Human Centered AI for Research Ethics and Transparency

In recent years, empirical sciences have faced a crisis of confidence, sparked by high-profile replication projects that failed to reproduce key findings. This "replication crisis" has prompted widespread reflection and a growing movement advocating for reform. Central to this movement is the belief that reproducibility, transparency, and rigorous evaluation are essential to scientific progress. Our research aligns with these goals, emphasizing how emerging technologies can offer innovative tools to address challenges in scientific integrity and reliability. This project outlines a research agenda focused on the use of artificial intelligence (AI) and hybrid human-AI systems to assess the replicability of scientific findings and to better understand science as a socially embedded practice. We develop and test a new class of AI-driven prediction markets, where algorithmic agents trade contracts tied to the outcomes of replication studies. Recognizing the value of human judgment, we explore hybrid prediction markets that integrate human participants alongside AI agents. These systems are examined through simulations and pilot experiments in real-world settings. To guide the design of these tools, we conduct in-depth surveys and interviews with researchers, gathering insights into the practical needs and concerns surrounding AI and hybrid systems in scientific workflows. Particular attention is paid to transparency, explainability, and the social and cultural dimensions that shape scientific practices. Our work aims to inform the development of technologies that not only support more reliable science but also respect and reflect the values of the scientific community.