Concession Curve Analysis for Inspire Negotiations

In the course of a negotiation it is often the case that the participants exchange packages of offers, which have, at least in the mind of the negotiators, a certain utility for them. We want to test whether the behaviour of the negotiators is reflected in the topology of the concession curve that plots each offer’s utility value in the course of a negotiation. In order to do this, we use data collected with the Inspire electronic negotiation support system, which records utility preference values for all issues under discussion, for each negotiator. We abstract the concession curves using a set of features, such as number of minima and maxima, slope of curve at the beginning and end, and then we use machine learning techniques to test whether we can predict negotiation outcome based on these concessions curve descriptions. We find that there are certain features of this curve, such as the number of minima and maxima, frequency of offers exchanged, that predict with high precision and recall the outcome of negotiations conducted with Inspire.