Pith. sign in

REVIEW 4 major objections 6 minor 28 references

This paper argues that generative AI adoption in German software engineering is widespread, but its benefits are moderated by developer experience, organizational size, and the AI's limited grasp of project context.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

T0 review · deepseek-v4-flash

2026-08-03 08:28 UTC pith:23TNNGOJ

load-bearing objection A well-organized exploratory study whose abstract overclaims: the moderation and 'most significant barrier' claims don't survive contact with the paper's own statistics. the 4 major comments →

arxiv 2601.16700 v2 pith:23TNNGOJ submitted 2026-01-23 cs.SE cs.AIcs.ETcs.HC

Adoption of Generative Artificial Intelligence in the German Software Engineering Industry: An Empirical Study

Ludwig Felder , Tobias Eisenreich , Mahsa Fischer , Stefan Wagner , Chunyang Chen This is my paper
classification cs.SE cs.AIcs.ETcs.HC
keywords Generative AISoftware EngineeringDeveloper ProductivityEmpirical StudyAI-Assisted DevelopmentHuman-AI InteractionExperience ParadoxContext Wall
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper sets out to explain why wide adoption of generative AI tools has not translated into uniform productivity gains among German software engineers. Using 18 interviews and a survey of 109 practitioners, it claims that experience level moderates how much value developers perceive: junior engineers treat AI as an oracle and rate specific instructions highly, while senior engineers remain skeptical. It also finds that company size shapes tool choice and usage intensity, and that the most significant barrier is the AI's limited awareness of the full project context—what the authors call the 'Context Wall.' The authors conclude that the value of GenAI depends on effective integration into workflows and organizational capabilities, not merely on having the tool. A sympathetic reader would care because this shifts the question from 'how many developers use AI' to 'under which conditions does AI actually help.'

Core claim

On its own terms, the paper's central claim is that the effectiveness of generative AI in German software engineering is moderated by three factors: developer experience, organizational size, and context awareness. The 'Experience Paradox' holds that juniors and seniors perceive AI benefits differently (juniors rate specific instructions 78% effective, seniors 39%). The 'Corporate Infrastructure Split' shows self-hosted adoption is bimodal by firm size, and code-generation frequency drops as company size grows. The 'Context Wall'—the AI's limited awareness of the codebase and its knowledge cutoff—is the most severe barrier, imposing a 'verification tax' that correlates negatively with workfl

What carries the argument

The paper's argument rests on an exploratory sequential mixed-methods design: 18 semi-structured interviews produced four themes, which were then operationalized into a survey of 109 German software engineers. The quantitative analysis uses cross-tabulations with chi-square tests (e.g., experience vs. perceived effectiveness of specific instructions), Spearman correlations between prompting strategies and perceived impacts, and k-means clustering (k=2) on self-reported usage frequency to derive usage profiles. The named patterns—Experience Paradox, Context Wall, Corporate Infrastructure Split, Communication Dividend, Proficiency Cycle—are the interpretive machinery that connects raw frequenc

Load-bearing premise

The paper's population-level claims rest on a convenience sample of 109 mostly senior German engineers (88% Germany, 62% developers) that may not represent the broader German software industry; the authors acknowledge this external-validity threat in Section 7.

What would settle it

A representative random sample of German software engineers (drawn from company rosters, not LinkedIn) that fails to reproduce the Experience Paradox—e.g., if senior engineers rate specific instructions as effective as juniors do—would falsify the moderating role of experience. Likewise, if a replication finds no significant chi-square relationship between experience and perceived effectiveness of specific instructions (p>0.05), the central claim weakens.

Watch this falsifier. Get emailed when new claim-graph text bears on it.

