Exploring the Strategy Space of Negotiating Agents by Tim Baarslag

Author:Tim Baarslag
Language: eng
Format: epub
Publisher: Springer International Publishing, Cham

Bayesian Scalable Model

0.0050

0.0058

0.0032

0.0053

Hypothesis 6.3

The more competitive an agent, the more it benefits from using an opponent model.

At each turn of a negotiation session, a set of possible agreements can be defined. This is the intersection of two sets: the set of bids that an agent considers for offering, and the set of all bids acceptable to the opponent. The more competitive the agent, the smaller the intersection between the two sets. When an agent concedes, the number of possible agreements increases at the cost of utility. An opponent model can help in finding possible agreements, preventing concession and therefore loss in utility. We therefore expected the gain for competitive agents to be higher, as the set of possible agreements each turn is smaller, and therefore an optimal bid is more easily missed by an agent not employing an opponent model. This is especially decisive in the last few seconds of the negotiation, when many agents concede rapidly to avoid non-agreement.

The hypothesis is confirmed by our experiments. In Table 6.5 there is a negative correlation between the concession speed and relative gain in performance. If we ignore the results of the three worst performing models, a small—albeit statistically significant—negative correlation of is found.

Download

Copyright Disclaimer:
This site does not store any files on its server. We only index and link to content provided by other sites. Please contact the content providers to delete copyright contents if any and email us, we'll remove relevant links or contents immediately.