The Opacity Tax

The less visible a system's reasoning and state, the more cognitive load the human bears. Opacity doesn't eliminate complexity — it shifts it to the user.

System Opacity doesn't make AI simpler — it makes the human's job harder. When the system doesn't show its reasoning, the user must infer it. When the system doesn't expose its state, the user must track it mentally. When the system doesn't provide audit trails, the user must verify everything manually. Every layer of opacity that the system hides is a layer of cognitive work that transfers to the human. The Opacity Tax means that opaque systems are effectively more expensive than transparent ones, because the human labor cost of compensating for opacity is real even if it doesn't appear on the AI invoice.

Why it matters

Teams evaluating AI systems often compare model costs while ignoring the Opacity Tax on their humans. A cheaper, more opaque system may cost less in tokens but more in human time. A more expensive, transparent system may cost more in tokens but save significant human effort. Total cost includes both.

In practice

Track not just token costs but human time spent on: verifying agent output, debugging unexpected behavior, re-doing agent work, and managing context. If human time costs exceed what transparency features would cost, you're paying the Opacity Tax.

← Previous The Amplification Principle Next → The Compounding Principle