arXiv is now an independent nonprofit! Learn more
License: CC BY 4.0
arXiv:2503.22238v1 [cs.SE] 28 Mar 2025arXiv:2503.22238v1 [cs.SE] 28 Mar 2025

Integrating LLMs in Software Engineering Education: Motivators, Demotivators, and a Roadmap Towards a Framework for Finnish Higher Education Institutes将大语言模型整合到软件工程教育中：激励因素、阻碍因素及面向芬兰高等教育机构框架的路线图

DOI: XXXXXXX.XXXXXXXISBN: 978-1-4503-XXXX-X/2018/06CCS: Software and its engineering Software creation and management
Maryam Khan Affiliation: Software Engineering Department, Lappeenranta-Lahti University of Technology, Lappeenranta, Finland email: maryam.khan@student.lut.fi , Muhammad Azeem Akbar Affiliation: Software Engineering Department, Lappeenranta-Lahti University of Technology, Lappeenranta, Finland email: azeem.akbar@lut.fi and Jussi Kasurinen Affiliation: Software Engineering Department, Lappeenranta-Lahti University of Technology, Lappeenranta, Finland email: Jussi.Kasurinen@lut.fi
20182018
Abstract.摘要：

The increasing adoption of Large Language Models (LLMs) in software engineering education presents both opportunities and challenges. While LLMs offer benefits such as enhanced learning experiences, automated assessments, and personalized tutoring, their integration also raises concerns about academic integrity, student over-reliance, and ethical considerations. In this study, we conducted a preliminary literature review to identify motivators and demotivators for using LLMs in software engineering education. We applied a thematic mapping process to categorize and structure these factors (motivators and demotivators), offering a comprehensive view of their impact. In total, we identified 25 motivators and 30 demotivators, which are further organized into four high-level themes. This mapping provides a structured framework for understanding the factors that influence the integration of LLMs in software engineering education, both positively and negatively. As part of a larger research project, this study serves as a feasibility assessment, laying the groundwork for future systematic literature review and empirical studies. Ultimately, this project aims to develop a framework to assist Finnish higher education institutions in effectively integrating LLMs into software engineering education while addressing potential risks and challenges.大语言模型（LLM）在软件工程教育中的应用日益广泛，既带来了机遇，也带来了挑战。LLM虽然具有增强学习体验、自动化评估和个性化辅导等优势，但其整合也引发了对学术诚信、学生过度依赖和伦理问题的担忧。在本研究中，我们进行了初步文献综述，以识别在软件工程教育中使用LLM的激励因素和阻碍因素。我们采用主题映射流程对这些因素（激励因素和阻碍因素）进行分类和结构化，从而全面了解其影响。我们共识别出25个激励因素和30个阻碍因素，并进一步将其归纳为四个高层主题。这一映射为理解积极和消极影响软件工程教育中LLM整合的因素提供了结构化框架。作为更大研究项目的一部分，本研究作为可行性评估，为未来的系统文献综述和实证研究奠定基础。最终，该项目旨在制定一个框架，帮助芬兰高等教育机构在应对潜在风险和挑战的同时，将LLM有效整合到软件工程教育中。

Keywords: 关键词：
Large Language Models, Software Engineering, Motivators, Demotivators 关键词：大语言模型、软件工程、激励因素、阻碍因素

1. INTRODUCTION1. 引言

The rise of artificial intelligence (AI) has transformed various industries towards automation (Acemoglu and Restrepo, 2018), including healthcare (Shaheen, 2021), finance (Cao, 2022), entertainment (Forbus and Laird, 2002), and notably education (Zhai et al., 2021; Jin et al., 2025), the primary focus of this study. LLMs lead the AI transformation and revolution, based on their core characteristics such as generating human-like responses to various prompts. LLMs are defined using deep learning algorithms and Natural Language Processing (NLP) techniques to develop coherent and relevant responses (Akbar et al., 2023). The models are based on transformer architectures and pre-trained on large amounts of textual data.人工智能（AI）的兴起已推动包括医疗保健（Shaheen, 2021）、金融（Cao, 2022）、娱乐（Forbus and Laird, 2002）以及尤为值得关注的教育（Zhai et al., 2021; Jin et al., 2025）在内的多个行业迈向自动化（Acemoglu and Restrepo, 2018），其中教育是本研究的主要关注点。LLM凭借其根据各种提示生成类似人类回答的核心特性，引领着AI的转型与变革。LLM的定义是使用深度学习算法和自然语言处理（NLP）技术来生成连贯且相关的回答（Akbar et al., 2023）。这些模型基于Transformer架构，并在大量文本数据上进行预训练。

The unique characteristics and expanding applications of LLMs introduce a spectrum of potential opportunities and distinct challenges, particularly in the landscape of software engineering education (He et al., 2024). LLMs hold significant potential for use in academia, especially in software engineering education—performing various software engineering activities. However, software engineering education is at a crucial stage due to AI-based technologies such as LLMs, which find applications in chatbots, content creation, and code development tools (Kirova et al., 2024). Therefore, the increasing use of LLMs necessitates restructuring software engineering education to integrate AI technologies (Kirova et al., 2024; Fan et al., 2023). Modern software engineering education programs need to holistically incorporate the unique applications of LLMs, ensuring that students are educated about the challenges and opportunities they present (Daun and Brings, 2023).LLM的独特特性和不断扩展的应用引入了一系列潜在机遇和独特挑战，尤其是在软件工程教育领域（He et al., 2024）。LLM在学术界具有巨大的应用潜力，特别是在软件工程教育中——可执行各种软件工程活动。然而，由于LLM等基于AI的技术被应用于聊天机器人、内容创作和代码开发工具，软件工程教育正处于一个关键阶段（Kirova et al., 2024）。因此，LLM日益广泛的使用要求重构软件工程教育，以整合AI技术（Kirova et al., 2024; Fan et al., 2023）。现代软件工程教育项目需要整体地融入LLM的独特应用，确保学生了解这些应用带来的挑战与机遇（Daun and Brings, 2023）。

Various research studies have been conducted focusing on the integration of AI technologies in education, such as software engineering and computer science (Banerjee et al., 2025; Feldt et al., 2018; Lo, 2023). However, the rapidly evolving AI landscape presents unique challenges for designing curricula and pedagogy. Incorporating AI tools, machine learning techniques, and the ethical implications of LLMs into software engineering education are active areas of research discussion and still remain relatively unexplored.已有大量研究聚焦于AI技术在教育（如软件工程和计算机科学）中的整合（Banerjee et al., 2025; Feldt et al., 2018; Lo, 2023）。然而，快速演进的AI格局给课程设计与教学法带来了独特挑战。将AI工具、机器学习技术以及LLM的伦理影响融入软件工程教育，仍是研究讨论的热点领域，同时也相对而言尚未被充分探索。

As highlighted in recent research, Finland progressive education system, known for its innovation and inclusivity, provides an ideal context for developing a tailored framework that aligns with national pedagogical goals and regulatory requirements (FattahiBavandpour, 2024). Recent work also emphasizes the importance of usability-focused AI tools in education, particularly in empowering teachers with AI-driven course management and curriculum planning solutions. These tools not only support educators but also enhance sustainable and work-life-relevant education by integrating LLM capabilities into existing learning platforms (Macias et al., 2024). However, despite these advancements, the application of LLMs in software engineering education in Finland remains highly unexplored. While AI-driven tools have been adopted in digital humanities and curriculum planning, their structured integration into SE education, particularly in Finnish universities, lacks systematic research and implementation.近期研究强调，芬兰进步的教育体系以其创新性和包容性著称，为制定符合国家教学目标和监管要求的定制化框架提供了理想环境（FattahiBavandpour, 2024）。近期工作还强调了以可用性为核心的AI工具在教育中的重要性，特别是通过AI驱动的课程管理和课程规划解决方案为教师赋权。这些工具不仅为教育者提供支持，还通过将LLM能力集成到现有学习平台，增强了可持续且与工作和生活相关的教育（Macias et al., 2024）。然而，尽管取得了这些进展，LLM在芬兰软件工程教育中的应用仍极不成熟。虽然AI驱动工具已被采用于数字人文学科和课程规划，但其在软件工程教育（尤其是芬兰大学中）中的系统性整合仍缺乏系统研究和实施。

It is acknowledged that while LLMs are powerful tools, they are not without limitations. Concerns have been raised about their susceptibility for biased decision-making, ethical issues, and their "black box" nature (Bender et al., 2021). These challenges are not just academic concerns; their practical implications make them highly relevant for software engineering students. The growing discussion on AI ethics has also captured significant attention in academia. This is reflected in the literature, where numerous studies have been conducted to emphasize the importance of understanding AI principles such as transparency, privacy, accountability, fairness, and autonomy (Khan et al., 2023). These ethical principles are crucial not only to the AI field but also to software engineering students who are primarily involved in developing AI-powered software solutions. It is essential to integrate the ethical aspects of LLMs into the software engineering curriculum (Khan et al., 2022).应当承认，LLM虽然是强大的工具，但并非没有局限性。人们已对其可能产生有偏见的决策、伦理问题及其“黑箱”性质提出担忧（Bender et al., 2021）。这些挑战不仅是学术问题；其实际影响使它们与软件工程学生高度相关。围绕AI伦理日益增长的讨论也引起了学术界的广泛关注。这反映在文献中，已有大量研究强调理解透明度、隐私、问责、公平和自主性等AI原则的重要性（Khan et al., 2023）。这些伦理原则不仅对AI领域至关重要，对那些主要参与开发AI驱动软件解决方案的软件工程学生而言同样至关重要。将LLM的伦理层面纳入软件工程课程是必不可少的（Khan et al., 2022）。

Therefore, this work outlines how software engineering institutes in Finland are incorporating LLMs into their educational systems to attain the promising benefits of AI technologies. This integration reflects the need for a deep understanding of these technologies—LLMs. This research focuses on several key areas of knowledge to provide a framework for Finnish software engineering research institutes. These include the challenges associated with integrating LLMs into educational systems, the causes of these challenges, and the best solutions. Furthermore, we aim to understand the existing guidelines and frameworks developed for integrating LLMs in education across other domains. Based on this, we will define a framework specific to software engineering education for Finnish higher education institutes. In the long term, this work aims to contribute to the ongoing discussions and dialogues concerning the future of software engineering education in an AI environment and serve as a set of guidelines for educational institutes, policymakers, students, and curriculum developers. To achieve the study aims and objectives, we have developed the core research questions and their rationale given in Table 1.因此，本工作概述了芬兰软件工程机构如何将LLM整合到其教育系统中，以获得AI技术带来的可观益处。这种整合反映出对这些技术——LLM——进行深入理解的必要性。本研究聚焦于若干关键知识领域，旨在为芬兰软件工程研究机构提供一个框架。这些领域包括将LLM整合到教育系统中所面临的挑战、这些挑战的成因及最佳解决方案。此外，我们旨在了解其他领域为将LLM整合到教育中而制定的现有指南和框架。在此基础上，我们将为芬兰高等教育机构定义一个针对软件工程教育的框架。从长远来看，本工作旨在为有关AI环境下软件工程教育未来的持续讨论与对话作出贡献，并作为一套指南，供教育机构、政策制定者、学生和课程开发者参考。为实现研究目的与目标，我们提出了核心研究问题及其依据，见表1。