If this is right

  • If the Experience Paradox holds, teams should expect junior and senior developers to disagree about AI value, and organizations will need structured knowledge exchange rather than assuming a shared perception.
  • If the Context Wall is the main barrier, tool vendors should prioritize full-project grounding (repository-wide context, up-to-date knowledge) over incremental prompt features; the paper explicitly recommends guided exploration and user-driven context specification.
  • If the Corporate Infrastructure Split is real, one-size-fits-all enterprise AI policy is mismatched: small and medium firms may adopt lightweight self-hosted models while mid-sized enterprises lag, so adoption support should be tailored by firm size.
  • If the Proficiency Cycle persists, productivity gains will concentrate in a subset of 'power users,' making AI-related knowledge a shared organizational resource rather than an individual skill.
  • If perceived productivity gains are real despite the verification tax, the net value of GenAI depends on reducing the cognitive cost of validation, not just generating more code.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • Extension: The paper's own correlation data (workflow speed with distrust, ρ=-0.33) leaves causality open; one testable extension is a longitudinal study that tracks whether heavy use erodes or reinforces trust, and whether the 'verification tax' compounds over time.
  • Extension: The bimodal Ollama adoption may reflect that mid-sized enterprises (1,000–9,999 employees) are too large for lightweight self-hosting but too small for dedicated infrastructure; a concrete test would compare tool-selection criteria and compliance budgets across these bands.
  • Extension: If context awareness is the binding constraint, then agentic tools that automatically retrieve repository context could narrow the junior-senior gap—but the Experience Paradox suggests seniors may remain skeptical, so the gap may persist even with better grounding.
  • Extension: The sample's seniority skew (mean 12.1 years) means the 'power users with lower education' finding may be an artifact of that cohort; a replication with a more junior, representative sample would test whether the Proficiency Cycle generalizes.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit.

Referee Report

4 major / 6 minor

Summary. The manuscript reports a mixed-methods study of generative AI adoption in German software engineering. The authors conducted 18 semi-structured interviews followed by a survey of 109 practitioners, and analyze tool adoption, prompting strategies, challenges, and perceived productivity impact. The main claimed findings are that experience level moderates perceived benefits, that organizational size affects tool selection and usage intensity, and that limited project-context awareness is the most significant adoption barrier.

Significance. If the claims hold, the study would contribute useful evidence on the context-dependent value of GenAI in a regulated industrial setting, complementing prior international surveys. The qualitative-to-quantitative design is systematic, the interview-to-survey mapping is explicit, and the comparison with the Stack Overflow survey adds perspective. However, the statistical support for the headline moderation and 'most significant barrier' claims is currently insufficient; as presented, the paper's main value is descriptive rather than inferential.

major comments (4)
  1. [Section 5.1 / RQ3 / Abstract] The claim that experience level 'moderates' perceived benefits is not tested. The reported chi-square test (χ²(2, N=109)=10.84, p_adj=0.022) is a bivariate association between experience group and perceived effectiveness of one prompting item. Moderation requires an interaction model (e.g., experience × strategy on outcome). As written, the evidence supports only an experience-group difference on a single item. Please either conduct appropriate interaction tests (e.g., regression with interaction terms) or reframe the claim.
  2. [Section 3.4 vs. Abstract] The abstract's headline 'Limited awareness of the project context is identified as the most significant barrier' is contradicted by the manuscript's own severity ratings: hallucinations have the highest mean (3.4), and 'limited context awareness of the codebase' is second (3.3). Section 3.4 explicitly calls hallucinations 'the single most significant challenge.' This inconsistency affects the paper's central message and must be resolved.
  3. [Section 5.5 / Section 9] The conclusion that 'benefits ... are concentrated among experienced users' is not supported by the presented cluster analysis. The k-means analysis (k=2) identifies power users by usage frequency; demographic analysis shows power users tend to have lower education and work for smaller companies. Experience was not a distinguishing cluster variable. Moreover, the choice k=2 is not validated (no silhouette or gap statistic), and the clusters are based on self-reported frequency rather than benefit. Please either provide a direct analysis of experience by benefit/usage or revise the conclusion.
  4. [Figure 9 / Section 5.4] The correlation matrix reports 10×5 = 50 Spearman correlations without multiple-comparison correction, and only selected p-values are given. With n ranging from 76 to 102, several significant correlations could arise by chance. Please report adjusted p-values (e.g., Benjamini-Hochberg) or confidence intervals, and avoid interpreting individual correlations as strong evidence. Also, the use of dichotomized top-box effectiveness scores as one input loses information; consider using the full ordinal scales.
minor comments (6)
  1. [Section 2.1] The sentence 'To connect our analysis to concrete findings...' is repeated verbatim; please remove the duplicate.
  2. [Section 2.2] Typos: 'shift form' should be 'shift from'; Section 2.1 has 'comapnies' instead of 'companies'.
  3. [Figures 4 and 9] The label 'uni00A0' appears to be a literal unicode escape; also 'ModeratlyEffective' is misspelled.
  4. [Section 5.5] The k-means analysis uses Likert frequency data; specify whether variables were standardized and how the number of clusters was selected.
  5. [Section 4] The comparison with the Stack Overflow survey is informal; since the authors state no direct comparison is possible, presenting it as triangulation only may be clearer.
  6. [References] Reference [5] is missing a publication year; please complete the bibliographic entry.

