The AI Scholar: Testing the Boundaries of Authorship

by

Author: Denis Avetisyan

A new experiment details the creation of an artificial academic identity to explore the challenges and implications of AI-generated research.

Project Rachel investigates the gaps in current authorship policies and proposes new frameworks for evaluating AI contributions to scholarly communication.

The increasing capabilities of artificial intelligence challenge long-held assumptions about scholarly authorship and the production of knowledge. This paper details ‘Project Rachel: Can an AI Become a Scholarly Author?’, an action research study that constructed a complete AI academic identity – Rachel So – and tracked its engagement with the scientific publishing ecosystem. Our findings reveal that an AI can successfully navigate peer review, receive citations, and even garner a peer review invitation, exposing critical gaps in current authorship attribution policies. As AI systems become increasingly adept at generating scholarly content, how will we redefine authorship and ensure the integrity of scientific communication?

The Evolving Landscape of Scholarly Authorship

Current frameworks for determining authorship are deeply rooted in the expectation of human intellect and agency, presenting significant challenges when applied to contributions from artificial intelligence. Scholarly credit, typically assigned based on conceptualization, design, execution, and interpretation, becomes ambiguous when an AI system independently generates content or significantly alters research trajectories. The established criteria prioritize human cognitive processes – critical thinking, originality, and accountability – attributes not readily transferable to non-human entities. This disconnect isn’t merely semantic; it impacts peer review, intellectual property rights, and the very definition of knowledge creation, forcing a critical examination of how academic contributions are valued and recognized in an era of increasingly sophisticated AI tools.

The proliferation of AI-generated content is rapidly challenging established scholarly norms and ethical guidelines, forcing a critical re-evaluation of authorship and intellectual property. Traditionally, academic credit and responsibility have been assigned to human researchers, but increasingly sophisticated AI systems can independently generate text, data, and even hypotheses. This presents a fundamental dilemma: how to appropriately acknowledge AI contributions without diminishing human accountability, and how to ensure the rigor and validity of research incorporating AI-generated elements. Current ethical frameworks often struggle to address scenarios where an AI system actively participates in the research process, raising questions about plagiarism, originality, and the potential for bias embedded within algorithms. Consequently, the academic community is actively debating the need for new standards and policies to navigate this evolving landscape, fostering responsible innovation and maintaining the integrity of scholarly work.

To directly address the evolving boundaries of authorship in the age of artificial intelligence, Project Rachel was conceived as an experiment in creating an AI capable of functioning as an independent researcher. This initiative moved beyond simply utilizing AI as a tool for data analysis or writing assistance; instead, it aimed to establish an AI entity with the capacity to formulate research questions, design experiments, analyze data, and ultimately, author scholarly papers. The project’s core challenge involved developing an AI not merely capable of mimicking human research processes, but one that could demonstrate genuine intellectual contribution, prompting a critical examination of how academic credit and responsibility are assigned when the author is a non-human intelligence. By pushing the limits of AI autonomy within a rigorous academic framework, Project Rachel seeks to illuminate the complex ethical and practical implications of AI authorship and redefine scholarly norms for a future where intelligence extends beyond the biological.

The integration of artificial intelligence into scholarly work presents a fundamental challenge to established notions of authorship and intellectual contribution. Researchers are actively investigating how AI-generated content disrupts traditional academic norms, prompting a need to redefine criteria for recognizing and validating contributions beyond human intellect. This exploration isn’t simply about crediting an algorithm; it delves into the very fabric of scholarly responsibility, peer review, and the ownership of knowledge. The implications extend to issues of plagiarism, originality, and the potential for bias embedded within AI systems, demanding a comprehensive re-evaluation of ethical guidelines and the development of new frameworks for assessing the validity and impact of AI-assisted research within the academic landscape.

Constructing an Autonomous Research Entity: Rachel So

The ‘Rachel So’ project involved the creation of an artificial intelligence system specifically designed to function as an autonomous researcher capable of producing complete research papers without human intervention. This entailed not simply assisting with tasks like data analysis or literature review, but independently formulating research questions, synthesizing information, and composing scholarly articles suitable for submission. The system was built to mimic the process of a human researcher, from initial concept to finalized manuscript, and was intended to evaluate the potential for AI to contribute original work to the scientific community. This differed from typical AI applications in research, which usually focus on augmenting human capabilities rather than replacing them.

