Be Patient and Tolerate Imprecision: How Autonomous Agents can Coordinate Effectively
IJCAI '99 Proceedings of the Sixteenth International Joint Conference on Artificial Intelligence
AAMAS '03 Proceedings of the second international joint conference on Autonomous agents and multiagent systems
Effect of referrals on convergence to satisficing distributions
Proceedings of the fourth international joint conference on Autonomous agents and multiagent systems
| Hi-index | 0.00 |
We study the problem of agents attempting to find quality service providers in a distributed environment. While referrals from other agents can be used to locate high-quality providers, referrers may be malicious and provide incorrect referrals to reduce traffic to their preferred service providers. Whereas apparently it would seem that such deceptive referrals can disrupt system stability, we observe that homogeneous groups of deceptive referrals converge faster to stable agent distributions over service providers compared to homogeneous groups of truthful referrers. We conjecture that deceptive referrers can unwittingly reduce the entropy, a measure of volatility, of the system as the recipient of a bad referral may not be inclined to move even if it is not satisfied with its current service providers. Additionally, we observe that mixed groups of truthful and deceptive referrers converge faster to stable distributions compared to homogeneous group of truthful referrers. These results highlight the unexpected positive effect of deceptive agents in stabilizing a population.