Interface design is often miscategorized as a subset of visual polish, usability, or product optimization.
This mistake is not semantic—it is structural, and it explains why the same failures recur across domains, decades, and technologies.

App UI/UX asks: Is this easy, pleasant, or engaging to use?
Interface design, properly understood, asks: What actions does this system permit, constrain, obscure, or normalize—and with what consequences?

The Category Error

UI/UX, as practiced in most commercial contexts, optimizes for:

  • engagement
  • speed
  • reduction of friction
  • conversion
  • retention

These are market metrics, not governance metrics.

Interface design, by contrast, governs:

  • authority
  • responsibility
  • reversibility
  • escalation
  • refusal
  • accountability

When these two are conflated, interfaces become efficient at the wrong things.

Why This Matters Historically

The most consequential interface failures in history were not “bad UX”:

  • Three Mile Island did not fail because controls were ugly.
  • Therac-25 did not fail because screens were confusing in isolation.
  • The Florida ballot did not fail because voters lacked intelligence.
  • The Hawaii missile alert did not fail because users clicked too fast.

They failed because interfaces encoded power relationships incorrectly.

The interface decided:

  • who could act,
  • how quickly,
  • with what confirmation,
  • and with what opportunity to stop.

Those are governance decisions.

The AI Parallel

Modern AI interfaces repeat this error at scale.

Chat-based systems collapse:

  • advisor vs authority
  • exploration vs execution
  • draft vs decision
  • possibility vs recommendation

into a single conversational surface.

This feels friendly.
It is also structurally dangerous.

Not because AI is malicious—but because the interface never signals when the system should not be trusted.

Why UX Framing Fails for AI

UX framing assumes:

  • reversible actions
  • low-stakes outcomes
  • user experimentation as learning
  • failure as acceptable friction

AI systems violate these assumptions.

AI outputs are:

  • copied
  • propagated
  • institutionalized
  • automated
  • cited

Long after the interaction ends.

A friendly interface that optimizes confidence without signaling limits becomes a force multiplier for error.

ACP’s Design Claim

ACP insists on a hard distinction:

  • UI/UX optimizes interaction
  • Interface design governs action

Good UX can exist inside a bad interface.
Bad UX can exist inside a good interface.

They are orthogonal.

ACP therefore evaluates interfaces not by delight, but by:

  • clarity of authority
  • visibility of uncertainty
  • disclosure of compression
  • presence of refusal
  • traceability of action
  • auditability after failure

Why This Is Uncomfortable

Treating interface design as governance means:

  • slowing things down
  • adding friction
  • exposing uncertainty
  • making refusal visible
  • limiting autonomy—sometimes intentionally

These are anti-patterns in consumer tech.
They are necessities in institutional systems.

Closing the Arc

The design arc is not about better screens.
It is about making power visible before it causes harm.

If governance is invisible, it will still operate—just without accountability.

Interface design determines whether institutions can see what their systems are doing before the accident report is written.

That is why interface design is governance.