Subproject A3

The market for services: Incentives, algorithms, implementation

In subproject A3 we model and analyze the market for composed services. The term "market" describes the "organization of economic interaction" between the different participants. To understand participants' incentives and the evolution of the market, we need to take into account specific characteristics of "On-The-Fly Computing". These include

(a) the possible trade relations between OTF provider and service provider that result from the technical composability of services,

(b) competition among service providers and among OTF providers, and

(c) imperfectly observable service quality.

Our market model which describes the "organization of economic interaction" between the different participants

Allocation

The OTF provider's procurement problem takes customers' needs into account and requires simultaneous transactions with suppliers. Subproject A3 analyzes different forms of organizing this allocation problem. We discuss models from cooperative bargaining theory, in particular multilateral interdependent problems and bargaining problems with incomplete information (about service quality). The study of matching markets and appropriate matching mechanisms complements the analysis.

Price and Competition

The central entity that drives competition among the various participants is the sales price of single services as well as service compositions. Since composed services are typically inhomogeneous products, a market equilibrium not only specifies prices but also further quantities such as product quality and the degree of product differentiation. This applies to competition among OTF providers as well as to competition among suppliers.

Evaluation of Quality

To minimize frictions in the market, it is important to adequately deal with the presence of asymmetric information on service qualities. We investigate in how far a reputation system can serve to reduce such asymmetries. The analysis has to include clients' strategic behavior, bounded rational behavior as well as learning algorithms that shall predict service qualities.

Methodology

For the analysis we use and develop methods from the fields of non-cooperative, cooperative, and algorithmic game theory. Furthermore, the study of boundedly rational behavior rests on methods from evolutionary game theory, behavioral economics and the theory of learning.