In this article, we present the preliminary literature review of this project to address RQ1.1, focusing on discussions surrounding the use of LLMs in software engineering education to establish the feasibility of this project. We identified 25 motivating and 30 demotivating factors influencing the adoption of LLMs in software engineering education. Furthermore, a thematic mapping of these factors was conducted to categorize them into high-level themes. In the future, we aim to conduct a comprehensive systematic literature review (SLR) and empirical studies to develop the final framework (See Section 2).在本文中，我们介绍了本项目为回答RQ1.1而进行的初步文献综述，重点关注围绕软件工程教育中使用LLM的讨论，以确定本项目的可行性。我们确定了影响软件工程教育中LLM采用的25个激励因素和30个阻碍因素。此外，我们对这些因素进行了主题映射，将其归入高层主题。未来，我们计划开展全面的系统文献综述（SLR）和实证研究，以制定最终框架（见第2节）。

The rest of the paper is structured as follows: The research methodology of the project and the preliminary review protocol are briefly discussed in Section 2. The study results are presented in Section 4. The implications of this study are detailed in Section 5, followed by a discussion on validity threats in Section 5.3. Finally, we conclude the study and outline future research directions in Section 7.本文其余部分的结构如下：第2节简要讨论了项目的研究方法和初步综述方案。第4节展示了研究结果。第5节详细阐述了本研究的意义，随后在第5.3节讨论了效度威胁。最后，第7节对研究进行总结并概述未来研究方向。

Table 1. Research Questions of this Project表1. 本项目的研究问题
RQ# Research Question Rationale
RQ1 How can LLMs be effectively integrated into Finnish higher education institutions to enhance teaching and learning in software engineering education? This research question aims to explore the integration of LLMs into Finnish higher education to potentially transform and enhance the pedagogical practices in software engineering education. It aims to identify effective strategies and practices for deploying LLMs that could improve both teaching methods and student learning outcomes. The focus on Finnish institutions provides context-specific insights that may lead to tailored solutions in educational technology.
RQ1.1 What are the motivators and demotivators of integrating LLMs into software engineering education? This research sub-question investigates the specific success factors (motivators) and challenges (demotivators) that should be considered when using LLMs in software engineering educational programs in Finnish higher education institutes. It aims to uncover technical, ethical, and pedagogical motivators and demotivators in using AI-driven tools within a curriculum traditionally focused on highly technical and specialized skills. Understanding these factors is crucial for developing strategies to effectively implement LLMs and increase their educational benefits.
RQ1.2 What are the causes of the identified challenges and mitigation practices? This question is defined to investigate the root causes of the challenges identified in RQ1.1. It seeks to understand the underlying factors contributing to these challenges and to explore potential mitigation practices. It will help in developing effective strategies and solutions that will ensure a successful integration of LLMs into educational frameworks (RQ3).
RQ2 What are the existing guidelines/frameworks in other domains that could assist higher education institutions in implementing LLMs? This RQ aims to explore existing guidelines and frameworks in various domains (such as computer science, business, etc.) and contexts (other countries) that could inform and support the integration of LLMs into software engineering higher education settings in Finland.
RQ3 What would be a framework to assist Finnish higher education institutions in implementing LLMs? The final RQ aims to develop a specific framework to assist Finnish higher education institutions in effectively integrating LLMs into software engineering education. It will provide a structured approach that addresses the unique educational, cultural, and technical contexts of Finnish institutions. By establishing a clear framework, the question seeks to ensure that the deployment of LLMs enhances teaching and learning outcomes while also navigating any potential challenges specific to the local academic environment. The inputs of this research question are RQ1-RQ2.

2. Research Methodology2. 研究方法

To address the study research questions, we will use mixed research methods, including a preliminary review (traditional literature survey) in the first phase of the research (this study), Systematic Literature Review (SLR), and empirical studies (questionnaire survey, interviews, and case study) in the second and final phase. The core steps of the proposed research methodology are discussed as follows and presented in Figure 1:为回答研究问题，我们将采用混合研究方法，包括研究第一阶段（即本研究）的初步综述（传统文献调查）、系统文献综述（SLR），以及第二和最后阶段的实证研究（问卷调查、访谈和案例研究）。所提研究方法的核心步骤如下所述，并在图1中呈现：

Refer to caption
Figure 1. Proposed Research Process图1. 拟议研究流程
  • •

    Preliminary Literature Survey In this article, we conducted a preliminary literature survey using the traditional literature survey approach (see Section 3) to identify motivators and demotivators of using LLMs in software engineering education to answer (RQ1.1). This preliminary study is conducted to evaluate the feasibility of this project in the long term. Additionally, it helps in identifying gaps in existing research and refining our research objectives for more targeted investigations. The insights gained from this phase provide a foundational understanding that informs subsequent stages of the study.初步文献调查 在本文中，我们采用传统文献调查方法（见第3节）进行了初步文献调查，以识别在软件工程教育中使用LLM的激励因素和阻碍因素，从而回答（RQ1.1）。初步研究的开展旨在长期评估本项目的可行性。此外，它有助于识别现有研究的空白，并完善我们的研究目标以进行更有针对性的调查。该阶段的见解提供了基础性理解，为后续研究阶段提供信息。

  • •

    Systematic Literature Review (SLR) The SLR approach is primarily chosen to explore the existing literature and provide a comprehensive overview of motivators, demotivators, their causes, and mitigation practices of using LLMs in Finnish software engineering institutes (RQ1.1, RQ1.2). . This phase builds upon the insights gained from the Preliminary Literature Survey, ensuring a structured and in-depth investigation of relevant studies. Following the principles of evidence-based software engineering, we ensure a rigorous review process by systematically examining all primary studies related to the research questions in an unbiased and iterative way (Kitchenham et al., 2007).系统文献综述（SLR） 选择SLR方法主要是为了探索现有文献，全面概述在芬兰软件工程机构中使用LLM的激励因素、阻碍因素、其成因以及缓解实践（RQ1.1、RQ1.2）。这一阶段建立在初步文献调查所获见解的基础上，确保对相关研究进行结构化、深入的调查。遵循循证软件工程的原则，我们通过以无偏且迭代的方式系统检查与研究问题相关的所有主要研究，确保严密的综述过程（Kitchenham et al., 2007）。

  • •

    Empirical Studies The SLR findings will be empirically evaluated by conducting a questionnaire-based survey, complemented with interviews and a case study:实证研究 SLR的发现将通过基于问卷调查的实证评估，并辅以访谈和案例研究：

    • –

      Questionnaire Survey: A detailed survey (Punter et al., 2003), incorporating both closed and open-ended questions, will be conducted with teachers, students, and IT administrators to evaluate the SLR findings, including the identified motivators, demotivators, their thematic mapping, causes of demotivators, and best practices. Additionally, the survey will help identify new motivators, demotivators, causes, and best practices that are not discussed in the existing literature, further enriching the SLR findings. Moreover, we will validate the thematic mapping of motivators and demotivators based on the perceptions of survey participants. Finally, the descriptive analysis will be used to analyze the collected data (Kitchenham et al., 2002).问卷调查：将面向教师、学生和IT管理人员开展一项详细调查（Punter et al., 2003），包括封闭式和开放式问题，以评估SLR的发现，包括所识别的激励因素、阻碍因素、其主题映射、阻碍因素的成因和最佳实践。此外，该调查将有助于识别现有文献中未讨论的新激励因素、阻碍因素、成因和最佳实践，进一步丰富SLR的发现。并且，我们将根据调查参与者的看法验证激励因素和阻碍因素的主题映射。最后，将使用描述性分析对收集的数据进行分析（Kitchenham et al., 2002）。

    • –

      Interviews: Comprehensive interviews with experts in software engineering education will be conducted to validate the questionnaire survey findings and understand their perspectives about the findings of previous phases. To analyze the interview data, the Grounded Theory (GT) (Glaser and Strauss, 2017) approach will be used to define and derive the core codes from raw data and develop their mapping.访谈：将与软件工程教育领域的专家进行全面访谈，以验证问卷调查结果，并了解他们对前几阶段研究结果的看法。为分析访谈数据，将采用扎根理论（GT）（Glaser and Strauss, 2017）方法，从原始数据中定义和提炼核心编码，并发展其映射。

    • –

      Case Studies: Based on the interview findings, the framework will be finalized and its real-world effectiveness evaluated using a case study approach (Runeson and Höst, 2009). Several Finnish universities will be targeted to deploy the framework and assess its effectiveness. Furthermore, it will be comparatively analyzed to determine if there are any existing practices of using LLMs in software engineering education. Eventually, the findings from Finland will be compared with those from other countries that have adopted similar technologies in education.案例研究：基于访谈结果，最终确定框架，并使用案例研究方法（Runeson and Höst, 2009）评估其在现实世界中的有效性。将选取若干芬兰大学部署该框架并评估其有效性。此外，将进行对比分析，以确定目前是否有在软件工程教育中使用LLM的实践。最后，将芬兰的发现与在其他国家采用类似教育技术的发现进行比较。

In this study, we explicitly focus on the first phase of the research by conducting a preliminary traditional review (see Section 3) to evaluate the feasibility of the project by answering RQ1.1.在本研究中，我们明确聚焦于研究的第一阶段，通过进行初步的传统综述（见第3节）来评估项目的可行性，以回答RQ1.1。

3. Proposed Research Approach3. 拟议研究方法

In the long term, we plan to conduct an SLR, empirical and case studies as described in Section 2. However, to assess the significance of the proposed project, we conducted a preliminary traditional literature review to understand the key motivators and demotivators (RQ1.1) of using LLMs in software engineering education. The aim of this study is to investigate the factors—both motivators and demotivators—that positively and negatively influence the integration of LLMs in software engineering education. In this literature survey, the snowballing approach is adopted, which consists of backward snowballing (selecting studies based on their references) and forward snowballing (identifying studies that cite the selected studies) (Wohlin, 2014). This approach gradually increases the pool of relevant studies. The studies identified through snowballing focus on the motivating factors (success factors) and demotivating factors (challenges) associated with using LLMs in software engineering education. We also included studies that do not explicitly mention these factors but focus on utilizing LLMs in software engineering education (Niazi, 2015; Khan et al., 2021). However, extracting these factors from such studies is relatively challenging, requiring an extensive and detailed review (Niazi, 2015).从长远来看，我们计划开展第2节所述的SLR、实证研究和案例研究。然而，为了评估拟议项目的重要性，我们进行了一项初步传统文献综述，以了解在软件工程教育中使用LLM的关键激励因素和阻碍因素（RQ1.1）。本研究旨在调查积极和消极影响软件工程教育中LLM整合的因素——包括激励因素和阻碍因素。该文献调查采用滚雪球法，包括向后滚雪球（根据参考文献选择研究）和向前滚雪球（识别引用所选研究的研究）（Wohlin, 2014）。这种方法逐步增加了相关研究的数量。通过滚雪球识别出的研究聚焦于在软件工程教育中使用LLM相关的激励因素（成功因素）和阻碍因素（挑战）。我们还纳入了一些并未明确提及这些因素但聚焦于在软件工程教育中利用LLM的研究（Niazi, 2015; Khan et al., 2021）。然而，从这类研究中提取这些因素相对困难，需要进行全面而详细的综述（Niazi, 2015）。