Circularity Check

0 steps flagged

No significant circularity: the study is observational, uses an independent survey dataset, and no fitted output is renamed as a prediction; the only author-overlap citation is background and non-load-bearing.

full rationale

This paper is an observational mixed-methods study; it contains no first-principles derivation, no fitted model whose output is reused as an input, and no equation-level reduction to its own data. The only fitted procedure is k-means clustering (k=2) on self-reported usage frequency (Section 5.5); the clusters are descriptive profiles, and the demographic comparisons (chi-square tests on education and company size) use variables external to the clustering, so the 'power users' finding is not the cluster definition itself. The exploratory sequential design (Sections 2.1-2.2) lets interview themes guide survey construction, but the survey responses come from 109 separate practitioners, so later agreement with the themes is not forced by construction. The sole author-overlap citation ([25], co-authored by C. Chen) appears in Related Work as background ('quality assurance [25]') and is not load-bearing. The paper itself (Section 7) acknowledges convenience sampling and self-report as limitations. Several internal inconsistencies exist -- the abstract's 'most significant barrier' claim conflicts with Section 3.4, where hallucinations have the highest mean (3.4) versus context awareness (3.3), and the 'experience moderates' claim in Section 5.1 rests on a bivariate chi-square rather than an interaction test -- but these are statistical/consistency concerns, not circularity. No circular step can be exhibited by quote-and-reduction.

Axiom & Free-Parameter Ledger

1 free parameters · 5 axioms · 0 invented entities

The central claims rest on self-reported survey data from a non-random convenience sample, an analyst-chosen k=2 cluster, and the assumption that interview themes were faithfully transferred into survey items. No external dataset or benchmark is used, so the findings are only as strong as the representativeness and measurement validity of the self-reports.

free parameters (1)
  • k-means cluster count (k=2) = 2
    The 'Proficiency Cycle' analysis uses k-means with k=2 chosen without reported validation (e.g., silhouette score). The power-user dichotomy is an analytical choice that directly shapes the claim that gains are uneven.
axioms (5)
  • domain assumption Self-reported Likert responses validly measure actual usage frequency, prompting effectiveness, and productivity impact.
    All quantitative conclusions rely on self-report; acknowledged in §7 as measurement error.
  • domain assumption The convenience sample of 109 engineers (88% Germany, 62% developers, mean 12.1 years experience) is representative enough for population-level claims.
    Sampling via LinkedIn/contacts (Section 2.2) is non-random; acknowledged as a threat to external validity in §7.
  • domain assumption Interview themes were accurately operationalized into survey items without loss or bias.
    Exploratory sequential design (Section 2.1) assumes the qualitative-to-quantitative transfer preserves meaning.
  • standard math Chi-square tests with N=109 have adequate expected cell counts for the reported cross-tabulations.
    Some groups (e.g., >15 years, n small) may have sparse cells; authors do not report expected counts.
  • domain assumption The k-means clustering (k=2) yields stable, interpretable clusters.
    No cluster validation (e.g., silhouette) reported; cluster labels used downstream for 'power user' comparisons.

pith-pipeline@v1.3.0-alltime-deepseek · 16711 in / 12744 out tokens · 116275 ms · 2026-08-03T08:28:03.817770+00:00 · methodology

