Carl Westin

4-2 Trust in and credibility of decision aids 67 source biases (e.g., perfect automation schema and disproportional negative effects of automation errors), 48, 58, 72, 179 and facilitate a more social interaction. 118 The Computers Are Social Actors (CASA) paradigm (also known as the media equation effect ) has provided evidence that people interact socially with computers (think household computers of the mid 90s) similar to how they interact with other humans, when computers personify human cues such as a gender, communications style, or personality. 118, 124 Beyond simple computers, research on automation eti- quette have shown that people perform better and trust an automated decision aids more if that aid adheres to human social communication “rules.” 114, 120 Research show that human speech is trusted more than artificial speech. 180 Noteworthy is that people have been found to rate their own voice (not knowing it is their own) as more attractive than others’ voices. 181 Waytz et al. 176 showed that an autonomous car, controlling steering and speed, was trusted more when portraying anthropomorphic features (name, gender, and voice). Effects of anthropomorphism on trust were also investigated by de Visser et al. 48 In a pattern recognition task, decision support was provided by three aids of varying anthropomorphic characteristics (by appearance and behavior): a computer (low), an avatar (intermediate), and a human (high). Increasing anthropomorphic characteristics increased trust, lessened the impact of errors on trust, and benefited the repair of trust. However, the drive for anthropomorphism in automated agents has also been cautioned as there is a risk for skewed beliefs and expectations of the automation’s capabilities and qualities. 48, 182, 183 Moreover, it is desirable to avoid transfer of weaknesses and biases associated with human sources, that can have undesirable effects on trust and acceptance. 72 In many contexts there may be a desire to retain the stereotype view of machines as more rational, consistent and fair. 48 4-2-4 The current study Findings from our own ATC related research on strategic conformance have shown that a controller’s acceptance and response to resolution advisories improve if ad- visories are based on the controller’s own solutions. 101 Such conformal advice can be considered to represent a next step of anthropomorphic characteristics, tied not to humans in general, but to the individual’s problem-solving preferences. 155 In this previous study, however, controllers were instructed that all advisories were automation generated, which may have limited their ability to anthropomorphize advisories. With consideration to previous research reviewed in this chapter (see the table in Appendix D for an overview), the perception of an aid partly depends on the expertise of the person who trust, and the framing of the aid (e.g., its credi- bility) to be trusted.