To ensure a broad search across digital repositories, we used Google Scholar as the primary search engine. It provides integrated access to various information technology and education-related repositories such as IEEE Xplore, Springer, ACM Digital Library, Wiley, etc. (Khan et al., 2021). This approach ensures wide coverage and reduces the risk of missing relevant studies.为确保在数字资源库中进行广泛检索，我们使用Google Scholar作为主要检索引擎。它可集成访问多个信息技术和教育相关资源库，如IEEE Xplore、Springer、ACM Digital Library、Wiley等（Khan et al., 2021）。这种方法可确保广泛的覆盖面，并降低遗漏相关研究的风险。

The first and second authors (supervisor) conducted the initial review process and resolved any disagreements in study selection through mutual discussion and consensus. Ultimately, a total of 45 studies were selected using the snowballing approach. These studies were preliminarily analyzed to extract a list of motivators and demotivators.第一作者和第二作者（导师）进行了初步综述过程，并通过相互讨论和共识解决了研究选择中的任何分歧。最终，通过滚雪球法共选择了45项研究。对这些研究进行了初步分析，以提取激励因素和阻碍因素清单。

The first and second authors reviewed all the selected studies and developed a list of motivators and demotivators for using LLMs in software engineering education. To eliminate interpersonal biases in the review process conducted by the authors, two external reviewers were invited. The reviewers were asked to randomly select ten studies from the list of selected studies and perform the review following the same procedure as the authors.第一作者和第二作者审阅了所有选定的研究，并制定了在软件工程教育中使用LLM的激励因素和阻碍因素清单。为消除作者综述过程中的人际偏见，邀请了两名外部评审人员。评审人员被要求从选定研究列表中随机选择十项研究，并按照与作者相同的流程进行综述。

Finally, Kendall’s coefficient of concordance (W) test is performed to assess interpersonal biases between the authors and the external reviewers. Kendall’s coefficient of concordance (W) is a well-known statistical test used to evaluate the degree of agreement among a group of people assessing a set consisting of n variables (Sidney, 1957). The W score ranges from 0 to 1, where 0 indicates complete dis-agreement and 1 represents strong agreement. We used R-3.6.3 to calculate the W value, and the results, shown in Table 2 (W = 0.82), indicate that the degree of agreement between the authors and the invited reviewers in the review process is positive. This suggests that there were no interpersonal biases in the review process performed by the authors. The R code developed to calculate the W value is as follows:最后，执行肯德尔和谐系数（W）检验，以评估作者与外部评审人员之间的人际偏见。肯德尔和谐系数（W）是一种著名的统计检验方法，用于评估一组人对由n个变量构成的集合进行评估时的同意程度（Sidney, 1957）。W值介于0到1之间，其中0表示完全不同意，1表示高度一致。我们使用R-3.6.3计算W值，结果如表2所示（W = 0.82），表明在综述过程中作者与受邀评审人员之间的同意程度是积极的。这表明作者所执行的综述过程中不存在人际偏见。用于计算W值的R代码如下：

⬇
library DescToolslibrary DescTools()
# Define the rankings given by two external reviewers and authors (10 studies)# Define the rankings given by two external reviewers and authors (10 studies)
LLMs Education data frameLLMs Education data frame_SE_ <- .(
externalexternal_rev1 (6, 10, 2),rev1 (6, 10, 2), = c 5, 7, 1, 4, 8, 3, 9,
external_rev2rev2 = c(6, 7, 5, 2, 1, 4, 10, 9, 8, 3),3),
authorsauthors = c(4, 15,(4, 15, 5, 7, 1, 2, 3, 9, 8, 6)
)
# Calculate Kendall’s coefficient of concordance (W)# Calculate Kendall’s coefficient of concordance (W)
result KendallW correct testresult KendallW correct test <- (LLMs_SE_Education, =TRUE, =TRUE)
# Print results# Print results
printprint(result)
Table 2. Kendall’s Coefficient of Concordance (W) Test Results for LLMs in Software Engineering Education表2. 肯德尔和谐系数（W）对软件工程教育中LLM的检验结果
Data Set Chi-Square df Vars Eval P-val W
LLMs in SE Education 34.733 14 15 3 0.001 0.826

We included the list of all 45 selected studies in the replication package (Khan, 2025) prepared for this study. It is not provided in the article due to page limitations.我们将全部45项选定研究的列表包含在为本次研究准备的复现包中（Khan, 2025）。由于篇幅限制，本文未提供该列表。

4. Results4. 结果

4.1. Identified Motivators4.1. 识别的激励因素

Following the review protocol discussed in Section 3, we identified a total of 25 motivators that could positively affect the use of LLM tools in software engineering education. To present these motivators in a meaningful structure, we conducted a thematic analysis following the guidelines defined by Braun and Clarke (Braun and Clarke, 2006). Thematic analysis is a well-known approach in social sciences for analyzing data to identify and report patterns (themes). The process consists of five core phases: (1) familiarization with the data (noting patterns, meanings, and initial ideas to develop a deep understanding of the content), (2) generating initial codes (systematically coding the data by identifying features of interest across the dataset), (3) searching for themes (sorting codes into potential themes by identifying broader patterns), (4) reviewing themes (checking the coherence of themes concerning the coded extracts), and (5) defining themes (identifying the essence of each theme and refining its scope).遵循第3节所述的综述方案，我们共识别出25个可能对软件工程教育中LLM工具的使用产生积极影响的激励因素。为以一种有意义的结构呈现这些激励因素，我们按照Braun和Clarke（Braun and Clarke, 2006）定义的指南进行了主题分析。主题分析是社会科学中用于分析数据以识别和报告模式（主题）的一种知名方法。该过程包含五个核心阶段：（1）熟悉数据（记录模式、意义和初始想法，以深入理解内容）；（2）生成初始编码（通过识别数据集中感兴趣的特征，对数据进行系统性编码）；（3）搜索主题（通过识别更广泛模式，将编码归类为潜在主题）；（4）审查主题（检查主题相对于编码摘录的一致性）；（5）定义主题（识别每个主题的本质并细化其范围）。

In this study, the first author primarily led the thematic mapping process and defined the themes and sub-themes of the identified motivators and demotivators. The second author continuously reviewed the mapping activities and collaborated with the first author to resolve any differences or address potential issues such as bias. Finally, the identified motivators are mapped across 9 sub-themes and 4 high-level themes as shown in Figure 2.在本研究中，第一作者主要主导了主题映射过程，并定义了已识别的激励因素和阻碍因素的主题和子主题。第二作者持续审查了映射活动，并与第一作者合作解决任何分歧或处理诸如偏见等潜在问题。最终，已识别的激励因素被映射到9个子主题和4个高层主题，如图2所示。

Refer to caption
Figure 2. Thematic Mapping of Identified Motivators图2. 已识别激励因素的主题映射

We now summarize the high-level themes and the details are as follows:下面概括高层主题，详情如下：

4.1.1. Enhancing Learning Experiences4.1.1. 提升学习体验

Enhancing Learning Experiences emerged as an important high-level theme consisting of total 9 motivating factors and 3 sub-themes. It mainly focuses on how LLMs contribute to creating a more personalized, interactive, and engaging educational environment for software engineering students (Cowan et al., 2023; Jošt et al., 2024). The use of LLMs in software engineering education significantly enhances learning experiences by fostering personalized and adaptive learning, engagement, and conceptual understanding (Cowan et al., 2023). LLMs provide tailored learning experiences, enabling students to learn at their own pace with AI-driven tutoring and adaptive recommendations. This level of customization ensures better knowledge retention and curriculum alignment to individual student needs (Korpimies et al., 2024). Additionally, LLMs increase student engagement and motivation by facilitating interactive learning environments that encourage curiosity-driven exploration, making learning more dynamic and accessible to diverse learning styles (Cambaz and Zhang, 2024). The ability of LLMs to simplify complex programming concepts with clear explanations fosters conceptual understanding and problem-solving skills, helping students develop critical debugging abilities and embrace iterative learning through experimentation (Cowan et al., 2023; Jošt et al., 2024). These advantages make AI a valuable tool in modern education, bridging the gap between students’ learning needs and the evolving demands of software engineering.提升学习体验作为一个重要高层主题，共包含9个激励因素和3个子主题。它主要关注LLM如何帮助为软件工程学生创造更个性化、互动性和参与性更强的教育环境（Cowan et al., 2023; Jošt et al., 2024）。在软件工程教育中使用LLM通过促进个性化自适应学习、参与度和概念理解，显著提升了学习体验（Cowan et al., 2023）。LLM提供量身定制的学习体验，使学生能够借助AI驱动的辅导和自适应推荐，按照自己的节奏学习。这种定制水平可确保更好的知识保持，并使课程与个体学生的需求对齐（Korpimies et al., 2024）。此外，LLM通过营造鼓励好奇心驱动探索的互动学习环境，提高了学生的参与度和学习动力，使学习更加生动，并适应多样化的学习风格（Cambaz and Zhang, 2024）。LLM能够以清晰的解释简化复杂编程概念，促进概念理解和问题解决技能，帮助学生培养关键的调试能力，并通过实验拥抱迭代学习（Cowan et al., 2023; Jošt et al., 2024）。这些优势使AI成为现代教育中的宝贵工具，弥合了学生学习需求与软件工程不断演化的要求之间的差距。

4.1.2. Assessment and Feedback in Education4.1.2. 教育中的评估与反馈

LLMs play a pivotal role in improving assessment and feedback mechanisms by automating grading, providing real-time formative feedback, and reducing instructor workload (Meißner et al., 2024). AI-powered automated assessment and grading systems offer instant feedback on assignments and projects, enabling students to quickly identify mistakes and improve their work (Rasnayaka et al., 2024; Lyu et al., 2024). Moreover, AI-assisted test case generation and evaluation contribute to more efficient and scalable assessment processes, particularly for large classes. Beyond grading, LLMs support formative feedback and learning guidance, allowing students to refine their solutions through real-time, interactive feedback (Meißner et al., 2024). This not only enhances learning but also alleviates the burden on educators by automating the evaluation of student progress. The ability of AI to provide consistent and objective feedback ensures a fair assessment process, making AI-assisted evaluation an essential tool for modern education (Meißner et al., 2024). Additionally, AI-enhanced assessments can detect patterns in student performance, helping instructors tailor their teaching strategies to address common difficulties (Hanifi et al., 2023). This data-driven approach ensures that students receive targeted support for their weaker areas, leading to more effective knowledge acquisition. Furthermore, LLMs can simulate real-world evaluation scenarios, preparing students for professional coding environments where automated testing and debugging are standard practices. By integrating AI into assessment strategies, educators can foster a more efficient, data-driven, and student-centered evaluation system that aligns with industry expectations.LLM在改进评估与反馈机制方面发挥着关键作用，可自动化评分、提供实时形成性反馈并减轻教师工作负担（Meißner et al., 2024）。借助AI的自动化评估和评分系统可对作业和项目提供即时反馈，使学生能够快速识别错误并改进作品（Rasnayaka et al., 2024; Lyu et al., 2024）。此外，AI辅助的测试用例生成和评估有助于实现更高效、可扩展的评估流程，尤其是对于大班教学。除评分外，LLM还可支持形成性反馈和学习指导，使学生能够通过实时互动反馈改进解决方案（Meißner et al., 2024）。这不仅促进了学习，还通过自动化评估学生进展减轻了教育工作者的负担。AI提供一致且客观反馈的能力确保了评估过程的公平性，使AI辅助评估成为现代教育的重要工具（Meißner et al., 2024）。此外，AI增强的评估可以检测学生表现中的模式，帮助教师调整教学策略以应对常见困难（Hanifi et al., 2023）。这种数据驱动的方法可确保学生在薄弱领域获得有针对性的支持，从而实现更有效的知识获取。进一步而言，LLM可以模拟真实场景的评估环境，使学生为自动化测试和调试成为标准实践的专业编码环境做好准备。通过将AI整合到评估策略中，教育工作者可以构建一个更高效、数据驱动且以学生为中心、符合行业期望的评估体系。