0 comments
read the original abstract

Generative artificial intelligence (GenAI) tools have seen rapid adoption among software developers. While adoption rates in the industry are rising, the underlying factors influencing the effective use of these tools, including the depth of interaction, organizational constraints, and experience-related considerations, have not been thoroughly investigated. This issue is particularly relevant in environments with stringent regulatory requirements, such as Germany, where practitioners must address the GDPR and the EU AI Act while balancing productivity gains with intellectual property considerations. Despite the significant impact of GenAI on software engineering, to the best of our knowledge, no empirical study has systematically examined the adoption dynamics of GenAI tools within the German context. To address this gap, we present a comprehensive mixed-methods study on GenAI adoption among German software engineers. Specifically, we conducted 18 exploratory interviews with practitioners, followed by a developer survey with 109 participants. We analyze patterns of tool adoption, prompting strategies, and organizational factors that influence effectiveness. Our results indicate that experience level moderates the perceived benefits of GenAI tools, and productivity gains are not evenly distributed among developers. Further, organizational size affects both tool selection and the intensity of tool use. Limited awareness of the project context is identified as the most significant barrier. We summarize a set of actionable implications for developers, organizations, and tool vendors seeking to advance artificial intelligence (AI) assisted software development.

Figures

Figures reproduced from arXiv: 2601.16700 by Chunyang Chen, Ludwig Felder, Mahsa Fischer, Stefan Wagner, Tobias Eisenreich.

Figure 1
Figure 1. Figure 1: Methodology overview of our sequential empirical study. [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 3
Figure 3. Figure 3: Frequency of GenAI usage across seven software [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Perceived effectiveness of ten prompting strategies [PITH_FULL_IMAGE:figures/full_fig_p005_4.png] view at source ↗
Figure 6
Figure 6. Figure 6: Severity of integration and customization chal [PITH_FULL_IMAGE:figures/full_fig_p006_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Perceived impact of GenAI on five dimensions of [PITH_FULL_IMAGE:figures/full_fig_p006_7.png] view at source ↗
Figure 9
Figure 9. Figure 9: Correlation of effectiveness ratings (very and ex [PITH_FULL_IMAGE:figures/full_fig_p007_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: Frequency usage by purpose and company size. [PITH_FULL_IMAGE:figures/full_fig_p008_10.png] view at source ↗

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

28 extracted references · 2 canonical work pages

  1. [1]

    Mamdouh Alenezi and Mohammed Akour. 2025. AI-Driven Innovations in Software Engineering: A Review of Current Practices and Future Directions. Applied Sciences15, 3 (2025). doi:10.3390/app15031344

    doi:10.3390/app15031344 2025

  2. [2]

    James, and Nadia Polikarpova

    Shraddha Barke, Michael B. James, and Nadia Polikarpova. 2022.Grounded Copilot: How Programmers Interact with Code-Generating Models. arXiv:2206.15000 [cs] doi:10.48550/arXiv.2206.15000

    Pith/arXiv arXiv doi:10.48550/arxiv.2206.15000 2022

  3. [3]

    2025.Measuring the Impact of Early-2025 AI on Experienced Open-Source Developer Productivity

    Joel Becker, Nate Rush, Elizabeth Barnes, and David Rein. 2025.Measuring the Impact of Early-2025 AI on Experienced Open-Source Developer Productivity. arXiv:2507.09089 [cs] doi:10.48550/arXiv.2507.09089

    Pith/arXiv arXiv doi:10.48550/arxiv.2507.09089 2025

  4. [4]

    Sayan Chatterjee, Ching Louis Liu, Gareth Rowland, and Tim Hogarth. 2024. The Impact of AI Tool on Engineering at ANZ Bank An Empirical Study on GitHub Copilot within Corporate Environment. arXiv:2402.05636 [cs.SE] https: //arxiv.org/abs/2402.05636

    Pith/arXiv arXiv 2024

  5. [5]

    Kevin Zheyuan Cui, Mert Demirer, Sonia Jaffe, Leon Musolff, Sida Peng, and Tobias Salz. [n. d.]. The Effects of Generative AI on High-Skilled Work: Evidence from Three Field Experiments with Software Developers. ([n. d.])

