Conceptual Policy – Treatment of Uncertainty

This page defines how uncertainty is handled in the data model.

It clarifies why the system does NOT implement a formal "Certainty" entity or certainty levels.

This decision is intentional and part of the conceptual design.

Principle

Historical research rarely produces absolute facts.

Most statements involve:

approximation
interpretation
incomplete sources
conflicting evidence

Uncertainty is therefore normal and expected.

The model must support nuance, not force artificial precision.

Decision

The data model does NOT implement:

Certainty entities
Certainty levels
Confidence scores
Probability flags
"High / Medium / Low" classifications

Uncertainty is not stored structurally.

It is expressed descriptively.

Rationale

1. Avoid false precision

Labels such as:

High
Medium
Low

suggest scientific measurement.

Historical interpretation is rarely measurable in this way.

Such labels create a false sense of objectivity.

2. Reduce editorial burden

Formal certainty fields would require editors to:

choose levels constantly
interpret ambiguous categories
make arbitrary decisions

This slows work and produces inconsistent data.

3. Preserve nuance

A short note such as:

"Only mentioned once in a newspaper article"

contains far more information than:

"certainty = low".

Free text preserves context and reasoning.

4. Simpler model

Removing certainty:

reduces schema complexity
simplifies forms
lowers cognitive load for contributors
improves usability

How uncertainty is expressed instead

Uncertainty is handled through:

precise wording
descriptive notes
date ranges
approximate values
explicit explanations
source citations (DigitalAssets)

Examples:

date: 1952
date: between 1950 and 1955
notes: probably owned by Company X
notes: attribution disputed in literature
citation to source DigitalAsset

Future reconsideration

If future requirements demonstrate a real need for structured certainty (e.g. automated reasoning or statistical analysis), this decision may be revisited.

Until such need appears, descriptive practice is preferred.

Status

This policy is part of the conceptual model – Version 3.2.

All schemas and implementations must follow this principle.

Procedure – Handling Uncertainty in Research Data

This page explains how researchers should record uncertain or incomplete information.

Uncertainty is normal in historical research.

The goal is clarity and transparency, not artificial precision.

General rule

When information is uncertain:

Describe the situation clearly.

Do not invent precision.

Dates

Use the most honest level of precision available.

Examples:

18/08/1952 → exact date known
1952 → year only known
1950–1955 → estimated range
empty → unknown

Do not add artificial detail.

Avoid:

guessing a full date
adding placeholders

Roles and attributions

If a role is uncertain, state this in words.

Examples:

architect (attributed)
probably owner
possibly involved in management

Use notes to explain the reasoning.

Avoid:

presenting assumptions as facts

Relationships

If a relationship is unclear:

Explain the source.

Examples:

mentioned in one newspaper article only
inferred from correspondence
not confirmed by archival records

Descriptions and notes

Use the Notes field to capture:

doubts
alternative interpretations
conflicts between sources
explanations of reasoning

Notes are for researchers and should be explicit.

Sources and citations

Always support claims with DigitalAssets.

Instead of marking certainty levels:

cite the source
let readers evaluate the evidence themselves

Example:

The sanatorium opened in 1923.<ref>DA:Opening ceremony article</ref>

What to avoid

Avoid:

inventing precise values
hiding uncertainty
using vague language without explanation
creating artificial certainty labels

Guiding principle

Prefer:

clear explanation

over

formal scoring.

Summary

Be honest. Be transparent. Document reasoning.

Uncertainty is normal and acceptable. Hidden assumptions are not.