Episode 7 of Institutional Design for the AI Age presents the controlled laboratory experiment that turns theoretical definitions into physical, measurable proof. We begin with the paradox that has paralyzed institutional theory: Douglass North and W. Richard Scott defined institutions in terms of latent, unobservable constructs—rules, norms, shared mental models—while digital platforms now generate oceans of behavioral telemetry that make those definitions irrelevant. This episode bridges that gap by introducing a 200-participant button-press experiment that physically measures the causal chain from informational signal to stabilized behavioral cluster, without a single survey question. No subjective report, no interpretation—just timestamped button presses that finally give institutional design for the AI age an empirical foundation.

We translate the abstract causal chain—signal → stimulus → reaction → pattern—into a physical setup: a bell and a visual sign act as the informational signals, and a single physical button (or an array of 30) captures the behavioral reaction. Four experimental conditions systematically test signal encoding (normative text vs. pure color), monetary incentives, and environmental complexity via single-button versus 30-button arrays. Crucially, this physical lab bypasses the algorithmic contamination and interference bias that render platform A/B tests unreliable—there is no recommendation algorithm secretly modulating exposure, no engagement optimization shifting signal distribution. The result is a clean measurement of how variations in signal type and environment shift the entire behavioral spectrum across exploratory and compliant clusters, not just a binary compliance rate.

The findings deliver the definitive operational definition: an institution is not the prescriptive text on the normative sign—it is the stabilized statistical distribution of physical button presses. That distribution is the institution. The experiment proves that the behavioral cluster, not the rule, is the observable object, and that institutional design for the AI age must start from this measurable reality.

🔥 What you’ll gain from this episode:

• The measurement gap exposed. Why North and Scott's definitions rely on unobservable mental models that leave no trace in behavioral telemetry—and why that's fatal in the platform era.
• The button-press experiment design. A 200-participant physical setup that translates the signal-reaction chain into a bell, a visual sign, and a physical button, with four conditions isolating signal encoding, incentives, and complexity.
• How the lab beats A/B testing. By eliminating algorithmic interference, the experiment delivers a clean view of how signal variations shift the entire behavioral spectrum, not just a single metric.
• The empirical redefinition of an institution. An institution is the stabilized statistical distribution of button presses—a behavioral cluster that can be directly observed and measured. No more guessing.

Stop defining institutions with words. Measure them with a button press, and build from what you can actually see.

📚 Dive deeper into the research and the book:

• Full paper: Read on ResearchGate
• Book: Habit Machine: AI Product Management
• AI A2A HUB: itinai.com

Connect with the author:

• Telegram: t.me/vlruso
• Email: vladimiruso@gmail.com
• LinkedIn: linkedin.com/in/uxproduct