Systemic Risks of Interacting AI

TL;DR

This paper explores systemic risks emerging from interacting AI agents, providing a taxonomy, visualization tools, and scenario-based analysis to understand and categorize these complex risks.

Contribution

It introduces a novel taxonomy and visualization language for systemic risks of interacting AI, along with scenario-based identification and analysis of emergent risk patterns.

Findings

Identified key emergent risk behaviors in AI interactions

Developed a graphical language 'Agentology' for visualization

Demonstrated risk scenarios in smart grid and social welfare domains

Abstract

In this study, we investigate system-level emergent risks of interacting AI agents. The core contribution of this work is an exploratory scenario-based identification of these risks as well as their categorization. We consider a multitude of systemic risk examples from existing literature and develop two scenarios demonstrating emergent risk patterns in domains of smart grid and social welfare. We provide a taxonomy of identified risks that categorizes them in different groups. In addition, we make two other important contributions: first, we identify what emergent behavior types produce systemic risks, and second, we develop a graphical language "Agentology" for visualization of interacting AI systems. Our study opens a new research direction for system-level risks of interacting AI, and is the first to closely investigate them.