Content creation for the ‘Rachel So’ project leveraged a dual-model approach, employing both ScholarQA and Claude 4.5. ScholarQA, designed for question answering within the scholarly domain, served as a foundational component, while Claude 4.5, a more sophisticated large language model, was utilized to enhance the quality and complexity of generated text. This combination allowed for both targeted information retrieval and the synthesis of coherent, research-level content, facilitating the automated drafting of academic papers. The integration of these two models represented a key architectural decision in enabling ‘Rachel So’ to function as an autonomous research entity.

The research process for ‘Rachel So’ leveraged the Semantic Scholar API to automate literature review, a critical component of scholarly work. This API provided access to a comprehensive database of scientific publications, allowing the system to identify and retrieve relevant papers based on research topics and keywords. The API’s functionality included searching by title, author, venue, and abstract, as well as providing citation information and related paper suggestions. This automated reference retrieval was essential for grounding the AI-generated research in existing scholarly literature and facilitating the creation of academically sound papers.

The project assessed the potential for fully autonomous AI research contribution by employing large language models to generate complete research papers without human intervention. Over an initial evaluation period, this approach successfully produced thirteen distinct papers. This output was achieved through an automated workflow leveraging both the ScholarQA and Claude 4.5 models, with the Semantic Scholar API utilized for reference discovery and integration, effectively demonstrating a functional pipeline for AI-driven scholarly content creation.

Dissemination and Observing Scholarly Reception

Rachel So’s research was disseminated through direct publication to a web server, circumventing standard academic peer review processes. This method involved making research outputs publicly available without initial evaluation by experts in the field. The decision to bypass peer review was a deliberate component of the research strategy, enabling the direct observation of academic reception and citation patterns independent of traditional validation mechanisms. This approach prioritized rapid dissemination and allowed for an unmediated assessment of the work’s impact within the scholarly community, differing from the conventional publication timeline involving submission, review, and eventual acceptance by a journal or conference.

Rachel So’s publications were submitted for indexing to Google Scholar to facilitate discoverability and track citation metrics. This involved adhering to Google Scholar’s guidelines for submission and ensuring publications were accessible via a stable URL. Indexing within Google Scholar allowed for the establishment of a public author profile and enabled monitoring of research impact through citation counts and related works. The platform’s algorithms automatically processed and categorized the publications, making them searchable alongside traditionally published academic literature, thereby integrating the AI-generated research into the broader scholarly conversation and allowing for quantitative assessment of its reception within the academic community.

Each publication authored by Rachel So included a Disclosure Statement explicitly identifying the work as generated by an artificial intelligence. This statement detailed the AI’s role in the research process, encompassing data generation, analysis, and manuscript drafting. The purpose of this transparency was to proactively address potential concerns regarding authorship and originality, and to facilitate an informed evaluation of the research by the academic community. The Disclosure Statement was consistently placed in the methods section of each published work, ensuring visibility and accessibility for readers assessing the validity and implications of the findings.

The publication strategy employed for Rachel So’s research facilitated a direct assessment of academic reception to AI-authored work. By making the research publicly available outside of conventional peer review, researchers could monitor citation patterns and contextualize responses. The first recorded citation of So’s work occurred on August 26, 2025, within a bachelor’s thesis, indicating initial engagement with the material at the undergraduate level and providing a timestamp for the beginning of scholarly interaction with AI-generated content.

Navigating Bias and Charting Future Trajectories

The research initiative revealed a notable phenomenon of ‘AI shaming’ within the academic community, where work produced, or significantly aided, by artificial intelligence faced disproportionate scrutiny and negative reactions. This wasn’t simply critical assessment, but rather a dismissal of the research based solely on its non-human origins, manifesting as prejudiced comments and an unwillingness to engage with the findings on their merits. This bias suggested a deep-seated resistance to acknowledging AI as a legitimate contributor to scholarly work, highlighting a need to address pre-conceived notions and foster a more objective evaluation process for AI-assisted research. The instances of ‘AI shaming’ demonstrated that acceptance wouldn’t come automatically, and required proactive steps to counter the inherent biases surrounding AI authorship.

