Threat AnalysisDEMONSTRATION CLASSIFICATION

Context, not keywords, decides the classification

Deterministic, explainable rules over synthetic records. This is not a machine-learning model.

Scenario:

Original synthetic content

PIE-0001

Low

Community workshop in Jos: how parents and teachers can talk to youths about gun violence and conflict prevention.

Matched keywords:gun

Available conversation context

Public, fictional thread

Parent · participant_1 (fictional)
Interesting point.
Matching message · original poster (fictional)
Community workshop in Jos: how parents and teachers can talk to youths about gun violence and conflict prevention.
Reply · participant_2 (fictional)
Thanks for sharing.

Source metadata & timestamps

Source category: Public forum
Public alias: Public forum · commenter_ox (fictional)
Language: English
Published: 04 Nov 09:00 UTC
Observed: 04 Nov 09:36 UTC
Region: Plateau, North Central

Open in review workspace →

Context classification

Demonstration Classification

EducationalUncertainty: Low

Keyword(s) "gun" appear, but the surrounding context is educational framing.

Rule trace: R4: educational framing with no sale/contact language · N: topic and political terms recorded as neutral — they never raise a classification on their own

Relevant / supporting indicators

No supporting threat indicators.

Contradictory / benign indicators

“workshop” — Educational framing. Content teaches prevention or self-protection
“prevention” — Educational framing. Content teaches prevention or self-protection

Neutral topic terms

“gun” — Weapon reference. Topic term — not suspicious on its own

Words such as “gun”, “drugs” or election references are recorded but never raise a classification by themselves.

Recommended analyst action

No action. Educational content — do not escalate.

Decisions are saved to shared demo state and also update the record's review status. Cleared by “Reset demo”.

Same keyword, different context

Why two posts with the same word receive different classifications

Keyword: “gun”

Keyword: “drugs”

Evaluation against known labels

154/154 match

Known ↓ / Predicted →	Benign	News	Educational	Cybersecurity	Potential	Insufficient
Benign / Informational	43	0	0	0	0	0
News Reporting	0	29	0	0	0	0
Educational	0	0	18	0	0	0
Cybersecurity Threat	0	0	0	26	0	0
Potential Illicit Activity	0	0	0	0	22	0
Public Safety Concern	0	0	0	0	0	0
Insufficient Evidence	0	0	0	0	0	16

Class	n	Prec.	Recall
Benign	43	100.0%	100.0%
News	29	100.0%	100.0%
Educational	18	100.0%	100.0%
Cybersecurity	26	100.0%	100.0%
Potential	22	100.0%	100.0%
Public	0	—	—
Insufficient	16	100.0%	100.0%

Calculated live from the 154 synthetic records and their curated labels. The rules were written with knowledge of the same synthetic templates, so this is a consistency check, not a measure of real-world accuracy. “—” means no records carry or received that label, so no figure can be calculated.

Classification providers

Integration abstraction

NCIC demonstration rule engineActive (demonstration rules)
Hand-written deterministic rules. Not a machine-learning model.
Approved model serviceAwaiting approval
Placeholder for a future, formally approved model service. No model is connected.