The benefits of search costs in multiagent exploration

Humans and software agents alike spend considerable time and effort in searching. Search enables finding the things that better fit and agent’s goals. But search can also be a costly process. Search costs can either come in the form of direct monetary payments, or in the form of time and resources spent. In general, the searcher must balance between the benefits provided by longer and broader search, on the one hand, and the associated increased cost, on the other. In economic literature search costs are often referred to as “environment friction” or “market inefficiency” and associated with reduced market performance [1]. Indeed, in the presence of search costs a rational player will not aim to find the best option, but rather settle for the “good enough”, beyond which the marginal cost of searching exceeds the marginal benefit of continuing the search. Thus, search costs promote sub-optimal results (or so it would seem). As such, the traditional wisdom is that when designing a MAS environment, search costs should be avoided or reduced to a minimum. Taking eCommerce as an example, most researchers see a great benefit in the ability of eMarketplaces to lower the buyers’ cost to obtain information (e.g. about the price features) from multiple sellers, as well as the sellers’ reduced costs to communicate their information [1]. The lowered search cost is associated in this case with increased economic efficiency and enable new markets to emerge. Similarly, many systems have been introduced in which central mechanisms or mediators are used in order to supply the agents complete information concerning market opportunities, eliminating the need to engage in costly search. In this paper we show that, not withstanding the above, search costs – “friction”, if and when applied appropriately,