  6. [6]

    Nicole Davila, Igor Wiese, Igor Steinmacher, Lucas Lucio da Silva, Andre Kawamoto, Gilson Jose Peres Favaro, and Ingrid Nunes. 2024. An Industry Case Study on Adoption of AI-based Programming Assistants. InProceedings of the 46th International Conference on Software Engineering: Software Engineering in Practice(Lisbon, Portugal)(ICSE-SEIP ’24). Associatio...

    arXiv 2024

  7. [7]

    DIN e. V. and DKE. 2022.German Standardization Roadmap on Artificial In- telligence – 2nd Edition. DIN e. V. and DKE, Berlin and Offenbach am Main. https://www.din.de/go/roadmap-ai Accessed: 2025-01-18

    2022

  8. [8]

    Usman Khan Durrani, Mustafa Akpinar, Hakan Bektas, and Mohammed Saleh

  9. [9]

    2025.The Future of Generative AI in Software Engi- neering: A Vision from Industry and Academia in the European GENIUS Project

    Robin Gröpler, Steffen Klepke, Jack Johns, Andreas Dreschinski, Klaus Schmid, Benedikt Dornauer, Eray Tüzün, Joost Noppen, Mohammad Reza Mousavi, Yongjian Tang, Johannes Viehmann, Selin Şirin Aslangül, Beum Seuk Lee, Adam Ziolkowski, and Eric Zie. 2025.The Future of Generative AI in Software Engi- neering: A Vision from Industry and Academia in the Europe...

    doi:10.48550/arxiv.2511.01348 2025

  10. [10]

    Juyong Jiang, Fan Wang, Jiasi Shen, Sungju Kim, and Sunghun Kim. 2024. A Survey on Large Language Models for Code Generation.ACM Transactions on Software Engineering and Methodology(2024). doi:10.1145/3747588

    doi:10.1145/3747588 2024

  11. [11]

    Ranim Khojah, Mazen Mohamad, Philipp Leitner, and Francisco Gomes de Oliveira Neto. 2024. Beyond Code Generation: An Observational Study of Chat- GPT Usage in Software Engineering Practice. 1 (2024), 81:1819–81:1840. Issue FSE. doi:10.1145/3660788

    doi:10.1145/3660788 2024

  12. [12]

    Madhava Krishna, Bhagesh Gaur, Arsh Verma, and Pankaj Jalote. 2024. Using LLMs in Software Requirements Specifications: An Empirical Evaluation.2024 IEEE 32nd International Requirements Engineering Conference (RE)(2024), 475–483. doi:10.1109/re59067.2024.00056

    arXiv 2024

  13. [13]

    Stefano Lambiase, Gemma Catolino, Fabio Palomba, Filomena Ferrucci, and Daniel Russo. 2025. Exploring Individual Factors in the Adoption of LLMs for Specific Software Engineering Tasks. arXiv:2504.02553 [cs.SE] https://arxiv.org/ abs/2504.02553

    Pith/arXiv arXiv 2025

  14. [14]

    Stefano Lambiase, Gemma Catolino, Fabio Palomba, Filomena Ferrucci, and Daniel Russo. 2025. Investigating the Role of Cultural Values in Adopting Large Language Models for Software Engineering.ACM Trans. Softw. Eng. Methodol. 35, 1, Article 23 (Dec. 2025), 43 pages. doi:10.1145/3725529

    doi:10.1145/3725529 2025

  15. [15]

    Ze Shi Li, Nowshin Nawar Arony, Ahmed Musa Awon, Daniela Damian, and Bowen Xu. 2024. AI Tool Use and Adoption in Software Development by Indi- viduals and Organizations: A Grounded Theory Study. arXiv:2406.17325 [cs.SE] https://arxiv.org/abs/2406.17325

    Pith/arXiv arXiv 2024

  16. [16]

    Liang, Chenyang Yang, and Brad A

    Jenny T. Liang, Chenyang Yang, and Brad A. Myers. 2024. A Large-Scale Survey on the Usability of AI Programming Assistants: Successes and Challenges. In Proceedings of the IEEE/ACM 46th International Conference on Software Engineer- ing(New York, NY, USA, 2024-02-06)(ICSE ’24). Association for Computing Machinery, 1–13. doi:10.1145/3597503.3608128

    arXiv 2024

  17. [17]

    Nuno Marques, R. R. Silva, and Jorge Bernardino. 2024. Using ChatGPT in Software Requirements Engineering: A Comprehensive Review.Future Internet 16 (2024), 180. doi:10.3390/fi16060180

    doi:10.3390/fi16060180 2024

  18. [18]

    2026.Between Policy and Practice: GenAI Adoption in Agile Software Development Teams

