Some journalists have succumbed to an understandable temptation to interpret our results as implying that chaos is likely to reign as e-commerce comes to rely increasingly on agent technology. Others have understood that, rather than trying to discourage the trend towards agent-based e-commerce, we believe in this vision of the future and are trying to ensure that it will work smoothly and effectively!  By studying price wars and other undesirable collective behaviors in our simulations, we hope to gain insights into how we can prevent these phenomena from occuring in the real world.

The results reported in some of our papers represent a first step in this direction. We have identified some major factors that can lead to price wars, including myopia and the greatly reduced friction that results from dramatic decreases in the costs of advertising, shopping, etc. in an information economy. Our research on preventive measures targets both of these effects.

We have begun to explore some machine learning techniques that allow agents to model one another's behavior in hopes that this will help them to anticipate possible retaliation. We have shown that two sellers can learn not to engage in a price war.  How this technique might scale to more complex situations with more sellers is an issue that we are studying very actively.  We have also started to experiment with various methods for introducing and/or exploiting frictional forces as a possible stabilizing force.

In short, we do not believe that the sky is in imminent danger of falling, and we are doing our part to ensure that it will stay up there where it belongs!