Canonical terms

These are the project definitions used across the research site.

Agent

A temporally persistent system that maps observations and internal state to actions in an environment according to a policy, in service of an objective. Russell & Norvig ; Sutton & Barto ; Åström & Murray

Single agent

One decision-making system acting on one environment. Sutton & Barto ; Åström & Murray

Policy

The rule or function that maps decision state to action. Sutton & Barto ; Åström & Murray

Objective

The latent criterion of success that the system is trying to achieve. Russell & Norvig ; AERA/APA/NCME validity standards overview

Goal

What counts as success for the system. Russell & Norvig ; Sutton & Barto

Environment

The world or system the agent interacts with and can affect. Russell & Norvig ; Sutton & Barto

Observation

What the agent receives from the environment. Åström & Murray ; Sutton & Barto

State

The information the agent uses to decide. State may include current observations, internal variables, estimates, and memory. Sutton & Barto ; Åström & Murray

Memory

Retained state across steps. This includes short-term context, long-term traces, prior task progress, and learned heuristics. Sutton & Barto ; POMDP belief-state framing

Action

What the agent does to affect the environment. Russell & Norvig ; Sutton & Barto

Error

The gap between current behavior or outcome and desired behavior or outcome. Åström & Murray ; optimal control framing

Steering

Changing behavior in response to error and feedback. Åström & Murray

Measurement

The observable signal produced by tests, traces, audits, judges, sensors, or logs. Cronbach & Meehl ; AERA/APA/NCME validity standards

Proxy

A measurable stand-in for the objective. Cronbach & Meehl ; Goodhart’s law overview

Evaluation

The function that aggregates measurements and proxies into a score or decision signal. AERA/APA/NCME validity standards ; balanced scorecard framing

Control input

A deliberate change applied to steer the system. Åström & Murray

Controller

The mechanism that selects control inputs or actions. Åström & Murray ; Sutton & Barto

Dynamics

How actions and environment evolution change future state. Åström & Murray

Feedback loop

Observe → interpret → decide → act → update. Åström & Murray

Stability

Whether the system converges, oscillates, or diverges under feedback. Åström & Murray

Robustness

Whether the agent still works under noise, delay, partial observability, or perturbation. Åström & Murray

Partial observability

When the agent cannot directly see the full state of the environment. POMDP framing ; Åström & Murray

Estimation

Inferring hidden or incomplete state from observations. Åström & Murray

Policy update

A change to the policy based on experience, learning, or feedback. Sutton & Barto ; adaptive control framing

Model

A representation of how the environment or system behaves. Russell & Norvig ; Åström & Murray

Goodhart effect

When optimization pressure causes a proxy to stop tracking the true objective well. Goodhart’s law ; variants of Goodhart’s law

When another page needs a definition, link here instead of restating the term.

Source basis