Definitions

Text (objective carrier)

Objective carriers exist regardless of who reads them: signals, traces, datasets, files, sensor streams, and physical marks.

Context (agent model)

The agent’s model: priors, expectations, goals, values, and policies used to interpret Texts and decide actions. Often denoted $Q$ when probabilistic.

Interpretation (mapping)

The mapping by which an agent transforms Text into meaning and action using its Context.

One Text, many Contexts (examples)

• Same courtroom transcript → different legal strategies for prosecution vs defense.
• Same radar return → different decisions for air‑traffic control vs weather service.
• Same royal ceremony → a cat notes a mouse; a lady‑in‑waiting notes dresses; a diplomat infers alliances.

Measurement hooks (surprisal, KL)

• Agent surprisal: $−log Q(i)$, how unexpected the observed $i$ is for this agent.
• Source surprisal: $−log P(i)$, how rare $i$ is under the source process.
• Mismatch: $D_{KL}(P \|\| Q)$, the gap learning aims to reduce.

Further reading: Dual Reality, Information: Objective vs Subjective, Agency & Delegation, Glossary.