AI evidence extraction

The rise of the research automaton: science as process or product in the era of generative AI?

Henrik Skaug Sætra · AI & Society · 2025

AI-generated evidence extraction, verified across multiple analytical personas. Not a substitute for the peer-reviewed original.

9/10

Relevance

4/4

Quality (LMQS)

Evidence

Citations

1.61

FWCI

Top 10%

Impact

This is an AI analysis. Read the peer-reviewed original at the publisher: https://doi.org/10.1007/s00146-025-02557-7

Methodology & findings

Study design

Narrative argumentative essay combining philosophical analysis, critical examination of technology and science literatures, and policy recommendations.

Main result

The paper argues that "the value of scientific inquiry lies not merely in its outputs, papers, patents, and products, but in the process itself." The author demonstrates that generative AI now enables the "Research Automaton" – "the automated researcher that allows for the production of science-like output without any specific human being meaningfully involved in a traditional research process" – and identifies both normative and pragmatic reasons to resist this automation, as "uncritical automation of these tasks is a risky endeavor for academics, academia, and the societies in which academia is embedded."

Reports effect sizes.

Research paradigm

Critical/normative philosophy of science and technology ethics

Author conclusions

The author concludes that "The path forward will be shaped by the conscious choices made by researchers, educators, institutions, policymakers, and publishers" and that "The key question is not simply can we automate science, but should we? We must ask ourselves: What kind of scientific future do we want to build? And what role can and should Gen AI have in ensuring that science provides individuals and society with meaningful lives and the knowledge we require to reach our goals?"

Risk of bias

Author positionality and selection of sources may reflect a particular stance skeptical of AI automation in research. The vignette scenario, while illustrative, is speculative and may not reflect actual adoption patterns or user experiences. Heavy reliance on historical analogies (printing press, air pump) that may not be directly comparable to contemporary AI applications.

Open questions raised

The paper identifies gaps in understanding: (1) whether machines can perform true science with meaningful human input, (2) how to preserve the formative aspects of research crucial for researcher development, (3) how to distinguish between using AI to augment discovery versus replace the human process, and (4) how to maintain research integrity and skill development in an age of automation.
The paper does not explicitly identify empirical research gaps, but implies the need for research on: (1) long-term impacts of research automation on skill development and innovation capacity in science; (2) empirical assessment of whether Gen AI-automated research maintains the values of Mertonian norms (universalism, disinterestedness, organized skepticism); (3) investigation of how automation affects the ethical and normative dimensions of scientific practice; (4) study of whether hybrid human-AI cognitive constellations can preserve the formative aspects of the research process.
The paper identifies the need for further work on: (1) developing frameworks for ethical oversight and governance of AI in research; (2) understanding how to preserve essential research skills (critical thinking, creativity) in an era of increasing automation; (3) reforming research assessment frameworks to value process integrity over output volume; (4) investigating the long-term effects of skill atrophy on scientific innovation capacity.

Data: No datasets used. Author states: "Data availability: There was no data used."Code: None mentioned.Extracted from: pdfAgreement 72%

Explore related topics

Responsible AI Use Human-AI Collaboration Position Paper