The AI Attribution Paradox: Transparency as Social Strategy in Open-Source Software Development

Obada Kraishan · 2025

Read The AI Attribution Paradox: Transparency as Social Strategy in Open-Source Software Development by Obada Kraishan (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2512-00867.txt.

Introduction

Obada Kraishan wrote a 2025 paper on a delicate choice programmers now face: whether to credit AI help. Coding assistants have changed how open-source software gets written. Developers must decide whether to note AI assistance in their commits, while weighing how their community will judge that honesty.

Kraishan calls this the AI attribution paradox and studies it in GitHub data. The analysis covers thousands of repositories and commits across several major AI tools in recent years. It looks at how developers disclose AI use and how communities respond. The findings matter as open-source projects write rules for fair credit.

Catalog record

CREATORObada Kraishan
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE81 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2512-00867

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

The AI Attribution Paradox: Transparency as Social Strategy in Open-Source Software Development Obada Kraishan Abstract AI coding assistants have transformed software development, raising questions about transparency and attribution practices. We examine the "AI attribution paradox": how developers strategically balance acknowledging AI assistance with managing community scrutiny. Analyzing 14,300 GitHub commits across 7,393 repositories from 2023-2025, we investigated attribution strategies and community responses across eight major AI tools. Results reveal widespread AI usage (95.2% of commits) but strategic attribution: only 29.5% employ explicit disclosure, with dramatic tool variation (Claude 80.5% versus Copilot 9.0%). Explicit attribution triggers modest scrutiny (23% more questions and 21% more comments), but tool choice matters 20-30 times more for predicting reception. Community sentiment remains neutral regardless of attribution type, suggesting curiosity rather than hostility. Temporal analyses show rapid norm evolution: explicit attribution increased from near-zero in early 2024 to 40% by late 2025, indicating community adaptation. These findings illuminate attribution as strategic communication rather than simple transparency, advancing understanding of algorithmic accountability and norm formation during technological transitions. We discuss implications for developers navigating disclosure decisions, platforms designing attribution mechanisms, and researchers studying emergent practices in AI-augmented collaborative work. K eywords human-computer interaction \cdot AI coding assistants \cdot attribution practices \cdot open source \cdot impression management \cdot transparency \cdot human-AI collaboration 1 Introduction The past three years have introduced a new attribution challenge in software development [ 1 ] . With the increasing incorporation of AI coding assistants such as GitHub Copilot, ChatGPT, Claude, and Cursor into developer workflows and development environments, these tools are now generating large parts of production code [ 2 , 3 , 4 , 5 , 6 ] . From this shift, a key question has arisen: do developers explicitly acknowledge those portions of the code they did not write? This question matters most in open-source ecosystems that value transparency, collaboration, and trust in the community [ 7 , 8 ] . At the same time, there are no clear norms or guidelines for how to attribute AI-written code, leaving developers to choose between being fully transparent or keeping AI involvement more vague [ 9 ] . Despite this escalating challenge, most research on AI coding assistant tools has mainly focused on adoption rates, productivity gains, and code quality [ 3 , 10 , 5 ] . However, we still do not have a clear understanding of how developers indicate the role of AI in their code and how their communities react when they do disclose this role. Attribution decisions are not neutral; those decisions help shape trust and collaboration in open-source communities [ 7 , 11 ] . Thus, questions about acknowledgement and attribution are not simply about labeling the role of AI; they are about how authorship and credit are reconfigured in human–AI collaboration [ 12 , 13 ] . This study focuses on what we call the “AI attribution paradox”: developers must balance two conflicting pressures when deciding how to credit their use of AI. On the one hand, community expectations around transparency encourage AI-augmented developers to disclose when a tool has contributed to their code [ 14 ] . On the other hand, fear of negative peer reactions to AI-generated code pushes some developers toward intentionally vague attribution [ 9 , 15 , 5 ] . The attribution practices developers inherited were developed to address human collaboration, not human–AI collaboration, so they offer limited guidance in this new context [ 12 , 2 ] . As a result, developers are mostly guessing about which attribution choices will work well. We therefore investigate three research questions about attribution patterns, community response, and how these change over time. To address these questions, our study analyzes 14,300 commits across 7,393 GitHub repositories from January 2023 to December 2025, a period when AI coding tools were widely adopted. Our data pipeline draws on natural language processing of commit messages with engagement measures like emoji reactions, comment sentiment, code reviews, and participation. Empirically, the AI attribution paradox appears in our data as follows: the majority of developers acknowledge AI tools in some way, yet very few explicitly credit them with writing code. This is not a single, consistent pattern; it changes across tools, with some showing much higher rates of explicit attribution than others. Notably, commits with explicit attribution receive only slightly more scrutiny than those without, and we see no signs of a negative backlash, suggesting that communities are more thoughtful about AI disclosure than expected. Building on these results, this study makes three contributions. First, we provide a large-scale view of how developers handle AI attribution in open-source projects, showing that developers frequently mention AI tools but rarely state that the AI wrote part of the code, and that disclosure practices differ sharply across tools and ecosystems. Second, we find that explicit attribution is linked to slightly more scrutiny but no negative reactions, a pattern consistent with impression management and computer-mediated communication [ 16 , 17 , 18 ] , algorithmic transparency [ 19 ] , and collaborative work in distributed communities [ 20 ] . Finally, we translate these findings into practical implications for developers choosing how to attribute AI, for platform designers deciding how to surface disclosure, and for policymakers concerned with the governance of AI-augmented work. 2 Related Work We have structured this section into three domains of research. First, we examine prior work on AI coding assistants and patterns of their adoption among developers. Second, we look at research on attribution and transparency practices in collaborative software ecosystems. Third, we discuss communication and impression management in online communities. Together, these literatures provide rich insight into AI tools and collaboration, but much less about attribution practices and community responses. 2.1 AI Coding Assistants in Software Development The recent spread of coding assistants based on large language models has pushed researchers to explore their use, effectiveness, and consequences for software development. Early research largely examined technical functionality and productivity improvements. For example, Barke et al. [ 2 ] studied how developers interact with Copilot and found that they use AI suggestions selectively and actively verify them. Weisz et al. [ 21 ] assessed productivity gains when developers used AI assistance to translate code, showing that the benefits varied with task complexity and developer expertise. Imai [ 22 ] further examined AI assistance in terms of code quality and maintainability, noting that it brings clear benefits but also new challenges. Besides technical and productivity concerns, recent studies have examined the social and collaborative roles of AI coding assistants. For example, Dakhel et al. [ 3 ] looked at how developers calibrate trust in AI tools, focusing on the factors that increase or decrease the likelihood of accepting AI suggestions. Similarly, Sarkar et al. [ 23 ] examined AI coding assistants in pair programming, explaining how they can strengthen collaboration but also introduce risks of over-reliance. In the area of security, Nguyen and Nadi [ 24 ] demonstrated that AI-generated code can include vulnerabilities that developers do not always catch. Taken together, these studies show that AI tools are changing the technical workflow, but they mostly treat AI assistance as an individual productivity tool, missing its intersection with community dynamics in

Read the complete text