4.1.3. Collaboration and Peer Learning4.1.3. 协作与同伴学习

LLMs contribute to collaboration and peer learning by acting as AI-assisted learning partners and enhancing project-based, inquiry-based learning approaches (Cambaz and Zhang, 2024; Vierhauser et al., 2024). AI serves as a virtual tutor for brainstorming and problem-solving, allowing students to bridge skill gaps and gain structured guidance in programming tasks (Wang et al., 2024). This fosters an environment where teamwork and knowledge sharing become more effective, as AI can facilitate discussions, provide relevant insights, and help students understand complex topics collaboratively. Additionally, AI improves group project coordination, offering structured guidance that ensures real-world application of engineering principles while supporting agile and iterative methodologies. This integration of AI in teamwork-driven learning encourages students to engage more actively in peer learning, strengthening both technical and collaborative skills essential for software engineering professionals (Kharrufa et al., 2024). Moreover, AI-driven learning environments can enable students to work asynchronously on projects, improving flexibility and accessibility for remote or international teams. AI can also provide dynamic feedback on collaborative efforts, helping teams refine their work based on AI-generated insights and best practices (Marques et al., 2024). By leveraging AI as a knowledge-sharing tool, students can expand their perspectives, learning from globalized AI resources beyond their immediate academic environment (Santos et al., 2024). This creates a more interconnected and informed student community, fostering a culture of collective learning and innovation in software engineering education.LLM通过作为AI辅助的学习伙伴，并增强基于项目和探究式学习方法，为协作与同伴学习作出贡献（Cambaz and Zhang, 2024; Vierhauser et al., 2024）。AI在头脑风暴和问题解决中充当虚拟导师，使学生能够弥合技能差距，并在编程任务中获得结构化指导（Wang et al., 2024）。这营造了一种团队合作和知识共享更高效的环境，因为AI可以促进讨论、提供相关见解并帮助学生协作理解复杂主题。此外，AI改善小组项目协调，提供结构化指导，在支持敏捷和迭代方法的同时确保工程原理在现实世界中的应用。AI在团队驱动的学习中的整合鼓励学生更积极地参与同伴学习，强化软件工程专业人员所需的技术与协作技能（Kharrufa et al., 2024）。再者，AI驱动的学习环境可使学生异步参与项目，提高远程或国际团队的灵活性和可及性。AI还可以对协作成果提供动态反馈，帮助团队根据AI生成的见解和最佳实践改进工作（Marques et al., 2024）。通过将AI用作知识共享工具，学生可以超越身边学术环境，从全球化的AI资源中学习以拓展视野（Santos et al., 2024）。这创造了一个更具互联性和信息灵通的学生社区，在软件工程教育中培养了集体学习与创新的文化。

4.1.4. Skill Development in Software Engineering Education4.1.4. 软件工程教育中的技能培养

LLMs significantly enhance skill development in software engineering education by offering programming assistance, debugging support, and software engineering process understanding. AI helps students tackle coding challenges by providing contextual explanations and step-by-step debugging guidance, making complex software development tasks more manageable (Meißner et al., 2024; Herden et al., 2024). By offering worked-out examples and best coding practices, AI ensures that students develop robust programming habits and reinforce essential software development skills (Boguslawski et al., 2025). Furthermore, AI supports a comprehensive understanding of the software engineering lifecycle, guiding students through documentation standards, software development processes, and real-world engineering workflows (Pereira et al., 2024). Exposure to industry-relevant software engineering methodologies through AI-driven insights prepares students for practical applications of their learning, ensuring they are equipped with the knowledge and skills necessary for professional success. Additionally, AI tools can simulate industry coding environments, allowing students to practice within real-world constraints such as efficiency, scalability, and security considerations (Zhang et al., 2023). This hands-on experience with AI guidance helps bridge the gap between theoretical knowledge and applied engineering skills, making students better prepared for the demands of the workforce. Moreover, AI-driven debugging assistance accelerates the learning curve for novice programmers by highlighting common mistakes and suggesting optimized solutions (Cowan et al., 2023; Cambaz and Zhang, 2024; Wang et al., 2024). By integrating AI into skill development strategies, software engineering education can produce more competent and industry-ready graduates, equipping them with the problem-solving capabilities essential for future technological advancements.LLM通过提供编程辅助、调试支持和软件工程流程理解，显著增强了软件工程教育中的技能培养。AI通过提供情境化解释和逐步调试指导来帮助学生应对编码挑战，使复杂的软件开发任务更易于管理（Meißner et al., 2024; Herden et al., 2024）。通过提供完整的示例和最佳编码实践，AI确保学生养成稳健的编程习惯，并强化基本的软件开发技能（Boguslawski et al., 2025）。此外，AI支持对软件工程生命周期的全面理解，引导学生掌握文档标准、软件开发流程和真实工程工作流（Pereira et al., 2024）。通过AI驱动的见解接触与行业相关的软件工程方法，可让学生做好将所学应用于实践的准备，确保他们具备职业成功所需的知识和技能。此外，AI工具可模拟行业编码环境，使学生能够在效率、可扩展性和安全性等真实约束下进行练习（Zhang et al., 2023）。在AI指导下的这种实践体验有助于弥合理论知识与实际工程技能之间的差距，使学生更好地适应劳动力市场的需求。并且，AI驱动的调试辅助通过指出常见错误和提出优化解决方案，加速了初学者程序员的学习曲线（Cowan et al., 2023; Cambaz and Zhang, 2024; Wang et al., 2024）。通过将AI整合到技能培养策略中，软件工程教育可以培养出更有能力、更具行业准备度的毕业生，使其具备推动未来技术进步所必需的问题解决能力。

In summary, while LLMs significantly enhance learning experiences, assessment processes, collaboration, and skill development, their effectiveness depends on structured implementation strategies that prevent over-reliance on AI-generated solutions. Educators must design curricula that leverage AI as a complement to traditional teaching methods rather than as a replacement. Ethical concerns regarding AI-assisted assessments, academic integrity, and student dependence on automation must also be addressed. Ensuring that LLMs serve as facilitators of deeper understanding rather than shortcuts to learning outcomes is crucial. The future of software engineering education will benefit from AI-driven tools, but their role should be carefully managed to maintain a balance between technological support and the cultivation of essential problem-solving skills among students.总而言之，虽然LLM显著增强了学习体验、评估过程、协作和技能培养，但其有效性取决于结构化的实施策略，以防止对AI生成解决方案的过度依赖。教育工作者必须设计将AI用作传统教学方法之补充而非替代的课程。还必须解决与AI辅助评估、学术诚信以及学生对自动化依赖相关的伦理问题。确保LLM充当加深理解的促进者，而非实现学习成果的捷径，这一点至关重要。软件工程教育的未来将受益于AI驱动的工具，但应审慎管理其角色，以在技术支持和学生基本问题解决技能的培养之间保持平衡。

4.2. Identified Demotivators4.2. 已识别的阻碍因素

Following the defined review protocol (Section 3), we identified 30 demotivators that could hinder the integration of LLMs in software engineering education. Additionally, we performed a thematic mapping of these demotivators, as outlined in Section 4.1. Our analysis resulted in the development of four high-level themes and ten sub-themes of demotivating factors, as illustrated in Figure 3. Below, we discuss each high-level theme in detail.按照既定的综述方案（第3节），我们识别出30个可能阻碍LLM在软件工程教育中整合的阻碍因素。此外，我们按照第4.1节所述的流程对这些阻碍因素进行了主题映射。我们的分析形成了四个高层主题和十个子主题的阻碍因素，如图3所示。下面我们详细讨论每个高层主题。

Refer to caption
Figure 3. Thematic Mapping of Identified Demotivators图3. 已识别阻碍因素的主题映射

4.2.1. Learning and Educational Challenges4.2.1. 学习与教育挑战

The high-level theme of Learning and Educational Challenges highlights critical concerns regarding the integration of LLMs in software engineering education, categorized into three key sub-themes: limitations in understanding and context, bias and hallucination in LLM outputs, and over-reliance on AI in learning. LLMs struggle with contextual understanding, leading to inaccurate handling of domain-specific topics and difficulties in maintaining conversation coherence, which can hinder deep learning and effective knowledge transfer (Sallou et al., 2024). Furthermore, bias and hallucination in LLM outputs pose a significant risk, as the presence of biased training data and misinformation can affect the credibility and quality of educational content, potentially misleading students (Jalil et al., 2023; Hussein, 2024). Inconsistencies in AI-generated explanations further exacerbate these issues, impacting learning accuracy and reliability. The third challenge, over-reliance on AI in learning, raises concerns about students becoming overly dependent on AI-generated responses rather than engaging in critical thinking, debugging, and problem-solving (Ozkaya, 2023). This reliance can lead to a decline in cognitive development and self-directed learning, reducing students’ motivation for independent exploration (Cambaz and Zhang, 2024; Santos et al., 2024). Collectively, these factors emphasize the need for a balanced approach in integrating LLMs into software engineering education, ensuring that AI complements rather than replaces essential learning and problem-solving skills.“学习与教育挑战”这一高层主题凸显了在软件工程教育中整合LLM的重要关切，并归结为三个关键子主题：理解与情境方面的局限性、LLM输出中的偏见与幻觉，以及学习中对AI的过度依赖。LLM在情境理解方面存在困难，导致对领域特定主题的处理不准确，并难以维持对话连贯性，这可能阻碍深度学习和有效的知识传递（Sallou et al., 2024）。此外，LLM输出中的偏见和幻觉构成重大风险，因为带有偏见的训练数据和错误信息会影响教育内容的可信度和质量，可能误导学生（Jalil et al., 2023; Hussein, 2024）。AI生成解释中的不一致进一步加剧了这些问题，影响学习的准确性和可靠性。第三个挑战——学习中对AI的过度依赖——引发了对学生过于依赖AI生成回答而非参与批判性思维、调试和问题解决的担忧（Ozkaya, 2023）。这种依赖可能导致认知发展和自主学习能力下降，降低学生独立探索的动力（Cambaz and Zhang, 2024; Santos et al., 2024）。总体而言，这些因素强调在软件工程教育中整合LLM时需要采取平衡的方法，确保AI补充而非替代基本的学习和问题解决技能。

