Search

In this report, we present results reproductions for several core algorithms implemented in the OpenSpiel framework for learning in games. The primary contribution of this work is a validation of OpenSpiel's re-implemented search and Reinforcement Learning algorithms against the results reported in their respective originating works. Additionally, we provide complete documentation of hyperparameters and source code required to reproduce these experiments easily and exactly.

The partially observable card game Hanabi has recently been proposed as a new AI challenge problem due to its dependence on implicit communication conventions and apparent necessity of theory of mind reasoning for efficient play. In this work, we propose a mechanism for imbuing Reinforcement Learning agents with a theory of mind to discover efficient cooperative strategies in Hanabi. The primary contributions of this work are threefold: First, a formal definition of a computationally tractable mechanism for computing hand probabilities in Hanabi. Second, an extension to conventional Deep Reinforcement Learning that introduces reasoning over finitely nested theory of mind belief hierarchies. Finally, an intrinsic reward mechanism enabled by theory of mind that incentivizes agents to share strategically relevant private knowledge with their teammates. We demonstrate the utility of our algorithm against Rainbow, a state-of-the-art Reinforcement Learning agent.

This paper examines the potential role of civil society action in increasing state accountability for development in Sub-Saharan Africa. It further develops the analytical framework of the World Development Report 2004 on accountability relationships, to emphasize the underlying political economy drivers of accountability and implications for how civil society is constituted and functions. It argues on this basis that the most important domain for improving accountability is through the political relations between citizens, civil society, and state leadership. The evidence broadly suggests that when higher-level political leadership provides sufficient or appropriate powers for citizen participation in holding within-state agencies or frontline providers accountable, there is frequently positive impact on outcomes. However, the big question remaining for such types of interventions is how to improve the incentives of higher-level leadership to pursue appropriate policy design and implementation. The paper argues that there is substantial scope for greater efforts in this domain, including through the support of external aid agencies. Such efforts and support should, however, build on existing political and civil society structures (rather than transplanting "best practice” initiatives from elsewhere), and be structured for careful monitoring and assessment of impact.