Our desire to explore the economic impact of shopbots in obtaining price and product information has led us to a model that is similar in spirit to several that have previously been investigated by economists interested in understanding the phenomenon of price dispersion. Our goals, however, are prescriptive, rather than descriptive, leading us to consider somewhat different causes and effects than are typical of price dispersion studies. Ultimately, we are interested in designing economically-motivated software agents, as well as an infrastructure that will support their interactions; thus, we have emphasized the constructive computation of price distributions and averages, rather than merely providing classical proofs of existence and other properties of equilibria.
Arguing that nonlinear search cost schedules are likely to exist naturally, or might even be adopted intentionally by shopbots, we studied their effect within the context of our model; our findings reveal that nonlinear search costs can lead to more complicated mixtures of buyer strategies and more extensive search than occur with linear costs. Another practical assumption, namely the existence of a positive number of uninformed buyers who do not use search mechanisms, can lead to similar outcomes. Taking evolutionary dynamics of buyer strategies into account, we found that the final equilibrium strategy vector depends on its initial value, and the route toward equilibrium can be surprisingly complicated.
Placing ourselves in the role of shopbot designers, we explored the strategic pricing of price information. Through modeling, analysis, and simulation, we validated our earlier assumption that nonlinear functions are reflective of search costs in electronic marketplaces by showing that nonlinear cost schedules come about as the natural consequence of economic incentives on the part of shopbots. Even in the face of competition from more costly search mechanisms, shopbots can wield a good deal of control over markets; specifically, they can manipulate prices so as to extract a large fraction of a market surplus. We demonstrated how buyers can benefit from this self-interested price manipulation. Shopbots' power would of course be diminished considerably if they were to compete amongst themselves. A study of coupled markets in which shopbots compete as providers of price and product information services to buyers or buyer agents would be fascinating.
In closing, we briefly mention two promising areas for future work. Firstly, combining the evolutionary dynamics of buyers with more realistic models of seller pricing behavior such as those described in [8, 12] seems likely to generate interesting, complex dynamics, and moreover, would be of practical significance. Secondly, since shopbots are starting to provide additional information about product attributes, it would also be of interest to analyze and simulate a model that accounts for both horizontal [1] and vertical differentiation.