Page 120 - Artificial Intelligence for the Internet of Everything

106 Artificial Intelligence for the Internet of Everything

6.2 METHOD
6.2.1 Participants
Six hundred and five US workers responded to an open call for participation
on Amazon’s Mechanical Turk (MTurk). The MTurk workers were all
employed at least part-time and were at least 18 years of age. No other
demographics were collected in this study. Participants were compensated
for their participation.

6.2.2 Study Description and Items
As part of a study focused on trust in automated technologies, participants
were asked to identify one “intelligent technology” that they use on a regular
basis. The following definition and description was provided to participants:
Intelligent technologies or autonomous systems are technologies that can decide
how and when to interact with you during tasks, communicate and/or dialogue
with you, and or technologies that can help you accomplish your goals. Examples
might include things like autonomous cars, service robots, industrial robots, robotic
assistants, navigation aids, Amazon Echo/Google Home, the Nest, Siri, etc.
Once a technology was identified, participants were asked to describe rea-
sons why they trust or distrust the technology. No explicit definitions of
trust were provided nor were any of the trust antecedents mentioned as pos-
sible reasons for trust/distrust. Participants were given an open text box to
respond. Next, participants were asked to characterize the relationship they
had with the technology as a teammate- or tool-like relationship. Then they
were asked to discuss why they characterized the relationship as a teammate
relationship or (if they earlier noted that the relationship was more tool-like)
what it would take for the relationship to be viewed as one of a teammate.
Thus in either characterization the present study sought to understand the
components of human-machine teaming—either as perceived currently
or as visualized for a future scenario involving this technology. These
open-ended responses were, in turn, coded according to the scheme
described later.

6.2.3 Coding Method

Four coders independently coded the open-ended items. Two raters coded
the entire set and two others coded a portion of the data. All raters were first
trained on the coding process and on the trust antecedents and human-
machine teaming dimensions. Next, all four raters coded the first 70