context retention, personalization of suggestions, communication tone, and sourcing external references.</div></div><div><h3>Conclusion:</h3><div>This study empirically validates the applicability of LLMs in early-stage software startups. It offers actionable guidelines for prompt and system instruction design and contributes both theoretical insights and a practical artifact — StartupGPT — that supports startup operations without necessitating costly LLM retraining.</div></div>","PeriodicalId":54983,"journal":{"name":"Information and Software Technology","volume":"187 ","pages":"Article 107832"},"PeriodicalIF":3.8,"publicationDate":"2025-07-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144596495","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}