Large Language Models (LLMs) such as ChatGPT are transforming how scientists conduct and validate research, offering promise as tools to improve scientific reproducibility. However, computational reproducibility and error detection remain expensive and labor-intensive. We experimentally test how collaboration between researchers and LLM assistants influences the reproduction of quantitative social science findings across different levels of AI autonomy. We randomly assigned 288 researchers to 103 teams working under three conditions: human-only, AI-assisted (using ChatGPT as a collaborative tool), or AI-led (ChatGPT operating with minimal human oversight). Teams reproduced published results from leading social science journals, detected coding errors, and proposed robustness checks. Human-only and AI-assisted teams achieved comparable reproduction rates (94% vs. 91%) and performed similarly on most outcomes, except human-only teams identified significantly more major coding errors. Both substantially outperformed AI-led teams, which achieved only a 37% reproduction rate, detected fewer errors across all categories, proposed weaker robustness checks, and required more time. This autonomous approach, however, likely represents only a lower bound of AI capabilities. Despite rapid model advances, expert human judgment currently remains indispensable for reliable empirical verification. While AI assistance did not degrade most outcomes, it provided no measurable advantages and was associated with reduced detection of major errors. However, the 37% autonomous reproduction rate indicates that AI could provide value in settings where scale or cost constraints preclude human review of papers, even though general-purpose LLMs offer no immediate advantages for human-supervised verification.
This study evaluates the effectiveness of varying levels of human and artificial intelligence (AI) integration in reproducibility assessments. We computationally reproduced quantitative results from published articles in the social sciences with 288 researchers, randomly assigned to 103 teams across three groups human-only teams, AI-assisted teams and teams whose task was to minimally guide an AI to conduct reproducibility checks (the "AI-led" approach). Findings reveal that when working independently, human teams matched the reproducibility success rates of teams using AI assistance, while both groups substantially outperformed AI-led approaches (with human teams achieving 57 pp higher success rates than AI-led teams). Human teams found significantly more major errors compared to both AI-assisted teams and AI-led teams. AI-assisted teams demonstrated an advantage over more automated approaches, detecting 0.4 more major errors per team than AI-led teams, though still significantly fewer than human-only teams. Finally, both human and AI-assisted teams significantly outperformed AI-led approaches in both proposing and implementing comprehensive robustness checks.
13 August 2022 in Journal of Economic Behavior & Organization
This paper suggests news coverage of a proposed prisoner exchange program (the Acuerdo Humanitario / AH ) systematically affected FARC ( Revolutionary Armed Forces of Colombia ) operations in the early 2000s. We propose AH news raised the value of political kidnappees through an agenda-setting-type mechanism. Our identification strategy draws on disaster deaths in countries favored by Colombian emigrants as an instrumental variable for AH news. Findings suggest AH coverage increased the number of political kidnappings, while decreasing the number of financial kidnappings. Overall, a simple back-ofthe-envelope calculation implies AH news de-escalated the conflict in the short run.