Interactive scenarios arise broadly, from robots collaborating with humans in dynamic environments, to recommender systems interacting with consumers, to intelligent control systems for power grids, transportation, and smart cities, to discovery agents interacting with literature databases and the internet. As foundation models grow more capable, they are being integrated into these interactive interfaces and autonomous systems that have physical or cyber components in the loop. This integration enables AI agents to collaborate with humans in assisting high-stakes decision-making across complex real-world domains. However, it also raises significant questions around how to ensure these models remain transparent, unbiased, robust, and aligned with human values.

This forward-looking workshop provides a forum to shape discussion on human-centric techniques that promote ethical and trustworthy integration of foundation models into interactive decision support systems. Sessions will review progress in areas like safe reinforcement learning and large language models (LLMs) based agents while also delineating open challenges around uncertainty quantification, safety evaluation, preference alignment, and responsible deployment in interactive settings.

By bringing together researchers across machine learning, human-computer interaction and AI safety, this workshop will facilitate cutting-edge thinking around trustworthy integration of foundation models into an emerging class of decision-making systems that closely link humans and AI. We look forward to your contributions on the techniques and responsible development practices that will shape the future of AI-assisted interactive decision making across a breadth of socially impactful domains.

• Workshop date: August 3, 2024

• Submission opens: April 1, 2024

• Submission deadline:  April 26, 2024

• Acceptance notification:  June 4, 2024

• Camera-ready and poster upload deadline: TBD