Directed Improvisation with AI

Redefining human agency in an AI-mediated world—from producing answers to directing how humans and AI produce them together

About the Course

Directed Improvisation with AI, introduced by Yuen Yuen Ang in 2026 as a PhD seminar, addresses a central challenge posed by generative AI across organizations: What becomes of human agency when AI can generate answers on demand—and what does it then mean to think and learn?

In higher education, this challenge appears as “students cheating with AI,” prompting bans or restrictions that seek to restore individual effort. This course takes a different approach. In a world where AI is ubiquitous, the task is not to eliminate AI, but to redefine human agency—from producing answers to directing how humans and AI produce them together.

Anchored in Ang’s model of Directed Improvisation, the course centers direction as a meta-level skill, treating AI not as a passive tool but as a complex, co-creative system whose outputs depend on how humans guide it. Moving beyond simple commands and prompts, it introduces a method of paradigm-level direction, where humans direct the underlying assumptions, values, positionality, and concepts of AI. We explore this method using using bespoke knowledge infrastructure designed around AIM (Adaptive, Inclusive & Moral Political Economy).

More broadly, the course advances a new conception of human agency in an AI-mediated world—metacognition and direction: the ability to think about thinking and shape the conditions of thinking.

What is Directed Improvisation?

Directed Improvisation (DI) is a model for enabling collective creativity that combines top-down direction with bottom-up improvisation, in contrast to rigid control or chaotic decentralization.

While DI has appeared in other creative domains such as computer programming and the performing arts, Ang is the first to apply it systematically to development and governance in How China Escaped the Poverty Trap (2016), with later extensions to U.S. innovation policy (2023–25).

Since then, scholars and practitioners have applied DI across a range of settings, for example:

  • Economic revitalization in America

  • Transformative governance in Africa

  • Local governance in India

  • Delegated growth strategies in the UK

In this course, introduced in 2026, Ang further extends DI to human–AI co-creation.

Generative AI is a transformative but also disruptive technology: it can produce fast, fluent answers across a range of tasks, seemingly replacing human cognitive effort. Thus, many have interpreted AI as a threat to learning and skilled employment.

In higher education, it appears as “students cheating with AI,” prompting policies that ban, restrict, or penalize its use (for example, see New York Times). Similar concerns arise in journalism, consulting, governance, and other domains: if AI can write essays, solve math problems, and pass bar exams, what remains of human judgment and creativity?

This course shifts the terms of the debate. In a world where AI will be ubiquitous, the task is not to eliminate AI, but to redefine human agency—from producing answers to directing how humans and AI produce them together. Rather than treating AI as a passive tool that executes commands, it treats AI as a co-creative system whose outputs depend on how humans guide it.

The meta-skill explored in the course is direction: the ability to shape how problem-solving or creativity unfolds by structuring its conditions. Contrary to claims that AI will make humans “lazy,” direction is a higher-order form of cognition—akin to directing an improvised play or cultivating an ecosystem—where outcomes can be influenced but not precisely predicted or controlled.

The course connects these practices to broader questions of pedagogy and governance. We explore why and how institutions often respond to AI through control and regulation. By the end of the course, students will gain awareness of a new form of human agency in an AI-mediated world—metacognition and direction: the ability to think about thinking and shape the conditions for it.

Learning Objectives

The theoretical foundation of the course is Directed Improvisation (DI). Extended to human–AI cocreation, DI reframes the researcher as director and the AI as improviser.

Building on this foundation, the course introduces a methodological approach to guiding AI through paradigm-level interventions. Central to this task is Ang’s AIM, a system of thought grounded in three redefined assumptions:

  • Adaptive: systems thinking instead of mechanical thinking

  • Inclusive: pluralistic solutions instead of a single universal ideal

  • Moral: awareness of how power shapes ideas instead of feigned neutrality

Through a series of experiments, students explore how AI responds under different assumptions and paradigms, rather than only adjusting prompts and adding context.

Theory & Methods