"An Ideal Human"

Abstract

Driven by state-of-the-art AI technologies, human-AI collaboration has become an important area in computer-supported teamwork research. While human-AI collaboration has been investigated in various domains, more research is needed to explore human perceptions and expectations of AI teammates in human-AI teaming. To achieve an in-depth understanding of how people perceive AI teammates and what they expect from AI teammates in human-AI teaming, we conducted a survey with 213 participants and a follow-up interview with 20 participants. Considering the context-dependency of teamwork, we chose to study human-AI teaming in the context of multiplayer online games as a case study. This study shows that people have mixed feelings toward AI teammates but hold a positive attitude toward future collaboration with AI teammates in general. Our findings highlight people's expectations for AI teammates in a rapidly changing collaborative environment (e.g., instrumental skills for in-game tasks, shared understanding between humans and AI, communication capabilities, human-like behaviors and performance), as well as factors that impact people's willingness to team up with AI teammates (e.g., pre-existing attitudes toward AI, previous collaboration experience with humans). We contribute to CSCW by shedding light on how AI should be structured in human-AI teaming to support highly complex collaborative activities in CSCW environments.

Artificial Intelligence and Groups: Effects of Attitudes and Discretion on Collaboration

“…The increasing use of AI makes understanding the interactions between humans and AI, and the implications for team processes, of critical importance (R. Zhang et al, 2020). We started with a search for “Artificial Intelligence” articles on Web of Science to identify the most highly cited papers across disciplines and then in business journals only.…”

“…Yet, the human ability to access attitudinal assessments and then the stability of those assessments is critical to our understanding of how attitudes connect to behaviors (Glasman & Albarracín, 2006; Kroenung & Eckhardt, 2015). Specifically related to AI, attitudes may play a role in trust evaluations (Glikson & Woolley, 2020; Tussyadiah et al, 2020) and willingness to team with AI (drawing from multiplayer online games, R. Zhang et al, 2020).…”

“…Negative attitudes toward AI technology can be rooted in a lack of trust, where a team associates AI with something they cannot learn or understand (Dawes, 1979). Or, people may fear errors and mistakes that may harm individuals in the team or collectively (R. Zhang et al, 2020). Astronaut John Glenn’s request to “Get the girl to check the numbers” is a historical illustration of how distrust can play out in terms of engagement.…”

Modelling intent for Manned-Unmanned Teaming : Exploring human-centric approaches for future combat aircraft systems

“…This question was posited by Baltrush and colleagues (2022), arguing for a shift from imposed collaboration to a more human-centric collaboration. Indeed, taking an even broader perspective, synthetic agents in a teaming context should be designed from the perspective of being a teammate rather than a replacement, requiring both human and artificial intelligence considerations (Zhang, McNeese, Freeman, & Musick, 2021). To this end, a more holistic perspective for modelling intent is necessary (Hiatt, Narber, Bekele, Khemlani, & Trafton, 2017;Van-Horenbeke & Peer, 2021), in which human factors approaches can provide novel insights for enabling intent-aware synthetic wingmen.…”

“…It has also been noted that these communities have branched into various specialized areas, focusing on different levels of analysis and challenges with few interactions and possible missed opportunities. This has, in turn, motivated proponents to argue for a more unifying perspective (Freedman & Zilberstein, 2019) and more holistic approaches (Hiatt et al, 2017;Van-Horenbeke & Peer, 2021) that accounts for both the human and artificial intelligence elements in teams (Zhang et al, 2021). However, despite these arguments for more holistic approaches, the tendency to rely on specific artificial intelligence techniques prevails, thus possibly missing the opportunities that can be gained from human factors methods.…”