Privacy, Trust, Agents & Users: A Review of Human-Factors Issues Associated With Trustworthy Software Agents

Developing software agents that users will trust with sensitive information is very difficult. The nature of the twice-removed relationship between the users and their tasks is described, and the concerns and opportunities of this relationship are explored. A model of agent acceptance is then proposed based on earlier work on user attitudes towards e-commerce transactions. This model describes how feelings of trust and perceptions of risk combine in opposite directions to determine a user's final acceptance of an agent technology. Specific factors that contribute to trust and risk are reviewed using both descriptive and prescriptive approaches where relevant research is reviewed and specific system and interface design features are recommended.