4.2.2. Assessment and Academic Integrity4.2.2. 评估与学术诚信

The integration of LLMs in software engineering education raises significant concerns regarding assessment and academic integrity, particularly in plagiarism, intellectual property, and ethical concerns in AI-assisted learning. AI-generated content can lead to increased instances of academic dishonesty, as students may use LLMs to complete assignments without genuine comprehension, making it difficult to verify originality (Hussein, 2024). This threatens the value of assessments and undermines learning objectives. Additionally, ethical concerns arise in AI-assisted assessments, where AI feedback may misalign with traditional grading criteria, leading to potential misuse by students to gain unfair advantages (Samhan et al., 2024). The ease of manipulating AI-generated solutions also creates risks in maintaining fair evaluation processes (Gao et al., 2024). The growing dependence on AI for content generation further blurs the line between original student work and AI-assisted outputs, making it imperative for institutions to develop stronger detection mechanisms for AI-generated plagiarism. Moreover, the risk of students misusing AI to bypass academic effort threatens the development of genuine expertise in software engineering disciplines (Bender et al., 2021). To mitigate these risks, universities must implement strict AI policies, promote academic integrity awareness, and redesign assessments to encourage critical thinking over rote AI-generated responses.LLM在软件工程教育中的整合引发了关于评估和学术诚信的重要关切，尤其在抄袭、知识产权以及AI辅助学习中的伦理问题方面。AI生成内容可能导致学术不端行为增多，因为学生可能使用LLM在没有真正理解的情况下完成作业，这使原创性验证变得困难（Hussein, 2024）。这会威胁评估的价值并破坏学习目标。此外，AI辅助评估中会出现伦理问题，因为AI反馈可能与传统的评分标准不一致，导致学生可能滥用AI以获取不公平优势（Samhan et al., 2024）。AI生成解决方案易于操纵也给维持公平的评估过程带来风险（Gao et al., 2024）。对AI生成内容日益增长的依赖进一步模糊了学生原创作品与AI辅助产出之间的界限，迫使各机构开发更强大的AI生成抄袭检测机制。此外，学生滥用AI以逃避学术努力的风险，威胁到软件工程学科真正专业能力的发展（Bender et al., 2021）。为减轻这些风险，大学必须实施严格的AI政策，提升学术诚信意识，并重新设计评估方式，以鼓励批判性思维，而非机械地复制AI生成的回答。

4.3. Student Skill Development and Cognitive Load4.3. 学生技能发展与认知负荷

A key challenge in adopting LLMs in education is their impact on student skill development and cognitive load, especially in critical thinking, problem-solving, and evaluating learning outcomes (Wang et al., 2025). Dependence on AI-generated solutions may diminish students ability to engage in critical reasoning and debugging, leading to superficial knowledge acquisition rather than deep learning. The automation of learning tasks reduces cognitive effort, which, while lowering barriers for understanding, may also stunt problem-solving development (Toma and Yánez-Pérez, 2024). Furthermore, assessing individual learning progress becomes difficult, as distinguishing between AI-assisted knowledge and independently acquired skills remains a challenge. The lack of established frameworks for evaluating AI-based learning outcomes further complicates assessment strategies, raising concerns about the effectiveness of AI-driven education in fostering essential analytical and problem-solving skills. Without proper guidance on the responsible use of AI, students may rely on LLMs to shortcut difficult tasks (Joshi et al., 2023), diminishing their resilience in tackling complex coding and debugging challenges. Additionally, the ease of access to AI-generated answers may reduce intellectual curiosity, preventing students from exploring alternative problem-solving strategies. A well-structured integration of AI should focus on enhancing learning rather than replacing cognitive engagement, ensuring that AI acts as a supportive tool rather than a substitute for fundamental engineering competencies.在教育中采用LLM的一个关键挑战是它们对学生技能发展和认知负荷的影响，尤其是在批判性思维、问题解决和学习成果评估方面（Wang et al., 2025）。对AI生成解决方案的依赖可能会削弱学生进行批判性推理和调试的能力，导致浅层知识获取而非深度学习。学习任务的自动化减少了认知努力，虽然降低了理解门槛，但也可能阻碍问题解决能力的发展（Toma and Yánez-Pérez, 2024）。此外，评估个体学习进展变得困难，因为区分AI辅助知识与独立习得的技能仍是一项挑战。缺乏评估基于AI学习成果的既定框架，使评估策略更加复杂，并引发了对AI驱动教育在培养基本分析和解决问题能力方面有效性的担忧。如果缺乏关于负责任使用AI的适当指导，学生可能会依赖LLM走捷径完成困难任务（Joshi et al., 2023），降低他们应对复杂编码和调试挑战的能力。此外，获取AI生成答案的便捷性可能降低求知欲，阻碍学生探索替代性的问题解决策略。一个结构良好的AI整合应侧重于促进学习而非取代认知参与，确保AI作为支持工具，而非基本工程能力的替代品。

4.3.1. Integration and Practical Implementation4.3.1. 整合与实际实施

The practical implementation of LLMs in software engineering education is hindered by challenges in course redesign, computational costs, and security concerns. Adapting existing curricula to incorporate LLM-based learning tools is complex and resource-intensive, as educators often resist shifting to AI-driven methodologies due to a lack of clear guidelines for ethical AI integration (Fuligni et al., 2025; Harvey et al., 2025). Additionally, high computational and infrastructure costs create accessibility barriers, limiting equitable access to AI-driven tools and widening learning gaps among students. Running large-scale AI models also raises energy consumption concerns, adding another layer of complexity to implementation (Belchev, 2025). Moreover, security, privacy, and data integrity risks emerge as critical issues, with the potential for LLMs to store or leak sensitive academic data and pose compliance challenges with educational data protection regulations (Yao et al., 2024). The lack of institutional support for AI-driven curriculum transformation makes the transition even more challenging, as many educators lack proper training on how to effectively leverage AI without compromising learning objectives. In addition, the absence of regulatory frameworks for AI in education leaves many ethical and security questions unanswered, making institutions hesitant to adopt these technologies on a large scale (Schiff, 2022). To ensure sustainable and fair AI integration, universities must invest in AI literacy programs, provide equitable access to AI resources, and establish strict security measures that protect students’ academic data while optimizing the benefits of AI-enhanced education.LLM在软件工程教育中的实际实施受到课程重新设计、计算成本和安全性问题等挑战的阻碍。调整现有课程以纳入基于LLM的学习工具既复杂又耗费资源，因为教育工作者往往因缺乏关于伦理AI整合的明确指南而抵制转向AI驱动的方法（Fuligni et al., 2025; Harvey et al., 2025）。此外，高昂的计算和基础设施成本造成了可及性障碍，限制了公平获取AI驱动工具的机会，并加剧了学生之间的学习差距。运行大规模AI模型还引发了能耗问题，为实施增加了另一层复杂性（Belchev, 2025）。另外，安全、隐私和数据完整性风险成为关键问题，因为LLM可能存储或泄露敏感的学术数据，并对教育数据保护法规的合规性构成挑战（Yao et al., 2024）。缺乏对AI驱动课程转型的机构支持使这一转变更具挑战性，因为许多教育工作者缺乏如何在不影响学习目标的前提下有效利用AI的适当培训。此外，教育领域AI监管框架的缺失使许多伦理和安全问题悬而未决，使各机构对大规模采用这些技术持谨慎态度（Schiff, 2022）。为确保可持续和公平的AI整合，大学必须投资于AI素养计划，提供平等的AI资源获取途径，并建立严格的安全措施以保护学生的学术数据，同时最大化AI增强教育的益处。

Overall, the integration of Large LLMs in software engineering education presents significant challenges across multiple dimensions, necessitating a balanced and well-regulated approach. Learning and educational challenges arise from LLMs’ limitations in contextual understanding, biases, hallucinations, and the risk of over-reliance, which may hinder students’ critical thinking and independent problem-solving skills. Assessment and academic integrity concerns highlight the risks of AI-assisted plagiarism, ethical dilemmas in grading, and the potential for academic dishonesty, necessitating stronger detection mechanisms and redesigned assessment strategies. The impact on student skill development and cognitive load further complicates AI adoption, as excessive reliance on LLMs can diminish deep learning, critical reasoning, and problem-solving abilities, making it crucial to establish clear evaluation frameworks that ensure AI supports rather than replaces cognitive engagement. Integration and practical implementation challenges, including course redesign, high computational costs, and security risks, create further barriers, requiring institutions to invest in AI literacy programs, infrastructure, and data protection strategies. Collectively, these challenges underscore the need for a structured and ethical AI integration framework that ensures LLMs enhance rather than disrupt software engineering education, supporting students in developing real-world competencies while maintaining academic integrity and inclusivity.总体而言，大语言模型（LLM）在软件工程教育中的整合在多个方面带来重大挑战，需要一种平衡且规范的方法。学习与教育挑战源于LLM在情境理解、偏见、幻觉和过度依赖风险方面的局限，这可能阻碍学生的批判性思维和独立问题解决能力。评估与学术诚信的关切凸显了AI辅助抄袭、评分中的伦理困境以及学术不端的潜在风险，因此需要更强大的检测机制和重新设计的评估策略。对学生技能发展与认知负荷的影响进一步使AI应用复杂化，因为过度依赖LLM会削弱深度学习、批判推理和问题解决能力，因此必须建立清晰的评估框架，确保AI支持而非取代认知参与。整合与实际实施的挑战，包括课程重新设计、高昂计算成本和安全风险，构成了进一步障碍，要求机构投资于AI素养计划、基础设施和数据保护策略。综合来看，这些挑战突出表明，需要一个结构化且符合伦理的AI整合框架，以确保LLM增强而非扰乱软件工程教育，在维护学术诚信和包容性的同时，支持学生发展现实世界中的能力。

5. Implications and Validity Threats5. 启示与效度威胁

We now summarise the key research and practical implications of this study results and the validity threats:我们现在总结本研究结果的关键研究意义、实践意义以及有效性威胁。

5.1. Research Implications5.1. 研究意义

The identified motivators and demotivators of using LLMs in software engineering education provide critical insights into the evolving role of AI in academic settings. From a research perspective, the motivators, such as enhanced learning experiences, personalized education, and improved assessment mechanisms, highlight the potential of LLMs to revolutionize software engineering education. These findings contribute to existing literature by offering evidence on how AI-driven tools can support adaptive learning and improve student engagement. Additionally, the demotivators, including concerns about academic integrity, AI biases, over-reliance on automation, and security risks, present new challenges that require further exploration.在软件工程教育中使用LLM的激励因素和阻碍因素为AI在学术环境中不断演变的作用提供了重要见解。从研究角度看，这些激励因素，如增强的学习体验、个性化教育和改进的评估机制，突显了LLM彻底改变软件工程教育的潜力。这些发现通过提供AI驱动工具如何支持适应性学习并提高学生参与度的证据，为现有文献做出了贡献。此外，阻碍因素，包括对学术诚信、AI偏见、过度依赖自动化和安全风险的担忧，提出了需要进一步探索的新挑战。

