An event is not merely something that happened. It is a bounded change through time, observed from a particular frame.
A protest, election, earthquake, merger, examination, publication, medical episode, launch, accident and scientific discovery are all events, but they differ in cause, duration, scale, evidence, actors, consequences and relationship to larger processes.
Categorising events well requires more than naming the topic. It requires deciding what the event is, where its boundary begins and ends, which dimensions matter, and how uncertainty is preserved.
Quick answer: how should events be categorised?
- Type: what kind of event is it?
- Time: when did it begin, peak and end?
- Duration: instantaneous, short, prolonged or recurring?
- Place: where did it occur?
- Actors: who or what participated?
- Cause: what conditions or triggers produced it?
- Scale: local, regional, national, global?
- Consequence: what changed afterward?
- Evidence: how do we know it happened?
- Confidence: which parts remain uncertain or disputed?
This is an application of the general framework in How to Categorise Anything.
1. Define what counts as the event
An event needs a unit. “The financial crisis” may refer to one market shock, a multi-year sequence, or a whole institutional episode.
2. Event boundaries are analytical choices
Start and end dates may be obvious for a scheduled examination and contested for a revolution, epidemic or social movement.
3. Separate event from process
An event may be a discrete episode inside a longer process. A bank failure is an event; financial deterioration can be a process.
4. Separate trigger from background condition
A trigger may initiate the visible event while deeper conditions made the event possible.
5. Event type is only one dimension
Calling something an election, accident or discovery does not fully describe its temporal, causal or geographic character.
6. Time should be explicit
Store date, time, period and timezone where relevant. Historical events may need uncertain or approximate dating.
7. Some events are instantaneous
A collision, signature, announcement or measurement can occur at a defined moment even if consequences continue afterward.
8. Some events have duration
Wars, strikes, festivals, outbreaks and negotiations unfold across intervals and may contain many sub-events.
9. Recurring events need series identity
An annual examination or recurring meeting is one event type with many instances. Preserve both series and occurrence.
10. Location may be point, area or network
An event can happen at one building, across a city, throughout a country, online, or across connected transport nodes.
11. Actors deserve typed roles
Initiator, participant, witness, regulator, victim, observer and responder are different roles.
12. Participation is not causation
An actor present in an event should not automatically be classified as its cause.
13. Causes can be layered
Distinguish trigger, enabling conditions, structural causes and contributing factors.
14. Causation may be disputed
Store competing causal claims with provenance rather than forcing one explanation prematurely.
15. Scale should match observed reach
Local, regional, national and global are useful only when their criteria are clear.
16. Magnitude is not the same as scale
A highly consequential local event may have greater impact than a low-intensity national event.
17. Consequences can be immediate and delayed
Store short-term effects separately from long-term outcomes where the distinction matters.
18. Consequences can be direct or indirect
A transport shutdown directly halts movement and indirectly affects work, supply and appointments.
19. Event severity needs domain-specific criteria
Severity in weather, medicine, finance and education cannot be defined by one universal scale.
20. Event evidence has multiple forms
Records, sensors, eyewitnesses, photographs, logs, official reports and physical traces may each support the event claim.
21. Evidence should be separated from interpretation
A video may show that something occurred without proving why it occurred.
22. Provenance makes event classification auditable
Store who recorded the event, when, from which source and under which classification scheme.
23. Historical events need source-aware classification
Different archives or historians may use different event boundaries or names. Preserve those viewpoints.
24. Events can nest
A conference contains sessions; a war contains battles; an election contains nomination, campaigning, polling and counting events.
25. Nesting should preserve level
Do not compare a whole war with one battle as though they were equivalent event units.
26. Events can overlap
Economic, political and cultural events may unfold simultaneously and influence one another.
27. Multi-label classification may be appropriate
One event can legitimately be both political and economic, or both scientific and institutional.
28. Event chains should represent dependency
Event A may enable B, trigger C or occur after D without being caused by D. Use typed relationships.
29. Sequence is not causation
Earlier events are not automatically causes of later events.
30. Event uncertainty should be local
The occurrence may be certain while date, actor, cause or consequence remains disputed.
31. Approximate dates need explicit representation
Use circa, ranges or confidence rather than converting uncertainty into false precision.
32. AI can extract candidate events
Models can detect event mentions, actors, times and locations from text, but extracted fields should remain evidence-linked.
33. Event extraction can confuse mention with occurrence
A document can discuss a hypothetical or denied event. Classification should detect modality and source context.
34. Event taxonomies need versioning
New event types appear and old concepts split. Preserve historical codes and crosswalks.
35. Search benefits from event facets
Users may filter by type, time, place, actor, severity and status independently.
36. Event classification can drive timelines
Once dates, types and relations are structured, events can be traversed as chronological and causal sequences.
37. Event classification can drive alerts
Operational systems can route event types to different responders, provided thresholds and consequences are governed.
38. A practical event record
- event ID;
- event type;
- start time;
- end time;
- location;
- actors and roles;
- cause claims;
- scale;
- severity;
- consequences;
- evidence;
- confidence;
- related events;
- taxonomy version.
39. Events are best understood as nodes in time
They connect states before and after, actors, locations, evidence and consequences.
40. The deeper idea
Categorising an event means deciding which change in the world deserves one identity and which dimensions must be preserved so that change remains interpretable.
An event is a cut through time. Good classification records where we made the cut and why.
Final answer
Categorise events by type, time, duration, location, actors, causes, scale, consequences and evidence. Separate events from processes, triggers from background conditions and sequence from causation. Preserve uncertainty and provenance, allow nesting and overlap, and use typed relationships to connect event chains.