Navigating AI-Accelerated Science: An Intellectual Virtue Approach to Responsible Inquiry

Abstract Although AI can accelerate scientific discovery, it also generates new epistemic risks that prevailing rule-based governance frameworks are structurally ill-equipped to address. We present four arguments for the claim that intellectual virtues offer an adequate response to the challenges of AI-accelerated science. The Adaptability Argument states that rules cannot keep pace with technological change. The Motivation Gap Argument shows that compliance does not amount to epistemic motivation. The Human-in-the-Loop Argument stresses that AI systems require virtuous human interpretation. Finally, the Fallibility-in-Principle Argument maintains that AI systems remain fallible and, therefore, require human oversight. Moreover, we develop the first systematic mapping of intellectual virtues onto the CRediT (contributor roles taxonomy) stages of the research process and extend this to encompass the process of scientific evaluation. In doing so, we bridge virtue epistemology and the governance of AI-accelerated science. We show how virtues such as intellectual justice, intellectual scrutiny, intellectual humility, curiosity, and intellectual courage constitute the intellectual infrastructure that enables scientists to evaluate AI-generated outputs and preserve epistemic agency in an era of increasing outsourcing of cognitive effort. This virtue-based framework offers a proactive and future-oriented normative foundation for scientific practice in the age of AI.