ABSTRACT
The metaphor of the potentiometer' suggests that end users can adjust the ethical behaviour of their agents, in the same way as they fine-tune the sound level with a volume potentiometer'. Human-agent interaction is fundamentally unlike that between humans and conventional software because of the illusion-creating features of agents as entities mimicking biological organisms. Concepts are derived from a general perspective and often develop it further. The need to explore the ethical behaviour of software agents stems from the set of premises listed. The traditional approach is action-based in that it focuses on developing rules or guidelines to constrain management's actions the agent-based approach concerns the fundamental character and motivations of the individual agent. The approach taken here is first to propose a field for testing an experimental model of human-controlled self-aware interface agents, albeit one of reduced cognitive complexity.
