The impact of behavioral diversity in multi-agent reinforcement learning

Many of the world's most pressing issues, such as climate change and global peace, require complex collective problem-solving skills. Recent studies indicate that diversity in individuals' behaviors is key to developing such skills and increasing collective performance. Yet behavioral diversity in collective artificial learning is understudied, with today's machine learning paradigms commonly favoring homogeneous agent strategies over heterogeneous ones, mainly due to computational consideration...