Conclusions and open questions

In this paper, we investigated the economic conditions governing the spontaneous development of specialized market niches in a population of competing information brokers. We found that the emergence of niches is sensitively dependent on the extrinsic costs of doing business. Further, the stability of the specialization process may be understood in terms of the preferred number of categories offered by a single broker without competition. We found several distinct regimes. When the extrinsic costs are such that a single broker prefers to specialize, the multi-broker system does as well. When the single-broker system would prefer to offer multiple categories, brokers in the multi-broker system find themselves in an ever-changing competitive web of partially overlapping categories. Even so, competition functions like an effective cost, causing brokers to offer fewer categories than in the single-broker case; if the extrinsic costs are not too low, the system may still specialize fully. If the extrinsic costs are sufficiently low, however, the perceived advantages of offering multiple categories overcome the disadvantages of competition, pushing the brokers into the ``spam'' regime of complex and disastrous price-and-category wars, with no stable niches emerging.

The experimental results reported here raise a number of interesting questions. For example, what are the factors governing the timescale of the transient sorting-out period? How can we best quantify the cost of competition, and how does it depend on the system parameters? And what are the conditions leading to the seemingly abrupt collapse from price wars to full specialization?

Another avenue of further enquiry would test the realism of the model's economic features. For example, no provision is made for the cost incurred by a broker when it changes its price category offerings. In any real situation, there is likely be a nonzero cost of adding or dropping a category--e.g., capital outlays for additional bandwidth, storage, or processing power. Probably there will be a cost associated with deciding which categories to add or drop and what new price to offer--i.e., doing market research. The observed rapid fluctuations in category offerings crucially depended on brokers being able to freely switch categories in pursuit of maximum profit. Some early investigations into the effects of switching costs show that they can sometimes quench the fluctuations, depending on the relationship between the magnitude of the switching cost and the various choices of interest vectors. Note, however, that an excessive switching cost can also prevent brokers from moving into a potentially profitable niche.

As noted at the end of section 2, many improvements are possible to the informational aspects of the model as well. We expect these will lead to novel and interesting behavior as the agents are made more sophisticated and the details of article selection by brokers and consumers change. The nature, and number, of potential niches is likely to change radically, and may itself become an emergent property. Even so, we expect that the overall dependence of specialization on extrinsic cost and cost of competition will be qualitatively similar to what we have found here.