Modelling Deliberative Exchange: The Value of Combining Ideas

Deliberation plays a central role in democratic theory and practice, yet its epistemic benefits remain hard to pin down precisely. This paper advances our understanding by analyzing deliberation through the lens of NK-landscape models, which simulate how agents search for solutions in complex, multidimensional spaces. Drawing on recent work by Wu (2023), we examine how different learning strategies affect group performance in these environments. The analysis focuses on two distinct mechanisms of social learning: simple imitation, where agents copy the best solutions from their neighbors, and recombination, where agents synthesize their solutions with those of others. We conjecture that our computational experiments will demonstrate that recombination outperforms imitation across various configurations. If confirmed, this finding has important implications for deliberative theory, as recombination in NK-landscapes shares key features with deliberative exchange. Just as deliberating agents combine different perspectives and arguments to reach new understanding, recombining agents in our model merge partial solutions to discover superior alternatives. This parallel offers novel support for deliberation's epistemic value. The paper concludes by discussing the broader implications for institutional design and democratic practice.