An Empirical Study on the Trustworthiness of Life-Like Interface Agents Susanne van Mulken, Elisabeth André, and Jochen Müller DFKI GmbH, Stuhlsatzenhausweg 3, D-66123, Saarbrücken, Germany 1 Introduction In this paper, we describe an experiment investigating the impact of the personification of interface agents as life-like characters on the agents‘ trustworthiness. Current research in intelligent interfaces moves strongly toward interfaces in which the communication between human and computer is mediated by intelligent interface agents---programs that may be authorized to autonomously act upon the user's behalf. A prerequisite for the acceptance of this new form of HCI is that the user trust the agent and indeed delegate tasks to it. Besides on factors such as a cost-benefit tradeoff and the possibility for the user to effectively communicate the task to the agent, delegation depends on the trustworthiness of the agent. Trustworthiness in turn may depend on the agent‘s underlying behaviour (e.g., the agent’s competence and its perceived predictability) and on the way the agent presents itself to the user (e.g., Milewsky & Lewis 1997). Related research suggests that it may be advantageous to make interface agents appear more human-like. For instance, there is evidence that when users interact with computer-based systems, they are biased at first toward distrust, whereas in interaction with other humans, there is a tendency toward trust (see Muir 1987). In addition, some empirical studies (e.g., Sproull et al. 1994) provide evidence that visually personifying the interface (e.g., through a computer-animated face) can lead to general social facilitation. Finally, facial expressions, body gestures, and vocal intonations are interpreted in terms of trustworthiness and can also be used to express trustworthiness (see DePaulo 1992). A more specific line of research in intelligent interfaces concentrates precisely on personifying the look of interface agents: Agents are personified by means of life-like characters which inhabit the user interface. The questions investigated in the current study are: How big is the impact of a life-like character on the trustworthiness of the system compared to that of other