Prometheus Principles

Overview

The Prometheus Principles are design rules for creating Human-Centered AI interfaces that achieve high automation AND high human control. Named after the Greek god who gave fire to humanity, technology as a gift that empowers rather than replaces.

The Core Principles

Consistent interfaces for intent expression Users need stable, predictable ways to form, express, and revise what they want. If the interface keeps changing how you tell it what to do, you can’t build competence.

Continuous visual display of objects and actions Users need to see the things they’re working with and what’s happening to them. Hidden state is lost control.

Rapid, incremental, and reversible actions Small steps that can be undone. Not big commits that can’t be taken back. This enables exploration and recovery from mistakes.

Informative feedback for every action Each user action gets acknowledged. The system shows it received the intent. Silence is ambiguity.

Progress indicators for status When operations take time, users need to see where things stand. Not just “working…” but “step 3 of 5” or “47% complete.”

Completion reports to confirm accomplishment When it’s done, say so clearly. Don’t leave users wondering whether their goal was achieved.

How These Enable High-High

These principles don’t reduce automation, they make automation controllable. A thermostat automates temperature regulation, but the principles ensure users understand current state, can express new intent, get feedback that the system received it, and see progress toward the goal. High automation, high control.

Examples in Practice

Elevators: Button lights confirm intent received, floor display shows progress, tone signals arrival. High automation (coordinated multi-elevator algorithms), high control (users always know state and can express destination).

Digital cameras: Viewfinder shows what you’ll get, settings are visible and adjustable, automatic adjustments (exposure, focus) happen visibly, capture is instantaneous with confirmation.

Thermostats: Current temperature visible, setting visible, feedback when adjustment registered, progress visible as temperature changes.

When Principles Conflict

These principles sometimes tension with each other. Informative feedback can clutter the interface. Progress indicators for sub-second operations are noise. The principles guide thinking, not mechanical application.

Limitations

The principles focus on individual user interfaces. They don’t address team coordination, organizational oversight, or system-level trust mechanisms, those require RST’s broader framework.

Related: 07-molecule—hcai-two-dimensional-framework, 07-molecule—ui-as-ultimate-guardrail, 01-atom—imperceptible-ai-problem