Systems Thinking and Complexity Theory for AI Architects

Explore essential systems thinking and complexity theory frameworks for designing responsible AI systems in this 49-minute conference talk by Nimisha Asthagiri from Thoughtworks. Learn to move beyond building Proof-of-Concepts without strategic consideration of unintended consequences as autonomous, self-learning multi-agent systems increase system complexity. Master frameworks like Cynefin and Iceberg models to identify and govern reinforcing loops that lead to algorithmic addiction, burnout, and ethical misalignment. Discover how to create Causal Flow Diagrams (CFD) to map AI risks including frontier models and fake alignment, while examining real-world examples from social media platforms to meeting scheduler agents. Understand the distinction between agents and microservices, explore multi-agent design patterns including RAG, Chain of Thought, and Reflection, and analyze agent topologies comparing orchestration versus decentralization tradeoffs. Apply the Cynefin Framework to distinguish between complicated and wicked complex problems, utilize the Iceberg Metaphor to examine events, patterns, structures, and mental models, and implement practical tools for behavioral observation and explainability using LIME and SHAP. Establish architectural boundaries with human-in-the-loop systems and governance agents while addressing the ethics of influencing behavior versus solving problems through comprehensive case studies and Q&A discussion.