Contestability
Ability for affected people to challenge, change, or overturn decisions with visible pathways.
View glossary entryPublic memory for contestability terms
A living record of how accountability language shifts so teams can cite stable definitions and avoid drift.
Summary
Keep contestability language anchored
This page tracks how governance terms shift over time so teams can keep requirements enforceable.
Key takeaways
- Document how popular terms drift into checklists without remedies.
- Use stable glossary anchors for contestability, accountability, and time transparency.
- Attach sources so policy language stays traceable to its origin.
Related sections
Why this matters
Language drift hides accountability gaps
When words evolve without evidence, accountability promises get weaker.
Contestability terms travel through policy, procurement, and public narratives. When a term like “responsible AI” becomes a label rather than a measurable commitment, people lose the ability to demand remedies. This page preserves the original intent behind governance language so it can be cited with precision.
- Terms must map to enforceable actions, not just internal checklists.
- Stable definitions make audits, procurement, and public accountability easier.
- Documenting drift keeps teams honest about what is actually delivered.
Timeline
Language drift and memory log
A snapshot of how accountability vocabulary has shifted across policy and practice.
-
2019
Trustworthy AI becomes the umbrella phrase.
OECD principles frame trustworthy AI around transparency, robustness, and accountability. The phrase later drifts into a compliance label without enforceable remedy pathways.
Source: OECD AI Principles -
2020
Responsible AI consolidates around vendor playbooks.
Industry guidance emphasizes responsible AI toolkits, but often omits binding ownership and appeal guarantees. The language shifts from shared obligations to internal process checklists.
Source: Google responsible AI practices -
2022
Rights framing enters public policy.
The White House AI Bill of Rights centers notice, explanation, and human alternatives, yet enforcement remains discretionary. The term “rights” is used while review clocks and escalation authority stay vague.
Source: Blueprint for an AI Bill of Rights (White House archive) -
2023
Risk management vocabulary takes hold.
NIST AI RMF codifies governance and accountability functions, but organizations often stop at risk scoring without publishing recourse evidence. The language drifts toward risk acceptance without remedy capacity.
Source: NIST AI RMF 1.0 -
2024
Regulatory language converges on transparency and oversight.
The EU AI Act emphasizes transparency, human oversight, and documentation. Without a shared contestability baseline, implementation can still omit appeal passage rates and enforceable reversals.
Source: EU AI Act overview
Stable definitions
Terms to cite verbatim
Use glossary anchors so governance language stays enforceable.
Explainability for accountability
Actionable explanations that name decision owners, evidence, and correction paths.
View glossary entryAppeal passage rate
Percentage of appeals resolved in favor of the user, signalling real contestability.
View glossary entryTime transparency
Clear, legible process state with clocks, blockers, and escalation triggers.
View glossary entryDesign authority
Named owners with the power to set safeguards and approve reversals.
View glossary entryHow to use
Turn memory into enforceable requirements
Treat this page as a living appendix for briefs, audits, and public statements.
- Cite glossary anchors in briefs, audits, and press summaries so terms stay stable.
- Pair policy language with measurable evidence (appeal passage rates, review clocks, remedy logs).
- Document drift: track when terms are used as labels without enforcement mechanisms.