5.2. Practical Implications5.2. 实践意义

The identified motivators suggest that institutions can leverage LLMs to enhance personalized learning, automate assessment, and support skill development in software engineering education. These insights provide a roadmap for integrating AI-driven tools into curricula while ensuring student engagement and pedagogical effectiveness. On the other hand, the demotivators highlight the need for proactive strategies to address ethical concerns, mitigate biases, and prevent over-reliance on AI-generated content. Universities and educators must establish clear policies on AI usage, promote academic integrity, and ensure that AI-assisted learning complements rather than replaces critical thinking and problem-solving skills. The findings also stress the importance of faculty training and AI literacy programs to equip educators with the necessary skills to implement and regulate AI-driven learning tools.已识别的激励因素表明，院校可以利用LLM来增强个性化学习、自动化评估，并支持软件工程教育中的技能发展。这些见解为将AI驱动工具整合到课程中提供了路线图，同时确保学生参与度和教学效果。另一方面，阻碍因素凸显了制定主动策略的必要性，以应对伦理问题、减轻偏见并防止过度依赖AI生成的内容。大学和教育工作者必须制定明确的AI使用政策，促进学术诚信，并确保AI辅助学习补充而非取代批判性思维和问题解决技能。研究结果还强调了教师培训和AI素养项目的重要性，以使教育工作者具备实施和规范AI驱动学习工具所需的技能。

Future research will be built on these findings to develop adaptive frameworks that will assist in bringing LLMs tools in Software Engineering education in Finnish higher education institutes.未来的研究将基于这些发现，开发适应性框架，以帮助芬兰高等教育机构在软件工程教育中引入LLM工具。

5.3. Threats to Validity5.3. 有效性威胁

Various threats could affect the validity of the study findings. In this study, we analyzed the potential threats in four main categories—internal, external, construct, and conclusion validity—as defined by Wohlin (Wohlin et al., 2012).多种威胁可能影响研究结果的有效性。在本研究中，我们按照Wohlin等人（Wohlin et al., 2012）的定义，从四个主要类别分析了潜在威胁：内部效度、外部效度、构念效度和结论效度。

5.3.1. Internal Validity5.3.1. 内部效度

Internal validity refers to whether the observed results are truly caused by the investigated factors rather than external influences. In this study, potential biases in data selection, researcher interpretations, and thematic mapping could pose threats. Additionally, using a traditional literature review instead of a systematic review (SLR) increases the risk of selection bias, incomplete coverage, and limited reproducibility. To mitigate these threats, we applied systematic data coding, cross-validated findings with external reviewers, documented the review process step by step, and assessed interpersonal bias using Kendall’s coefficient of concordance analysis.内部效度指的是观察到的结果是否真正由所研究的因素引起，而非外部影响所致。在本研究中，数据选择、研究者解释和主题映射中的潜在偏差可能构成威胁。此外，使用传统文献综述而非系统综述（SLR）增加了选择偏差、覆盖不完整和可重复性有限的风险。为减轻这些威胁，我们采用了系统的数据编码，与外部审阅者交叉验证研究结果，逐步记录综述过程，并使用肯德尔和谐系数分析评估人际偏差。

5.3.2. External Validity5.3.2. 外部效度

External validity refers to the extent to which the study findings can be generalized. In this study, the identified motivators and demotivators were derived from a traditional literature review, which may limit the breadth of included studies. Additionally, the thematic mapping, while structured, is influenced by selected studies and may not capture all possible themes. To mitigate this, we ensured diverse sources, followed a structured review process, and validated the thematic mapping through a formal thematic approach and multiple reviewers (authors) to reduce bias.外部效度指的是研究结果可推广的程度。在本研究中，已识别的激励因素和阻碍因素来源于传统文献综述，这可能限制了所纳入研究的广度。此外，主题映射虽有结构性，但受所选研究的影响，可能无法涵盖所有可能的主题。为减轻这一问题，我们确保了来源的多样性，遵循了结构化的综述过程，并通过正式的主题分析方法和多位审阅者（作者）验证了主题映射，以减少偏差。

5.3.3. Construct Validity5.3.3. 构念效度

Construct validity assesses whether the study accurately measures the intended concepts. In this study, threats may arise from misclassification of motivators and demotivators or inconsistencies in thematic mapping, which could affect the reliability of identified themes. To mitigate this, we followed a formal thematic analysis process, used clearly defined coding schemes, and involved multiple reviewers to ensure consistency and reduce subjective bias.构念效度评估研究是否准确测量了预期的概念。在本研究中，威胁可能源于激励因素和阻碍因素的错误分类或主题映射的不一致，这可能影响所识别主题的可靠性。为减轻这一问题，我们遵循了正式的主题分析流程，使用了明确定义的编码方案，并邀请了多位审阅者参与，以确保一致性并减少主观偏差。

5.3.4. Conclusion Validity5.3.4. 结论效度

Conclusion validity ensures that the study’s findings and inferences are logically sound and well-supported by the data. Threats in this study may arise from biases in data interpretation, limited dataset coverage, or overfitting of thematic patterns, potentially leading to misleading conclusions. To mitigate these threats, we followed a structured review process, critically evaluated emerging patterns through consensus meetings among all authors, and ensured clear documentation of the review process, including a replication package (Khan, 2025) for transparency and reproducibility.结论效度确保研究的发现和推论在逻辑上合理并得到数据的充分支持。本研究中的威胁可能源于数据解释的偏差、数据集覆盖范围有限或主题模式的过度拟合，可能导致误导性结论。为减轻这些威胁，我们遵循了结构化的综述流程，通过所有作者的共识会议批判性评估新兴模式，并确保清晰记录综述过程，包括用于透明度和可重复性的复制包（Khan, 2025）。

6. Related Work6. 相关工作

The integration of LLMs in education has been extensively studied, with various works focusing on their pedagogical applications, ethical concerns, and impact on student learning. However, while prior research provides valuable insights into the broader implications of LLMs in education, most studies do not specifically analyze the motivators and demotivators of LLM adoption in software engineering education within Finnish higher education institutions.LLM在教育中的整合已被广泛研究，许多工作聚焦于其教学应用、伦理问题和对学生学习的影响。然而，尽管先前的研究对LLM在教育中的广泛影响提供了宝贵的见解，但大多数研究并未专门分析芬兰高等教育机构中软件工程教育采用LLM的激励因素和阻碍因素。

For instance, study conducted by Arto et al. (Hellas et al., 2023) examines how LLMs such as OpenAI Codex and GPT-3.5 respond to programming help requests in an introductory course at Aalto University, Finland. This study evaluates the accuracy, reliability, and limitations of LLM-generated feedback, finding that while GPT-3.5 identifies errors effectively, it frequently produces false positives and unnecessary suggestions. The study highlights both the benefits and risks of LLM-assisted learning in programming education, but it does not investigate the broader challenges of integrating LLMs into structured curricula nor the pedagogical motivators and demotivators of their adoption in software engineering education. Samuli et al. (Laato et al., 2023) explores the potential and challenges of AI-driven learning tools in Finnish university education. It systematically maps ChatGPT’s capabilities and identifies thirteen key implications for AI-assisted learning, including its benefits in personalized tutoring, accessibility, and interactive learning. However, the study also raises concerns about academic integrity, over-reliance on AI, and the need for educators to adapt teaching strategies to integrate LLMs effectively. While this work provides valuable insights into the general use of LLMs in higher education, it does not explicitly address their impact on software engineering education. Reza (FattahiBavandpour, 2024) conducts a systematic review of LLMs in education, focusing on their scalability, ethical concerns, and implementation strategies. This study examines the applications of AI chatbots, automated feedback, and personalized learning, while also addressing challenges related to bias, data privacy, and responsible AI use. The research presents a roadmap for integrating LLMs into Finland’s education system, particularly in the context of higher education institutions. However, it does not provide an examination of specific motivators and demotivators nor does it focus on software engineering students’ learning experiences and assessment-related challenges. Melany et al. (Macias et al., 2024) presents two LLM-based educational tools developed at Metropolia University of Applied Sciences in Finland. These tools include a Moodle11 1 https://moodle.org/ AI plugin designed to assist teachers in managing course content and a curriculum planning tool that aligns course descriptions with strategic educational goals. This study takes a usability-focused, user-centric approach to integrating AI into existing educational systems while ensuring compliance with data protection regulations such as GDPR22 2 https://gdpr-info.eu/. While this research highlights the practical applications of AI-driven tools for educators, it does not investigate how LLMs impact students’ learning experiences, particularly in software engineering education.例如，Arto等人（Hellas et al., 2023）开展的研究考察了OpenAI Codex和GPT-3.5等LLM如何响应芬兰阿尔托大学入门课程中的编程帮助请求。该研究评估了LLM生成反馈的准确性、可靠性和局限性，发现虽然GPT-3.5能有效识别错误，但经常产生误报和不必要的建议。该研究强调了LLM辅助学习在编程教育中的益处和风险，但未调查将LLM整合到结构化课程中的更广泛挑战，也未探讨其在软件工程教育中采用的教学激励因素和阻碍因素。 Samuli等人（Laato et al., 2023）探讨了AI驱动学习工具在芬兰大学教育中的潜力和挑战。该研究系统性地梳理了ChatGPT的能力，并确定了AI辅助学习的十三个关键影响，包括其在个性化辅导、可访问性和互动学习方面的优势。然而，该研究也提出了对学术诚信、过度依赖AI以及教育工作者需要调整教学策略以有效整合LLM的担忧。虽然这项工作为LLM在高等教育中的普遍使用提供了宝贵见解，但并未明确阐述其对软件工程教育的影响。 Reza（FattahiBavandpour, 2024）对教育中的LLM进行了系统综述，重点关注其可扩展性、伦理问题和实施策略。该研究考察了AI聊天机器人、自动反馈和个性化学习的应用，同时解决了与偏见、数据隐私和负责任AI使用相关的挑战。该研究为将LLM整合到芬兰教育系统（尤其是高等教育机构背景下）提供了路线图。然而，它并未对具体的激励因素和阻碍因素进行考察，也未关注软件工程学生的学习体验和与评估相关的挑战。 Melany等人（Macias et al., 2024）介绍了在芬兰Metropolia应用科学大学开发的两款基于LLM的教育工具。这些工具包括一个Moodle AI插件（https://moodle.org/），旨在帮助教师管理课程内容，以及一个课程规划工具，使课程描述与战略教育目标保持一致。该研究采用以用户为中心的可用性导向方法，将AI整合到现有教育系统中，同时确保符合数据保护法规，如GDPR（https://gdpr-info.eu/）。虽然这项研究突出了AI驱动工具对教育工作者的实际应用，但并未调查LLM如何影响学生的学习体验，尤其是在软件工程教育方面。

6.1. Comparison with Proposed Work6.1. 与拟议工作的比较

Now we provide the comparison between the proposed study and the related studies discussed in the previous section (See Table 3). The proposed study differentiates itself by reviewing the state-of-the-art literature on motivators and demotivators of using LLMs in software engineering education, providing a structured thematic mapping of factors that influence their adoption. Unlike previous studies that primarily focus on ChatGPT general applications in education or the usability of AI tools for educators, this research examines LLMs from the perspective of both students and educators in the domain of software engineering. It provides a literature review of 25 motivators and 30 demotivators, categorizing them into themes such as enhancing learning experiences, assessment and feedback, collaboration and peer learning, and skill development. Furthermore, the study extends beyond theoretical discussions by offering a roadmap (understanding of motivators and demotivator factors) that Finnish universities can use to integrate LLMs effectively into their software engineering curricula while addressing key concerns such as over-reliance, academic integrity, and ethical AI use.现在我们提供拟议研究与前一节中讨论的相关研究之间的比较（见表3）。拟议研究的区别在于，它回顾了在软件工程教育中使用LLM的激励因素和阻碍因素的最新文献，并对影响其采用的因素进行了结构化的主题映射。与先前主要关注ChatGPT在教育中的普遍应用或AI工具对教育工作者的可用性的研究不同，本研究从软件工程领域的学生和教育工作者两个角度审视LLM。它提供了对25个激励因素和30个阻碍因素的文献综述，并将其归类为诸如增强学习体验、评估与反馈、协作与同伴学习以及技能发展等主题。此外，该研究超越了理论讨论，提供了芬兰大学可以使用的路线图（对激励因素和阻碍因素的理解），以有效地将LLM整合到其软件工程课程中，同时解决过度依赖、学术诚信和伦理AI使用等关键问题。

