Presentation Information
[U03-05]Open Science in the Age of Generative AI
— The Hacking of the Academic Ecosystem —★Invited Papers
*Kazuhiro Hayashi1 (1.National Institute of Science and Technology Policy)
Keywords:
Open Science,Generative AI,Academic Ecosystem,Scholarly Communication,Research Integrity
Introduction
The rapid advancement of generative artificial intelligence (Generative AI) is transforming research practices and academic publishing. Tasks such as literature search, summarization, translation, and manuscript drafting are increasingly supported by AI systems, leading to deeper integration of AI into researchers’ cognitive processes. Generative AI is no longer merely a tool for efficiency; it has become an active element in knowledge creation.
This paper examines how generative AI, combined with open science, is reshaping the academic ecosystem. While open science has traditionally focused on open access and data sharing, generative AI introduces a new dimension in which openly available knowledge is continuously recombined and regenerated. This shift presents both significant opportunities and critical challenges for the reliability and governance of scholarly communication.
1. Transformation of Knowledge Creation
Large language models (LLMs) enable researchers to process vast volumes of academic information across disciplines. As a result, researchers can rapidly gain overviews of relevant literature and identify potential research questions. In this sense, generative AI functions as a cognitive amplifier rather than a simple automation tool.
However, AI-generated outputs are not inherently accurate or verifiable. The problem of hallucinations—plausible but incorrect information—poses a serious risk to research quality. As AI-generated text becomes increasingly fluent, even experts may find it difficult to detect errors based solely on textual evaluation.
2. Hacking the Academic Ecosystem
The widespread use of generative AI is beginning to “hack” existing academic systems. Here, hacking refers to the disruption of established mechanisms—such as peer review, publishing, and research evaluation—through uses not anticipated in their original design.
Recent cases have shown that papers containing AI-generated errors or fabricated references have passed multiple rounds of expert peer review. This suggests that traditional peer review, which relies heavily on human judgment of coherence and novelty, may be insufficient in an AI-assisted research environment.
Additionally, falsified author identities and fabricated datasets have been reported, exploiting systems that presuppose trust and authenticity. These developments challenge the fundamental credibility of scholarly communication.
3. Institutional Responses and Challenges
In response, publishers and research institutions are revising review criteria and policies, placing greater emphasis on data authenticity, methodological transparency, and reproducibility. Open data and metadata standardization play a crucial role in enabling verification and accountability in AI-assisted research.
Nevertheless, major challenges remain, including clarifying responsibility for AI-generated content, strengthening AI literacy among researchers and reviewers, and redesigning research evaluation systems to reflect new modes of knowledge creation.
Conclusion
The convergence of generative AI and open science accelerates knowledge production while exposing structural vulnerabilities in the academic ecosystem. These changes require more than technical solutions; they call for a fundamental redesign of scholarly communication and governance. In the generative AI era, open science must evolve to support not only openness but also trust and sustainability.
The rapid advancement of generative artificial intelligence (Generative AI) is transforming research practices and academic publishing. Tasks such as literature search, summarization, translation, and manuscript drafting are increasingly supported by AI systems, leading to deeper integration of AI into researchers’ cognitive processes. Generative AI is no longer merely a tool for efficiency; it has become an active element in knowledge creation.
This paper examines how generative AI, combined with open science, is reshaping the academic ecosystem. While open science has traditionally focused on open access and data sharing, generative AI introduces a new dimension in which openly available knowledge is continuously recombined and regenerated. This shift presents both significant opportunities and critical challenges for the reliability and governance of scholarly communication.
1. Transformation of Knowledge Creation
Large language models (LLMs) enable researchers to process vast volumes of academic information across disciplines. As a result, researchers can rapidly gain overviews of relevant literature and identify potential research questions. In this sense, generative AI functions as a cognitive amplifier rather than a simple automation tool.
However, AI-generated outputs are not inherently accurate or verifiable. The problem of hallucinations—plausible but incorrect information—poses a serious risk to research quality. As AI-generated text becomes increasingly fluent, even experts may find it difficult to detect errors based solely on textual evaluation.
2. Hacking the Academic Ecosystem
The widespread use of generative AI is beginning to “hack” existing academic systems. Here, hacking refers to the disruption of established mechanisms—such as peer review, publishing, and research evaluation—through uses not anticipated in their original design.
Recent cases have shown that papers containing AI-generated errors or fabricated references have passed multiple rounds of expert peer review. This suggests that traditional peer review, which relies heavily on human judgment of coherence and novelty, may be insufficient in an AI-assisted research environment.
Additionally, falsified author identities and fabricated datasets have been reported, exploiting systems that presuppose trust and authenticity. These developments challenge the fundamental credibility of scholarly communication.
3. Institutional Responses and Challenges
In response, publishers and research institutions are revising review criteria and policies, placing greater emphasis on data authenticity, methodological transparency, and reproducibility. Open data and metadata standardization play a crucial role in enabling verification and accountability in AI-assisted research.
Nevertheless, major challenges remain, including clarifying responsibility for AI-generated content, strengthening AI literacy among researchers and reviewers, and redesigning research evaluation systems to reflect new modes of knowledge creation.
Conclusion
The convergence of generative AI and open science accelerates knowledge production while exposing structural vulnerabilities in the academic ecosystem. These changes require more than technical solutions; they call for a fundamental redesign of scholarly communication and governance. In the generative AI era, open science must evolve to support not only openness but also trust and sustainability.