Impact of Shared Mental Models on Human-AI Interaction and Mission Effectiveness
Abstract
The objective of this research is to develop a shared mental model that is accessible and updatable by both humans and AI, and to demonstrate that joint human-AI systems which include a SMM perform better at dynamic decision-making tasks. Central to this research is the belief that a human must be supported in an intelligible way (i.e. the human must have some understanding of the AI-system) and the AI must have an understanding of its human teammates thought processes. This concept of a mutual understanding of the problems, goals, information cues, strategies, and roles of each teammate is referred to as SMM. SMM are considered crucial to high-functioning human teams (Fiore, Salas, & Cannon-Bowers 2001; Mathieu et al. 2000). We aim to create a SMM of the decision task between the AI-agent and the human decision maker. Psychology research shows that when people interact with any complex system, they create a mental model which facilitates their use of the system (Norman, 1988). Similarly, humans also create mental models of AI-agents, through their interactions with these agents (Kulesza et al.,2012). Borrowing from this idea, there has been a recent surge in development of proactive and social robotic agents for human-robot teaming which stemmed from the AI agents ability to create mental models of humans. Creating SMMs is a multifaceted issue, wherein the AI must be transparent enough for the human to understand why the AI is making a suggestion (or at the very least whether the suggestion is reliable), while at the same time developing an accurate model to identify the humans expectations, wants, and needs. We will use a combination of theories from robotics, computer science, and psychology, alongwith the cascading decision making (CDM) model from our previous research to develop a proactive AI-agent to advise a human decision-maker, acting as a teammate rather than a tool. This AI-agent will improve human decision making by utilizing not only the problem parameters but developing cognizance of both the heuristic and analytic strategies on which human decisionmakers rely during high pressure decision tasks. If successful, we will have pioneered methods to create shared mental models between humans and AI systems that will allow improved human-AI teaming in a variety of domains. Further, we will have demonstrated a mechanism for allowing AI-agents to model humans utilizing heuristics, enabling them to anticipate human behavior and to provide appropriate decision support. Once more, it is an important step toward allowing humans to be both "on the loop" while not being "out of the loop", by providing insight into AI processes.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Aug 31, 2020
- Source ID
- N000142012577
Entities
People
- Karen Feigh
Organizations
- Georgia Tech Research Corporation
- Office of Naval Research
- United States Navy