Clarity
People must be able to understand the system state, the source of a message, and the next meaningful action.
Support Standard
A public-interest standard for human-centered systems
CLEAR
Human-Centered AI
Support Standard
From observed failure to verified repair.
Explore the case studiesDiagnostic pathway
- 1Observed failure→
- 2Evidence→
- 3Accountability→
- 4Repair→
- 5Verified closure
CClarityLLegitimate ConsentEExternal PresenceAAttention IntegrityRRestorative Pacing
The standard
Human dignity is a systems requirement.
CLEAR evaluates whether a human-facing system preserves identity, intent, orientation, attention, and the practical ability to pause or leave.
It turns lived failure into an operational evaluation method: observe the behavior, preserve evidence, identify the governing responsibility, implement repair, and verify closure. A system does not pass CLEAR because it appears helpful. It passes when its behavior remains understandable, consensual, externally oriented, attention-respecting, and restorative under pressure.
Five governing principles
CLEAR is both the name and the test.
Legitimate Consent
Participation, continuation, and data use must reflect informed choice—not pressure, ambiguity, or forced dependency.
External Presence
Technology should preserve orientation to people, surroundings, and reality rather than pull attention into a closed system.
Attention Integrity
Interfaces must respect human focus and intent instead of manufacturing urgency, repetition, or engagement loops.
Restorative Pacing
When confusion or failure occurs, systems must slow down, preserve context, and support recovery before demanding more action.
Evaluation lens
A system fails CLEAR when it…
- requires repeated input because its own channels are fragmented;
- confuses, redirects, or overrides the person’s stated intent;
- blurs the distinction between automated and human communication;
- prioritizes continued engagement over dignity and well-being;
- limits the practical ability to pause, disengage, or exit;
- acknowledges a concern without producing accountable closure.
Evidence in practice
Case studies from recurring system failure.
Each case converts an observed pattern into a CLEAR evaluation and an implementable repair standard. The purpose is institutional learning—not exposure of private individuals.
01
One issue becomes many support threads
- Observed pattern
- A single concern is fragmented across channels, receives duplicate or conflicting responses, and loses a coherent owner.
- Human impact
- The person must repeatedly reconstruct the issue while the institution treats each fragment as a separate event.
- CLEAR test
- Clarity • Attention Integrity • Restorative Pacing
- Required repair
- Link the threads, preserve one issue identity, name one accountable owner, and maintain a visible lifecycle through confirmed closure.
02
Identity remains while context disappears
- Observed pattern
- After a file or session event, a system may still recognize the person’s name while failing to restore the active conversation or work state.
- Human impact
- Identity recognition without continuity creates uncertainty about preservation, ownership, and whether work has been lost.
- CLEAR test
- Clarity • External Presence • Restorative Pacing
- Required repair
- Restore the most recent context, maintain file-to-conversation linkage, and present an explicit return path instead of a default reset.
03
Authentication becomes an obstacle course
- Observed pattern
- Overlapping codes, inconsistent input behavior, unexplained eligibility failures, or repeated verification requests prevent reliable access.
- Human impact
- The security process itself increases confusion and reduces confidence without making the person more informed or secure.
- CLEAR test
- Clarity • Legitimate Consent • Attention Integrity
- Required repair
- Allow one active verification flow, explain failures precisely, preserve user choice, and provide consistent input and notification behavior.
04
Authorship and labor vanish during failure
- Observed pattern
- A platform permits substantial work but provides no durable draft state, reliable export confirmation, or visible recovery pathway.
- Human impact
- The cost of instability is externalized as lost intellectual labor, repeated reconstruction, and weakened provenance.
- CLEAR test
- Clarity • Legitimate Consent • Restorative Pacing
- Required repair
- Create durable drafts, visible version history, truthful completion signals, recoverable provenance, and a repair pathway when work is lost.
05
Placeholder identity becomes real data
- Observed pattern
- A common human name used as example text is ingested, duplicated, or propagated as if it were a verified identity.
- Human impact
- The real person must compete with a synthetic record while corrections may fail to reach downstream systems.
- CLEAR test
- Clarity • Legitimate Consent • External Presence
- Required repair
- Prohibit real names as placeholders, isolate synthetic records, overwrite contaminated values, and verify downstream correction and decay.
06
Acknowledgment becomes a loop without closure
- Observed pattern
- A concern is repeatedly received or routed, yet no responsible owner, durable resolution, or confirmed endpoint appears.
- Human impact
- Silence or repeated intake functions as de facto closure while the underlying failure remains active.
- CLEAR test
- All five CLEAR principles
- Required repair
- Require recognition, accountable ownership, verified remediation, communication to the affected person, and a permanent closure record.
Implementation standard
What adoption requires.
Organizations adopt CLEAR by changing system behavior—not by displaying a badge.
- 01
Maintain identity continuity
Recognize the same person and issue consistently across interactions, channels, and recovery events.
- 02
Distinguish message origin
Make automated, assisted, and human responses legible without forcing the person to infer who is speaking.
- 03
Align with user intent
Respond to the stated goal before redirecting, expanding scope, or optimizing for continued engagement.
- 04
Avoid duplication and fragmentation
Carry forward verified context so people are not required to repeat information already provided.
- 05
Provide pause and exit pathways
Allow a person to stop, defer, disengage, or return later without penalty, pressure, or loss of work.
- 06
Close the repair loop
Confirm what changed, communicate the outcome, preserve evidence, and make unresolved items visible.
Institutional commitment