error
Let an agent recognise, record, correct and communicate errors honestly, including its own, and distinguish error types and their consequences.
Research draft, second pass
A second pass drafted this model: the structure a model of this thing needs, and what is known about it in the world. The line under this one says how the second half was obtained - researched against sources, or recalled without web access, in which case nothing here was read anywhere and every claim is a lead to verify. Unreviewed either way.
written by Claude from model knowledge without web access - no source was read, every claim is a lead to verify
Researched by: Claude
Purpose and description
Let an agent recognise, record, correct and communicate errors honestly, including its own, and distinguish error types and their consequences.
A deviation from what is correct, true or intended, such as a mistake in a text, a measurement error, a software fault or a wrongful judicial decision.
What it is for: Errors are what quality, correction and accountability processes exist to find and fix.
It can be detect and report it; correct it and publish a correction; quantify it, as measurement error; learn from it to prevent recurrence.
Distinguishing features
Deviation from what is correct
Random and systematic errors differ
Consequences range from trivial to grave
Correction requires evidence and transparency
What it looks like
Varies: a typo, a wrong number, a failed test, a miscarriage of justice.
How it is recognised
Discrepancy between a claim and evidence
Corrections and errata notices
Error messages and logs in software
Related models
is a kind of - defect
is measured in - uncertainty
is corrected by - remedy
can cause - effects
In practice
Families and kinds
measurement error: random and systematic
human error
software bugs
errors in publications and records
judicial errors such as wrongful convictions
Standards and regulation
ISO/IEC Guide 98-3 (measurement uncertainty)
Press codes on corrections
Legal procedures for reviewing wrongful convictions
Failure modes and hazards
Errors left uncorrected
Blaming individuals instead of fixing systems
Overconfident claims hiding error
Also called
+140
Where this came from
wikidata · CC0 1.0
Also registered as vr.tr.error-communication
Drafted structure
Bundle to layer to finding to question, as the second pass will find it: 4 bundles · 8 layers · 8 findings · 16 questions.
Detection Finding the error.
Errors must be found before they can be fixed.
Evidence
What shows it is wrong.
Evidence
Evidence of error.
- What evidence shows this is an error? provenance
- Who found it? provenance
Type
Kind of error.
Type
Error type.
- Is it random, systematic, human or software error? definition
- What caused it? definition
Impact What it affected.
Impact decides urgency.
Scope
Who and what.
Scope
Affected items.
- What decisions or records did the error affect? boundary
- Who was harmed? boundary
Severity
How bad.
Severity
Severity.
- How serious are the consequences? measurement
- Is harm ongoing? boundary
Correction Fixing it.
Corrections must be visible.
Fix
Corrective action.
Correction
Correction made.
- How was the error corrected? action
- Was the correction published? provenance
Remedy
Redress for harm.
Remedy
Redress.
- What redress is owed to those harmed? boundary
- Through which process? provenance
Prevention Stopping recurrence.
Systems prevent repeats.
Root cause
Why it happened.
Root cause
Root cause analysis.
- What was the root cause? definition
- Was it systemic? boundary
Controls
Checks.
Controls
Preventive controls.
- Which checks would prevent it next time? action
- Who owns them? provenance
What the second pass must settle
- Should error types be separate entries?
- How should an agent record and disclose its own errors?
- The registry entry merges judicial and editorial errors; should they be split off?