Table 3. Proposed Study Comparison With Related Studies. Note: (✓: included, ✗: not included, *: abstract overview, SE: Software Engineering)表3. 拟议研究与相关研究的比较。注：（✓：包含，✗：不包含，*：摘要概述，SE：软件工程）
Study Focus Area Motivators/Demotivators SE Education Finland Context Review Paper
Arto et al. (Hellas et al., 2023) LLMs in coding education ✗ ✓(*) ✓ ✗
Samuli et al. (Laato et al., 2023) AI in university education ✗ ✗ ✓ ✗
Reza (FattahiBavandpour, 2024) AI Chatbots & Business Models ✗ ✗ ✓ ✓
Melany et al. (Macias et al., 2024) AI tools for digital pedagogy ✗ ✗ ✓ ✗
Proposed Study LLMs in SE Education ✓ ✓ ✓ ✓

7. Conclusions and Future Work7. 结论与未来工作

This study examined the integration of LLMs in software engineering education by identifying key motivators and demotivators from the literature. The findings reveal that LLMs provide several advantages, including personalized learning, enhanced engagement, and automated assessments, making education more efficient and adaptive. These benefits indicate that LLMs can significantly improve the learning experience by offering real-time feedback, supporting skill development, and facilitating collaborative learning environments. However, the study also uncovered major concerns, such as ethical issues, biases in AI-generated content, the risk of over-reliance on automation, and challenges in academic integrity. These demotivators highlight the need for a structured and well-regulated approach to integrating LLMs into education, ensuring that they complement rather than replace essential problem-solving and critical thinking skills.本研究通过从文献中识别关键的激励因素和阻碍因素，考察了LLM在软件工程教育中的整合。研究结果表明，LLM提供了若干优势，包括个性化学习、增强的参与度和自动化评估，使教育更加高效和适应性更强。这些益处表明，LLM通过提供实时反馈、支持技能发展和促进协作学习环境，能够显著改善学习体验。然而，研究也揭示了重大问题，如伦理问题、AI生成内容中的偏见、过度依赖自动化的风险以及学术诚信方面的挑战。这些阻碍因素凸显了以结构化且规范化的方式将LLM整合到教育中的必要性，确保其补充而非取代基本的问题解决和批判性思维技能。

Building on these findings, the next phase of this research will involve empirical studies, including surveys, interviews, and case studies, to validate and expand upon the insights gained from the literature review in the domain of Finnish higher education institutes. These studies will provide deeper insights into how LLMs are perceived and utilized in real-world educational settings, helping to refine the understanding of both their advantages and challenges. The ultimate objective is to develop a well-defined framework tailored for Finnish higher education institutions, offering guidelines for responsible and effective LLM integration. This framework will ensure that LLMs enhance teaching and learning outcomes while addressing potential risks and maintaining educational integrity. Through this structured approach, the research aims to contribute to the ongoing discussion on AI in education and support institutions in making informed decisions about LLM adoption.在这些发现的基础上，本研究的下一阶段将涉及实证研究，包括调查、访谈和案例研究，以验证和扩展从芬兰高等教育机构领域的文献综述中获得的见解。这些研究将更深入地了解LLM在真实教育环境中如何被感知和使用，有助于完善对其优势和挑战的理解。最终目标是为芬兰高等教育机构量身定制一个明确的框架，为负责任且有效的LLM整合提供指导方针。该框架将确保LLM增强教学成果，同时应对潜在风险并维护教育诚信。通过这种结构化方法，本研究旨在为关于AI在教育中应用的持续讨论做出贡献，并支持机构就LLM的采用做出明智决策。

8. Acknowledgments and Authors Contribution8. 致谢与作者贡献

The authors acknowledge the use of generative AI tools, such as ChatGPT, along with the writing assistant Grammarly, and writefull (Overleaf tool), to enhance the clarity and coherence of this paper. However, these tools were employed as aids rather than replacements for the authors input. All authors performed significant revisions and thorough reviews to refine the content, and the authors assume full responsibility for the final version of the manuscript.
Maryam Khan (PhD Researcher) was primarily responsible for data collection, analysis, and reporting, with equal assistance from Muhammad Azeem Akbar (Co-Supervisor). The third author, Jussi Kasurinen (Main Supervisor), contributed to reviewing, writing, and finalizing the manuscript.作者感谢使用生成式AI工具（如ChatGPT）以及写作辅助工具Grammarly和Writefull（Overleaf工具）来提高本文的清晰度和连贯性。然而，这些工具仅是辅助手段，而非替代作者的贡献。所有作者都对完善内容进行了重要修订和全面审阅，并对稿件的最终版本负全部责任。 Maryam Khan（博士研究员）主要负责数据收集、分析和报告，Muhammad Azeem Akbar（共同导师）提供了同等协助。第三作者Jussi Kasurinen（主要导师）参与了稿件的审阅、撰写和定稿。

