Humans naturally combine pre-existing knowledge with observations and contextual clues to construct rich mental models of the world around them. These models help us evaluate goals, perform thought experiments, make predictions, and update our situational understanding. When interacting with others, we use a skill called theory of mind (ToM) to infer their mental states and predict future actions. In team settings, these models can become highly complex. High-performing teams naturally align key aspects of their mental models to create shared understandings of their environment, equipment, team dynamics, and strategies. ToM and the ability to create shared mental models are critical elements of human social intelligence, forming the foundation for collaboration in various settings, from sports to military missions.
Despite the advancements in artificial intelligence (AI), AI technologies still struggle to understand the most crucial component of their operating environments: humans. This gap in understanding hampers the development of safe, efficient, and productive human-machine teams. The Artificial Social Intelligence for Successful Teams (ASIST) program aims to bridge this gap by developing AI systems with the basic social skills required to understand human partners’ goals and situational knowledge, predict their needs, and offer context-aware actions, ultimately functioning as adaptable and resilient AI teammates.
The ASIST Program: Goals and Objectives
The ASIST program seeks to create AI agents capable of demonstrating machine ToM and participating effectively in teams by representing and maintaining shared models. This involves developing foundational AI theory and systems that can infer human partners’ goals, predict their needs, and engage in context-aware actions. The ultimate goal is to enhance human-machine collaboration across various domains.
Key objectives of the ASIST program include:
- Developing Machine Theory of Mind (ToM): AI agents will be designed to infer human mental states from observed actions and context, similar to how humans use ToM to understand and predict others’ behaviors.
- Creating Shared Mental Models: AI agents will help maintain shared mental models within teams, aligning their understanding of the environment, equipment, and strategies with their human partners.
- Building a Customizable Testbed: A testbed for evaluating AI agents will be created, featuring customizable open-world environments and standardized interfaces. This testbed will include standard sensing channels (information streams from physical and virtual sensors) and communication/action channels (mechanisms for agents to convey information to human teammates and engage with the team).
Challenges and Innovations
The ASIST program faces several challenges, including:
- Complex and Specialized Environments: AI agents must operate in increasingly complex and specialized environments, requiring advanced capabilities for situational understanding and decision-making.
- Adaptability to Perturbations: AI agents must be adaptable to sudden changes in the mission or team, such as the loss of communication with a key teammate.
- Noisy Multi-Channel Observations: AI agents must use noisy multi-channel observations to represent the world accurately and perform complex inference and prediction tasks.
Despite these challenges, the ASIST program is poised to make significant innovations in AI technology. By developing AI agents that can infer human goals, predict needs, and engage in context-aware actions, the program aims to create AI systems that function as effective, adaptable, and resilient teammates.
The Future of Human-Machine Collaboration
The ASIST program represents a crucial step forward in the development of AI systems that can understand and collaborate with humans effectively. By focusing on machine ToM and shared mental models, the program aims to create AI agents that enhance human-machine collaboration across various domains. From improving teamwork in military missions to enhancing coordination in industrial settings, the potential applications of ASIST technology are vast and varied.
As AI continues to evolve, the ability to understand and work alongside humans will become increasingly important. The ASIST program is at the forefront of this evolution, paving the way for a future where AI and humans can collaborate seamlessly and effectively. Through innovative research and development, ASIST is set to transform the landscape of human-machine teamwork, unlocking new possibilities for productivity, safety, and efficiency in diverse environments.