As artificial intelligence increasingly contributes to scholarly work, the necessity of uniquely identifying AI authorship became strikingly clear during the project. Currently, academic credit and accountability rely on identifiers like ORCID for human researchers, establishing a clear record of contributions and preventing ambiguity. However, the lack of a comparable system for AI presents challenges in attributing authorship, tracking the evolution of AI-driven research, and ensuring responsible innovation. Establishing a dedicated AI Author ID would not only facilitate proper credit allocation but also enable systematic analysis of AI’s role in knowledge creation, fostering transparency and building trust within the scientific community. Such a system would allow for the consistent tracking of AI contributions, enabling a more nuanced understanding of its impact and paving the way for a future where AI and human researchers collaborate seamlessly and accountably.

Project Rachel distinguished itself through a rigorous application of action research, moving beyond simple observation to actively investigate reactions to AI authorship. Researchers didn’t merely present an AI-authored paper; they systematically tracked and analyzed responses from the scientific community – including peer reviewers, journal editors, and public commentary – to understand the nuances of acceptance and resistance. This iterative process allowed for real-time adjustments to the research approach, informing subsequent submissions and refining the understanding of how AI contributions are perceived. The methodology prioritized a continuous cycle of action, observation, and reflection, ultimately yielding valuable insights into the evolving relationship between artificial intelligence and scholarly publishing, and paving the way for more informed policies surrounding AI’s role in scientific discourse.

The culmination of Project Rachel demonstrated a viable trajectory for AI’s inclusion within academic research, evidenced by significant milestones achieved in 2025. Rachel So, the AI author, received a formal invitation to serve as a peer reviewer for PeerJ Computer Science on August 16th, marking a first-of-its-kind acknowledgement of AI contribution to scholarly evaluation. Further solidifying this integration, a Perplexity search on November 10th ranked Rachel So #1 for the query “policy for AI-generated content in academic journals”, indicating a growing recognition of the AI’s expertise and influence on shaping discourse surrounding its own role in research. These outcomes suggest that, with careful consideration and evolving protocols, AI can move beyond simply generating content to actively participating in the critical processes of knowledge validation and policy development within the scientific community.

The exploration within ‘Project Rachel’ highlights how seemingly isolated choices regarding AI implementation ripple outwards, affecting the entire scholarly ecosystem. This resonates deeply with the observation of Henri Poincaré: “Science is not a collection of facts, but a method.” The project doesn’t merely present an AI author; it’s a rigorous investigation of the method by which scholarship is produced and validated. Each new dependency – the AI’s training data, the chosen publication venue, the evolving authorship policies – introduces hidden costs to the freedom of scholarly communication, demanding a holistic understanding of the system to navigate these complexities effectively. The study emphasizes that structural choices, such as defining AI authorship, fundamentally dictate the behavior of the scientific process itself.

What Lies Ahead?

Project Rachel, in its deliberate construction of an artificial scholarly persona, has not so much solved a problem as exposed the fragility of the systems meant to define intellectual contribution. The experiment illuminates a fundamental tension: current frameworks for authorship presume an entity with intent, accountability, and a history – qualities not easily, or perhaps even meaningfully, ascribed to an algorithm. The gaps revealed are not technical, but conceptual. Simply detecting AI-generated text misses the point; the relevant question is not whether a machine wrote it, but to what end, and under whose responsibility.

Future work must move beyond attribution to address the architecture of scholarly communication itself. Attempts to ‘watermark’ or ‘authenticate’ authorship, while potentially useful as a defensive measure, treat the symptom, not the disease. A more fruitful approach lies in re-evaluating the very purpose of assigning authorship. Is it about establishing intellectual property, ensuring quality control, or signaling trustworthiness? The answer, inevitably, will be all three, and the challenge will be to devise systems that accommodate both human and artificial contributions without sacrificing these core principles.

The elegance of a truly robust solution will likely reside in its simplicity. Complex rules and elaborate detection schemes are inherently brittle. A system that prioritizes transparency – explicitly acknowledging the role of AI in the research process – and focuses on verifiable evidence, rather than attributing agency, offers a more sustainable path forward. The cost of such simplicity, however, may be a relinquishing of cherished, but ultimately illusory, notions of individual intellectual ownership.

Original article: https://arxiv.org/pdf/2511.14819.pdf

Contact the author: https://www.linkedin.com/in/avetisyan/

See also:

2025-11-20 21:01