Page 126 - Artificial Intelligence for the Internet of Everything

112 Artificial Intelligence for the Internet of Everything

to trust perceptions (Hoff & Bashir, 2015; Lee & See, 2004). The combina-
tion of reliability and predictability allows users of automated systems to
develop appropriate expectations of the technology in task contexts. Sup-
port in completing one’s tasks (not merely the notion of high reliability)
was found to be important, and this is logical as technology that makes peo-
ple more effective at something (tasks or social functions) should be trusted
more. The idea of proactive support was also relevant. This is consistent with
the construct of benevolence. Despite the machine ontology of the systems
considered by the humans, they appeared to value when the machine pro-
actively monitors the environment and reacted on the behalf of the human
without prompts by the humans. This behavior may be most relevant for
systems that have both high competence and high decision authority
(Lyons, 2013).
Interestingly, the facets of transparency involving understanding of the
logic for decisions, state awareness, and intent did not play prominently
in the participants’ descriptions of trust rationale. There are a couple poten-
tial explanations for these findings. First, it is possible that transparency of
intent overlapped with proactive support, as this is similar to benevolence,
which is a form of transparency of intent. Second, it is possible that the tech-
nologies used as referents in the present study did not contain decision-
making affordances. Prior research has found that transparency of decision
logic has focused on complex automation, where an automated aid recom-
mended a course of action in a highly complex situation (Lyons, Koltai,
et al., 2016b). Further, recent research—which found benefits for added
state awareness and projection—utilized highly complex automation and
complex military scenarios (see Mercado et al., 2016). Thus it is plausible
that greater emphasis would be placed on transparency in cases of highly
complex systems that were being used in dynamic and multifaceted scenar-
ios. Future research should examine this speculation, and future systems
developed to exploit the IoT may enable these more complex scenarios.
There were also few mentions of the constructs of liking, familiarity, and
social interaction. Again, it is possible that the types of technology consid-
ered by the participants influenced these findings. All of the technologies
considered were systems that are currently in use (i.e., they were not notion,
prototype, or futuristic systems), which could have created a ceiling effect
for familiarity (i.e., it is unlikely that individuals referred to a technology that
they were unfamiliar with). For this same reason it is possible that there was a
ceiling effect for liking, as it is unlikely that participants used a technology
referent for a system that was highly disliked. Future research can mitigate