From Full Autonomy to Manual Control Specifying Negotiating Agents

The C-IPS approach provides a way to structure negotiating agents at an abstract level that is above strategic and protocol-related aspects. The benefits to modelling and implementation of negotiating agents have been shown previously. Based on the modularisation imposed by the C-IPS approach we extend previously specified fully autonomous agents by a support for dynamic degrees of delegation. Artificial agents whose architecture supports dynamic degrees of delegation enable their users to delegate only subsets of their decisions to artificial agents. The delegation can be changed at runtime. Furthermore, the different modules can be developed independently of each other as long as the specified IPS dynamics are implemented. This paper shows that due to its abstraction level the C-IPS approach can be used to specify autonomous agents as well as interface agents.