Files
otivm/docs/training/corpus/Layer_2--Uncertainty/CORPUS-0009-same-event-different-knowledge.md
2026-04-30 08:33:12 -04:00

4.2 KiB

CORPUS-0009

Same Event, Different Knowledge

Status: Training Corpus Seed

Layer: Layer_2--Uncertainty

Purpose: Teach that different actors may know different parts of the same event because of position, status, timing, and access

Repository Path: docs/training/corpus/Layer_2--Uncertainty/CORPUS-0009-same-event-different-knowledge.md


0. Scenario

A timber barge is delayed near Ostia.

The event is one event.

But no single actor automatically knows the whole event.

A porter, trader, clerk, official, buyer, and cart owner may each know a different part.

The model must not treat “the city knows” as one shared knowledge state.


1. Hidden True Event

The simulation true state:

Hidden True State Value
Event timber barge delayed
Cause towline broke
Cargo dry beam stock and wheel blanks
Delay one day minimum
Damage minor
Destination buyers two carpentry shops
Official notice not yet filed

The true state exists.

Most actors know only fragments.


2. Actor Knowledge States

Actor What They Know
towman towline broke; delay likely one day
porter a timber barge is late
trader timber buyers are asking questions
warehouse clerk reserved storage may remain empty today
cart owner return load may be delayed
official no formal notice yet
carpenter expected timber has not arrived
buyer in another district may know nothing

Each actor has partial knowledge.


3. Why The Differences Matter

Different knowledge creates different actions.

The towman may seek rope.

The porter may repeat a vague delay rumor.

The trader may reserve alternative timber.

The warehouse clerk may release storage.

The cart owner may demand waiting pay.

The carpenter may bid for substitute material.

No one needs the entire truth to act.


4. Incorrect Model Behavior

The model should not:

  • assume all actors know the same facts
  • treat a city as one mind
  • give the trader official knowledge without a path
  • assume formal notice arrives before street knowledge
  • assume partial knowledge is useless
  • ignore actor position and access
  • collapse true event into public knowledge

5. Correct Model Behavior

The model should track:

Category Meaning
true_event_state what actually happened
actor_observed_signal what each actor directly observes
actor_report what each actor says or repeats
actor_inference what each actor believes follows
confidence how strongly each actor trusts the information
action what each actor does from partial knowledge

Knowledge should be actor-specific.


6. Decision Example

A trader hears from a porter:

Timber is late.

This is vague.

He then sees two carpenters near the quay asking about substitutes.

That visible behavior increases confidence that the delay matters.

He still does not know:

  • cause of delay
  • cargo condition
  • duration of delay
  • whether the timber is already reserved
  • whether alternate supply exists

The trader may still act before full knowledge.


7. Same Event, Different Profits

Different actors may benefit from different fragments:

Actor Possible Use
rope seller sells replacement towline
cart owner charges waiting time
trader sources substitute timber
carpenter secures scarce material early
clerk rents storage elsewhere
creditor checks exposed contractor

The same event creates different opportunities because knowledge differs.


8. Layer-0 And Layer-1 Concepts Used

This example uses:

  • Layer_0/CORPUS-0006-delay-is-economic-cost
  • Layer_0/CORPUS-0007-information-arrives-unevenly
  • Layer_0/CORPUS-0008-rumor-is-uncertain-information
  • Layer_0/CORPUS-0011-status-changes-access
  • Layer_0/CORPUS-0018-rivalry-changes-conditions
  • Layer_1/CORPUS-0007-rival-buys-the-cart-space
  • Layer_1/CORPUS-0011-round-trip-cart-value
  • Layer_2/CORPUS-0005-hidden-true-state-vs-known-state

9. Success Condition

If the model stops treating an event as equally known by all actors and starts tracking who knows which fragment, from what position, and with what confidence, this file is functioning correctly.