Background and aim: Rapid advancements in artificial intelligence (AI) have triggered a leap in science, technology, and innovation systems, facilitating the transition from e-government to smart governance. This study aimed to systematically identify and analyze the transformative functions of AI within the cycle of science and technology assessment, evaluation, and policymaking.
Methodology: This study was applied in nature and employed a systematic review methodology incorporating reflexive thematic analysis. Guided by the PRISMA 2020 statement, the study conducted a systematic review and thematic analysis of 98 credible domestic and international sources (published between 2018 and May 2026). During the full-text screening phase, each source was scored out of 5 based on its title, abstract, introduction, methodology, and findings; sources scoring below 3 were excluded. Data were analyzed using reflexive thematic analysis, and theme citation frequencies were reported descriptively.
Result: The results identified 7 themes and 40 components. Of the 135 coded citation instances, 69 (51.1%) related to AI functions in analysis and mapping, research evaluation, decision support, policy simulation, monitoring, and science-society communication, while 66 (48.9%) pertained to ethics and governance. Of these 66 items, 22 addressed the ethical and epistemological risks associated with the application of artificial intelligence such as bias, model opacity, and the erosion of human responsibility while 44 focused on the governance of AI itself, covering national strategies, comparative policy analysis, regulation, public trust, and infrastructure. Only 39 of the 98 sources cited in the component tables dealt with the functions of AI in science and technology assessment and policymaking.
Conclusion: Although AI enhances policymakers' capabilities by acting as a "cognitive exoskeleton," it faces challenges such as epistemic injustice and algorithmic bias. Successful integration requires adherence to guiding ethical principles, the development of open data infrastructure, and the preservation of human judgment's pivotal role within a "human-in-the-loop" framework. This review contributes by distinguishing AI as a tool for assessment and policymaking from AI governance as a policy subject, and by presenting a dual-aspect conceptual model alongside institutional recommendations tailored to science and technology governance in Iran.