A model of preference elicitation: The case of distributed resource allocation

Market mechanisms are deemed promising for distributed resource allocation settings by explicitly involving users into the allocation process. The market considers the users’ and providers’ valuations to generate efficient resource allocations and prices. In theory, valuations are assumed to be known to the user. In practice, however, this is not the case. It is a complex burden for both users and providers to assess their true valuation for a certain combination of resources and services and to efficiently communicate this valuation to the market. This paper contributes to the theory of designing distributed allocation models in that (i) we propose a model for preference elicitation, which allows users and providers to assess their valuations as a function of their resource requirements and strategic considerations, (ii) we show how this model can be encoded within so-called bidding agents which interact with the market on behalf of the user, and (iii) we evaluate our approach in a numerical experiment to illustrate how the bidding agent adapts to the dynamic market situation. As this evaluation shows, the model outperforms technical schedulers and can thus be used for decision support in electronic markets.