Improving Human Autonomy Teaming Efficacy Through a Voice Communication Interface

  • Adam Bogg

    Student thesis: Doctoral ThesisDoctor of Philosophy

    Abstract

    Continuous advancements in information technology have significantly changed the scale of use and scope of application of automation, to the point where future applied technologies, such as Unmanned Aircraft Vehicle Traffic Management systems, will be critically dependent upon integral intelligent automation capability. Whilst the implementation of intelligent automation generally results in an increase in system and operator performance and capability, research has shown that the high levels of automation can lead to operators losing essential environmental or system Situation Awareness (SA) and becoming ‘out-of-the-loop’ and at risk of missing or making safety critical errors.

    Research into Human Autonomy Teaming (HAT) proposes that improved communication between the human and automation of a system can address this problem. HAT proposes changing the relationship between the human and automation away from a human using a machine to one of a human teaming with an autonomous synthetic agent. Typically, experimental platforms for research into HAT communication use graphical or text-based communication channels, with the result that knowledge is lacking in the literature about the effect of implementing speech or audio-voice communication on human team-member SA, performance and perception of teaming.

    This research project proposes that speech conversation delivered through the audio-voice medium would provide the optimum form of communication for a HAT. Audio-voice communication would offer many human factors advantages, including ease of use, focused exchange of SA information, cognitive timesharing, and anthropomorphic appeal.

    The aim of this research project is to investigate the effect the use of audio-voice intra-team communication has on the SA, performance, and subjective perception of teaming of the human team-member. The objective of the research is to provide consolidated advice and guidance on the use of audio-voice based communication that can be used for future systems interface design.

    The project initiated with a review of research literature on SA that exposed a potential issue in core SA theory and resulted in the determination to utilise decision-making results as a secondary measure of SA to support evidence gathered using the SAGAT methodology. This was followed by a qualitative review of key factors identified as likely to impact on the implementation and structure of audio-voice communication such as direction of communication, teaming structure, evidence of cognition or reasoning, and automation degradation. The outcome of the initial review was a proposed framework for designing and implementing HAT messages drawing upon the aviation best practice human-human teaming model of Crew Resource Management (CRM).

    To conduct the investigation into the effect of audio-voice communication, a series of three experimental studies were conducted during which the key factors identified were systematically varied using an aviation-based traffic control simulator. The three experiments consisted of:

    1. A pilot study to evaluate the effect of synthetic agent communication as either audio or text messages against a range of teaming structures based upon four Levels Of Automation (LOA );
    2. A study into the specific effect of the human speaking to the synthetic agent;
    3. A final study examining the impact of increasing the synthetic agent message content to include evidence of cognition and reasoning.

    The findings of all three studies were essentially positive, provided measured results and evidence that synthetic agent voice communication can have a significant positive effect on human SA, performance, decision-making and perception of teaming, improving human trust of the synthetic agent and influencing human decision-making. Importantly, it was found that whilst providing the autonomy with a human voice can encourage humans to adopt synthetic agent recommendations, the additional provision of reasoning information embedded within the synthetic voice messages ensured human trust was optimised, preventing over-trust and unsafe behaviour.

    However, the studies also provided unexpected negative findings on the use of SAGAT, with the method consistently providing apparently unusually low absolute measures and values of SA that largely contradict the positive performance and decision-making findings. Practical and theoretical analysis of these findings raise questions about the validity of the SAGAT methodology and, by extension, the SA theory that it is based upon. The same analysis supports using measures of decision-making as a more accurate and extensive approach to evaluating SA. It is recommended that future research continues to examine this approach, especially in the context of complex systems of multiple human and synthetic actors.

    The conclusion of the research project is that cumulatively, the results of the studies indicate that there are significant positive advantages to providing a HAT with an audio-voice communication interface that lever off the best practice methodologies of human-human team systems such as CRM, and that audio-voice communication can improve overall systems safety by assisting the human team-member build SA and make more balanced and considered decisions.
    Date of AwardJan 2022
    Original languageEnglish
    Awarding Institution
    • Coventry University
    SupervisorMike Bromfield (Supervisor), Thomas Statheros (Supervisor), Stewart Birrell (Supervisor) & Andrew Parkes (Supervisor)

    Cite this

    '