[Home] [Tech Library]

4. Initial strategies for trust

Many of the above algorithms make strategic sense only if one object can trust another object to follow them. For example, there are direct financial incentives to embezzle funds or misreport earnings by violating the dividend algorithm, and there are market-share incentives to produce falsely low cost estimates by violating the base-demand algorithm. Further, improper market intelligence (who is using what services?) can be gleaned by comparing alert- ID values arriving via different consultants. Thus, one needs what may be called initial strategies for trust.

The simplest strategy is for an object to trust whatever existing objects it is initially instructed to trust. This need not lead to great inflexibility or put a great burden on the programmer. Standard initial market strategies for resource management can be provided by a programming environment. In one kind of implementation, a wide range of objects will use instances of the same, small set of initial-strategy objects; these objects will recognize and trust each other, and will be able to interact with other objects in ways that do not assume trust. (Unforgeable identities are an essential foundation for trust.) Thus, use of standard initial strategies can itself be an initial strategy for trust.

Other means of building trust are discussed in [II]. They include creating or noticing situations having the characteristics of indefinitely-iterated prisoner's dilemma games [10] (see also the discussion in [I]), use of posted bonds, use of positive-reputation systems, and use of behavior-certification agencies.

Previous Next

[Home] [Tech Library]