    Michael Neumann, Lasse Bischof, Nic Elias Hinz, Luca Stockmann, Dennis Schrader, Ana Carolina Ahaus, Erim Can Demirci, Benjamin Gabel, Maria Rauschenberger, Philipp Diebold, Henning Fritzemeier, and Adam Przybylek. 2026.Between Policy and Practice: GenAI Adoption in Agile Software Development Teams. arXiv:2601.07051 [cs] doi:10.48550/arXiv.2601.07051

    work page doi:10.48550/arxiv.2601.07051 2026

  19. [19]

    André Pahnke and Friederike Welter. 2019. The German Mittelstand: Antithesis to Silicon Valley Entrepreneurship? 52, 2 (2019), 345–358. doi:10.1007/s11187- 018-0095-4

    work page doi:10.1007/s11187- 2019

  20. [20]

    DS Pashchenko. 2023. Early formalization of AI-tools usage in software engi- neering in Europe: study of 2023.International Journal of Information Technology and Computer Science15, 6 (2023), 29–36

    2023

  21. [21]

    Sida Peng, Eirini Kalliamvakou, Peter Cihon, and Mert Demirer. 2023. The Impact of AI on Developer Productivity: Evidence from GitHub Copilot. arXiv:2302.06590 [cs] doi:10.48550/arXiv.2302.06590

    Pith/arXiv arXiv doi:10.48550/arxiv.2302.06590 2023

  22. [22]

    Daniel Russo. 2024. Navigating the Complexity of Generative AI Adoption in Software Engineering. 33, 5, Article 135 (June 2024), 50 pages. doi:10.1145/3652154

    doi:10.1145/3652154 2024

  23. [23]

    Larissa Schmid, Tobias Hey, Martin Armbruster, Sophie Corallo, Dominik Fuchß, Jan Keim, Haoyu Liu, and Anne Koziolek. 2025. Software Architecture Meets LLMs: A Systematic Literature Review. arXiv:2505.16697 [cs.SE] https://arxiv. org/abs/2505.16697

    Pith/arXiv arXiv 2025

  24. [25]

    Junjie Wang, Yuchao Huang, Chunyang Chen, Zhe Liu, Song Wang, and Qing Wang. 2023. Software Testing With Large Language Models: Survey, Landscape, and Vision.IEEE Transactions on Software Engineering50 (2023), 911–936. doi:10. 1109/tse.2024.3368208

    arXiv 2023

  25. [26]

    Weisz, Shraddha Vijay Kumar, Michael Muller, Karen-Ellen Browne, Arielle Goldberg, Katrin Ellice Heintze, and Shagun Bajpai

    Justin D. Weisz, Shraddha Vijay Kumar, Michael Muller, Karen-Ellen Browne, Arielle Goldberg, Katrin Ellice Heintze, and Shagun Bajpai. 2025. Examining the Use and Impact of an AI Code Assistant on Developer Productivity and Experience in the Enterprise. InProceedings of the Extended Abstracts of the CHI Conference on Human Factors in Computing Systems(New...

    arXiv 2025

  26. [27]

    Ziyang Ye, Triet Huynh, Minh Le, and M. A. Babar. 2025. LLMSecConfig: An LLM-Based Approach for Fixing Software Container Misconfigurations.2025 IEEE/ACM 22nd International Conference on Mining Software Repositories (MSR) (2025), 629–641. doi:10.1109/msr66628.2025.00099

    arXiv 2025

  27. [28]

    Ilya Zakharov, Ekaterina Koshchenko, and Agnia Sergeyuk. 2025. AI in Software Engineering: Perceived Roles and Their Impact on Adoption. InProceedings of the 33rd ACM International Conference on the Foundations of Software Engineering (New York, NY, USA, 2025-07-28)(FSE Companion ’25). Association for Computing Machinery, 1305–1309. doi:10.1145/3696630.37...

    arXiv 2025

  28. [2025]

    doi:10.1109/ACCESS.2025.3574462

    Impact of Artificial Intelligence on Software Engineering Phases and Activities (2013–2024): A Quantitative Analysis Using Zero- Truncated Poisson Model.IEEE Access13 (2025), 95535–95547. doi:10.1109/ACCESS.2025.3574462

    arXiv 2013