Artificial Intelligence for assessing confidence in published research

Over the past two decades or so, a crisis of confidence has impacted most if not all scientific communities engaged in empirical research. High profile attempts to reproduce published findings have reported startlingly low success and have catalyzed a movement to improve scientific practices, from research design to evaluation and publication. One hope is that we might leverage advanced technologies, e.g., machine learning and artificial intelligence (AI) to support these efforts.

This talk will detail our work to develop and use AI to assign “confidence scores” to research claims published in the social and behavioral sciences literatures.  Our AI is actually a market populated by bot traders who buy and sell assets corresponding to research outcomes. Our work is part of the Defense Advanced Research Projects Agency (DARPA)’s Systematizing Confidence in Open Research and Evidence (SCORE) program. The talk will provide an overview of this program and our role within it.

• Beverages will be offered.
• Visitors are asked to follow current masking guidelines while indoors.