References

  • Acemoglu and Restrepo (2018) Daron Acemoglu and Pascual Restrepo. 2018. Artificial intelligence, automation, and work. In The economics of artificial intelligence: An agenda. University of Chicago Press, 197–236.
  • Akbar et al. (2023) Muhammad Azeem Akbar, Arif Ali Khan, and Peng Liang. 2023. Ethical aspects of ChatGPT in software engineering research. IEEE Transactions on Artificial Intelligence (2023).
  • Banerjee et al. (2025) Paramarshi Banerjee, Anurag K Srivastava, Donald A Adjeroh, Ramana Reddy, and Nima Karimian. 2025. Understanding chatgpt: Impact analysis and path forward for teaching computer science and engineering. IEEE Access (2025).
  • Belchev (2025) Boris Belchev. 2025. Code Green: Evaluating the Carbon and Energy Implications of LLM Integration in Software Development. Master’s thesis. University of Twente.
  • Bender et al. (2021) Emily M Bender, Timnit Gebru, Angelina McMillan-Major, and Shmargaret Shmitchell. 2021. On the dangers of stochastic parrots: Can language models be too big?. In Proceedings of the 2021 ACM conference on fairness, accountability, and transparency. 610–623.
  • Boguslawski et al. (2025) Samuel Boguslawski, Rowan Deer, and Mark G Dawson. 2025. Programming education and learner motivation in the age of generative AI: student and educator perspectives. Information and Learning Sciences 126, 1/2 (2025), 91–109.
  • Braun and Clarke (2006) Virginia Braun and Victoria Clarke. 2006. Using thematic analysis in psychology. Qualitative research in psychology 3, 2 (2006), 77–101.
  • Cambaz and Zhang (2024) Doga Cambaz and Xiaoling Zhang. 2024. Use of AI-driven code generation models in teaching and learning programming: a systematic literature review. In Proceedings of the 55th ACM Technical Symposium on Computer Science Education V. 1. 172–178.
  • Cao (2022) Longbing Cao. 2022. Ai in finance: challenges, techniques, and opportunities. ACM Computing Surveys (CSUR) 55, 3 (2022), 1–38.
  • Cowan et al. (2023) Brendan Cowan, Yutaka Watanobe, and Atsushi Shirafuji. 2023. Enhancing programming learning with llms: Prompt engineering and flipped interaction. In Proceedings of the 2023 4th Asia Service Sciences and Software Engineering Conference. 10–16.
  • Daun and Brings (2023) Marian Daun and Jennifer Brings. 2023. How ChatGPT will change software engineering education. In Proceedings of the 2023 Conference on Innovation and Technology in Computer Science Education V. 1. 110–116.
  • Fan et al. (2023) Angela Fan, Beliz Gokkaya, Mark Harman, Mitya Lyubarskiy, Shubho Sengupta, Shin Yoo, and Jie M Zhang. 2023. Large language models for software engineering: Survey and open problems. In 2023 IEEE/ACM International Conference on Software Engineering: Future of Software Engineering (ICSE-FoSE). IEEE, 31–53.
  • FattahiBavandpour (2024) Reza FattahiBavandpour. 2024. Advancing Education with Large Language Models: A Systematic Review of Potential, Limitations, and Business Opportunities. Master’s thesis. Lappeenranta–Lahti University of Technology LUT, Finland.
  • Feldt et al. (2018) Robert Feldt, Francisco G de Oliveira Neto, and Richard Torkar. 2018. Ways of applying artificial intelligence in software engineering. In Proceedings of the 6th International Workshop on Realizing Artificial Intelligence Synergies in Software Engineering. 35–41.
  • Forbus and Laird (2002) Kenneth D Forbus and John Laird. 2002. Guest editors’ introduction: AI and the entertainment industry. IEEE Intelligent Systems 17, 04 (2002), 15–16.
  • Fuligni et al. (2025) Caterina Fuligni, Daniel Dominguez Figaredo, and Julia Stoyanovich. 2025. " Would You Want an AI Tutor?" Understanding Stakeholder Perceptions of LLM-based Chatbots in the Classroom. arXiv preprint arXiv:2503.02885 (2025).
  • Gao et al. (2024) Rong Gao, Qin Ni, and Bingying Hu. 2024. Fairness of Large Language Models in Education. In Proceedings of the 2024 International Conference on Intelligent Education and Computer Technology. 1.
  • Glaser and Strauss (2017) Barney Glaser and Anselm Strauss. 2017. Discovery of grounded theory: Strategies for qualitative research. Routledge.
  • Hanifi et al. (2023) Khadija Hanifi, Orcun Cetin, and Cemal Yilmaz. 2023. On chatgpt: Perspectives from software engineering students. In 2023 IEEE 23rd International Conference on Software Quality, Reliability, and Security (QRS). IEEE, 196–205.
  • Harvey et al. (2025) Emma Harvey, Allison Koenecke, and Rene F Kizilcec. 2025. " Don’t Forget the Teachers": Towards an Educator-Centered Understanding of Harms from Large Language Models in Education. arXiv preprint arXiv:2502.14592 (2025).
  • He et al. (2024) Junda He, Christoph Treude, and David Lo. 2024. LLM-Based Multi-Agent Systems for Software Engineering: Literature Review, Vision and the Road Ahead. ACM Transactions on Software Engineering and Methodology (2024).
  • Hellas et al. (2023) Arto Hellas, Juho Leinonen, Sami Sarsa, Charles Koutcheme, Lilja Kujanpää, and Juha Sorva. 2023. Exploring the responses of large language models to beginner programmers’ help requests. In Proceedings of the 2023 ACM Conference on International Computing Education Research-Volume 1. 93–105.
  • Herden et al. (2024) Olaf Herden et al. 2024. Integration of Chatbots for Generating Code into Introductory Programming Courses. In Conference Proceedings. The Future of Education 2024.
  • Hussein (2024) Dania Hussein. 2024. Usability of LLMs for Assisting Software Engineering: a Literature Review. (2024).
  • Jalil et al. (2023) Sajed Jalil, Suzzana Rafi, Thomas D LaToza, Kevin Moran, and Wing Lam. 2023. Chatgpt and software testing education: Promises & perils. In 2023 IEEE international conference on software testing, verification and validation workshops (ICSTW). IEEE, 4130–4137.
  • Jin et al. (2025) Yueqiao Jin, Lixiang Yan, Vanessa Echeverria, Dragan Gašević, and Roberto Martinez-Maldonado. 2025. Generative AI in higher education: A global perspective of institutional adoption policies and guidelines. Computers and Education: Artificial Intelligence 8 (2025), 100348.
  • Joshi et al. (2023) Ishika Joshi, Ritvik Budhiraja, Pranav Deepak Tanna, Lovenya Jain, Mihika Deshpande, Arjun Srivastava, Srinivas Rallapalli, Harshal D Akolekar, Jagat Sesh Challa, and Dhruv Kumar. 2023. " With Great Power Comes Great Responsibility!": Student and Instructor Perspectives on the influence of LLMs on Undergraduate Engineering Education. arXiv preprint arXiv:2309.10694 (2023).
  • Jošt et al. (2024) Gregor Jošt, Viktor Taneski, and Sašo Karakatič. 2024. The impact of large language models on programming education and student learning outcomes. Applied Sciences 14, 10 (2024), 4115.
  • Khan et al. (2023) Arif Ali Khan, Muhammad Azeem Akbar, Mahdi Fahmideh, Peng Liang, Muhammad Waseem, Aakash Ahmad, Mahmood Niazi, and Pekka Abrahamsson. 2023. AI ethics: an empirical study on the views of practitioners and lawmakers. IEEE Transactions on Computational Social Systems 10, 6 (2023), 2971–2984.
  • Khan et al. (2022) Arif Ali Khan, Sher Badshah, Peng Liang, Muhammad Waseem, Bilal Khan, Aakash Ahmad, Mahdi Fahmideh, Mahmood Niazi, and Muhammad Azeem Akbar. 2022. Ethics of AI: A systematic literature review of principles and challenges. In Proceedings of the 26th international conference on evaluation and assessment in software engineering. 383–392.
  • Khan et al. (2021) Arif Ali Khan, Mohammad Shameem, Mohammad Nadeem, and Muhammad Azeem Akbar. 2021. Agile trends in Chinese global software development industry: Fuzzy AHP based conceptual mapping. Applied Soft Computing 102 (2021), 107090.
  • Khan (2025) M. Khan. 2025. Motivators and Demotivators of using LLMs in Software Engineering Education. https://doi.org/10.5281/zenodo.15013312 [Data set].
  • Kharrufa et al. (2024) Ahmed Kharrufa, Sami Alghamdi, Abeer Aziz, and Christopher Bull. 2024. LLMs Integration in Software Engineering Team Projects: Roles, Impact, and a Pedagogical Design Space for AI Tools in Computing Education. arXiv preprint arXiv:2410.23069 (2024).
  • Kirova et al. (2024) Vassilka D Kirova, Cyril S Ku, Joseph R Laracy, and Thomas J Marlowe. 2024. Software engineering education must adapt and evolve for an llm environment. In Proceedings of the 55th ACM Technical Symposium on Computer Science Education V. 1. 666–672.
  • Kitchenham et al. (2007) Barbara Kitchenham, Stuart Charters, et al. 2007. Guidelines for performing systematic literature reviews in software engineering.
  • Kitchenham et al. (2002) Barbara A Kitchenham, Shari Lawrence Pfleeger, Lesley M Pickard, Peter W Jones, David C. Hoaglin, Khaled El Emam, and Jarrett Rosenberg. 2002. Preliminary guidelines for empirical research in software engineering. IEEE Transactions on software engineering 28, 8 (2002), 721–734.
  • Korpimies et al. (2024) Kai Korpimies, Antti Laaksonen, and Matti Luukkainen. 2024. Unrestricted use of LLMs in a software project course: Student perceptions on learning and impact on course performance. In Proceedings of the 24th Koli Calling International Conference on Computing Education Research. 1–7.
  • Laato et al. (2023) Samuli Laato, Benedikt Morschheuser, Juho Hamari, and Jari Björne. 2023. AI-assisted learning with ChatGPT and large language models: Implications for higher education. In 2023 IEEE International Conference on Advanced Learning Technologies (ICALT). IEEE, 226–230.
  • Lo (2023) Chung Kwan Lo. 2023. What is the impact of ChatGPT on education? A rapid review of the literature. Education sciences 13, 4 (2023), 410.
  • Lyu et al. (2024) Wenhan Lyu, Yimeng Wang, Tingting Chung, Yifan Sun, and Yixuan Zhang. 2024. Evaluating the effectiveness of llms in introductory computer science education: A semester-long field study. In Proceedings of the Eleventh ACM Conference on Learning@ Scale. 63–74.
  • Macias et al. (2024) Melany Vanessa Macias, Lev Kharlashkin, Leo Einari Huovinen, and Mika Hämäläinen. 2024. Empowering Teachers with Usability-Oriented LLM-Based Tools for Digital Pedagogy. In Proceedings of the 4th International Conference on Natural Language Processing for Digital Humanities. 549–557.
  • Marques et al. (2024) Nuno Marques, Rodrigo Rocha Silva, and Jorge Bernardino. 2024. Using chatgpt in software requirements engineering: A comprehensive review. Future Internet 16, 6 (2024), 180.
  • Meißner et al. (2024) Niklas Meißner, Sandro Speth, Julian Kieslinger, and Steffen Becker. 2024. Evalquiz–llm-based automated generation of self-assessment quizzes in software engineering education. In Software Engineering im Unterricht der Hochschulen 2024. Gesellschaft für Informatik eV, 53–64.
  • Niazi (2015) Mahmood Niazi. 2015. Do systematic literature reviews outperform informal literature reviews in the software engineering domain? An initial case study. Arabian Journal for Science and Engineering 40 (2015), 845–855.
  • Ozkaya (2023) Ipek Ozkaya. 2023. Application of large language models to software engineering tasks: Opportunities, risks, and implications. IEEE Software 40, 3 (2023), 4–8.
  • Pereira et al. (2024) Juanan Pereira, Juan-Miguel López, Xabier Garmendia, and Maider Azanza. 2024. Leveraging Open Source LLMs for Software Engineering Education and Training. In 2024 36th International Conference on Software Engineering Education and Training (CSEE&T). IEEE, 1–10.
  • Punter et al. (2003) Teade Punter, Marcus Ciolkowski, Bernd Freimut, and Isabel John. 2003. Conducting on-line surveys in software engineering. In 2003 International Symposium on Empirical Software Engineering, 2003. ISESE 2003. Proceedings. IEEE, 80–88.
  • Rasnayaka et al. (2024) Sanka Rasnayaka, Guanlin Wang, Ridwan Shariffdeen, and Ganesh Neelakanta Iyer. 2024. An empirical study on usage and perceptions of llms in a software engineering project. In Proceedings of the 1st International Workshop on Large Language Models for Code. 111–118.
  • Runeson and Höst (2009) Per Runeson and Martin Höst. 2009. Guidelines for conducting and reporting case study research in software engineering. Empirical software engineering 14 (2009), 131–164.
  • Sallou et al. (2024) June Sallou, Thomas Durieux, and Annibale Panichella. 2024. Breaking the silence: the threats of using llms in software engineering. In Proceedings of the 2024 ACM/IEEE 44th International Conference on Software Engineering: New Ideas and Emerging Results. 102–106.
  • Samhan et al. (2024) Ahmad Samhan, Suhair AlHajHassan, Sara Abu Dabaa’t, and Ali Elrashidi. 2024. A Review of AI-Assisted Impact Analysis for Software Requirements Change: Challenges and Future Directions. In 2024 25th International Arab Conference on Information Technology (ACIT). IEEE, 1–13.
  • Santos et al. (2024) Patricia Santos, Keysha Urgel, and Verónica Moreno. 2024. Generative Artificial Intelligence in teaching and learning of ICT engineering education: A literature review and illustrative scenarios. In 2024 47th MIPRO ICT and Electronics Convention (MIPRO). IEEE, 1338–1343.
  • Schiff (2022) Daniel Schiff. 2022. Education for AI, not AI for education: The role of education and ethics in national AI policy strategies. International Journal of Artificial Intelligence in Education 32, 3 (2022), 527–563.
  • Shaheen (2021) Mohammed Yousef Shaheen. 2021. Applications of Artificial Intelligence (AI) in healthcare: A review. ScienceOpen Preprints (2021).
  • Sidney (1957) Siegel Sidney. 1957. Nonparametric statistics for the behavioral sciences. The Journal of Nervous and Mental Disease 125, 3 (1957), 497.
  • Toma and Yánez-Pérez (2024) Radu Bogdan Toma and Iraya Yánez-Pérez. 2024. Effects of ChatGPT use on undergraduate students’ creativity: a threat to creative thinking? Discover Artificial Intelligence 4, 1 (2024), 74.
  • Vierhauser et al. (2024) Michael Vierhauser, Iris Groher, Tobias Antensteiner, and Clemens Sauerwein. 2024. Towards integrating emerging ai applications in se education. In 2024 36th International Conference on Software Engineering Education and Training (CSEE&T). IEEE, 1–5.
  • Wang et al. (2025) Minkai Wang, Di Zhang, Jingdong Zhu, and Hanjie Gu. 2025. Effects of Incorporating a Large Language Model-Based Adaptive Mechanism Into Contextual Games on Students’ Academic Performance, Flow Experience, Cognitive Load and Behavioral Patterns. Journal of Educational Computing Research (2025), 07356331251321719.
  • Wang et al. (2024) Tianyu Wang, Nianjun Zhou, and Zhixiong Chen. 2024. Enhancing computer programming education with llms: A study on effective prompt engineering for python code generation. arXiv preprint arXiv:2407.05437 (2024).
  • Wohlin (2014) Claes Wohlin. 2014. Guidelines for snowballing in systematic literature studies and a replication in software engineering. In Proceedings of the 18th international conference on evaluation and assessment in software engineering. 1–10.
  • Wohlin et al. (2012) Claes Wohlin, Per Runeson, Martin Höst, Magnus C Ohlsson, Björn Regnell, Anders Wesslén, et al. 2012. Experimentation in software engineering. Vol. 236. Springer.
  • Yao et al. (2024) Yifan Yao, Jinhao Duan, Kaidi Xu, Yuanfang Cai, Zhibo Sun, and Yue Zhang. 2024. A survey on large language model (llm) security and privacy: The good, the bad, and the ugly. High-Confidence Computing (2024), 100211.
  • Zhai et al. (2021) Xuesong Zhai, Xiaoyan Chu, Ching Sing Chai, Morris Siu Yung Jong, Andreja Istenic, Michael Spector, Jia-Bao Liu, Jing Yuan, and Yan Li. 2021. A Review of Artificial Intelligence (AI) in Education from 2010 to 2020. Complexity 2021, 1 (2021), 8812542.
  • Zhang et al. (2023) Quanjun Zhang, Chunrong Fang, Yang Xie, Yaxin Zhang, Yun Yang, Weisong Sun, Shengcheng Yu, and Zhenyu Chen. 2023. A survey on large language models for software engineering. arXiv preprint arXiv:2312.15223 (2023).