Platforms for Assessing Relationships: Trust with Near Ecologically-Valid Risk, and Team Interaction

• Published in: Engineering Artificially Intelligent Systems Vol. 13000; pp. 209 - 229
• Main Authors: Marble, Julie L., Greenberg, Ariel M., Bonny, Justin W., Kain, Sean M., Scott, Brandon J., Hughes, Ian M., Luongo, Mary E.
• Format: Book Chapter
• Language: English
• Published: Switzerland Springer International Publishing AG 2021; Springer International Publishing
• Series: Lecture Notes in Computer Science
• Subjects: Autonomous systems; Human machine teaming; Human-machine interaction; Research platform; Trust; Virtual reality
• ISBN: 9783030893842; 3030893847
• ISSN: 0302-9743; 1611-3349
• DOI: 10.1007/978-3-030-89385-9_13

Assessment of human-machine trust is difficult because of confounds in context, system capability and reliability. Trust indicates willingness to be vulnerable to the variable and unpredicted actions of another actor. Making people vulnerable to risk from decisions made by an intelligent agent is difficult to justify for research ethics purposes. Making expensive, physical intelligent agents vulnerable to human decisions is an inhibiting factor to exploring the development of trust or teams with embodied systems. These confounds can be addressed through use of virtual reality and immersive gaming systems. This chapter describes the development of two platforms, PAR-TNER and PARTI, for the exploration of human collaboration with autonomous systems, and provides an overview of a limited initial pilot of PAR-TNER. In PAR-TNER and PAR-TI, the test participant teams with either humans or machines to escape from a room collaboratively. PAR-TNER leverages virtual reality to stimulate risk, while PAR-TI allows researchers to explore team dynamics. While the data from the pilot test of PAR-TNER are limited, they indicate the ability to leverage the research platforms to discern trust from perceived capability.