An expert-model and machine learning hybrid approach to predicting human-agent negotiation outcomes in varied data

Abstract

We present the results of a machine-learning approach to the analysis of several human-agent negotiation studies. By combining expert knowledge of negotiating behavior compiled over a series of empirical studies with neural networks, we show that a hybrid approach to parameter selection yields promise for designing more effective and socially intelligent agents. Specifically, we show that a deep feedforward neural network using a theory-driven three-parameter model can be effective in predicting negotiation outcomes. Furthermore, it outperforms other expert-designed models that use more parameters, as well as those using other techniques (such as linear regression models or boosted decision trees). In a follow-up study, we show that the most successful models change as the dataset size increases and the prediction targets change, and show that boosted decision trees may not be suitable for the negotiation domain. We anticipate these results will have impact for those seeking to combine extensive domain knowledge with more automated approaches in human-computer negotiation. Further, we show that this approach can be a stepping stone from purely exploratory research to targeted human-behavioral experimentation. Through our approach, areas of social artificial intelligence that have historically benefited from expert knowledge and traditional AI approaches can be combined with more recent proven-effective machine learning algorithms.