PEAS in AI
PEAS stands for Performance measure, Environment, Actuators, and Sensors. It is a structured framework introduced by Russell and Norvig for precisely chara...
Describing Agents with PEAS
PEAS stands for Performance measure, Environment, Actuators, and Sensors. It is a structured framework introduced by Russell and Norvig for precisely characterizing an intelligent agent before building or analyzing it. Without a clear PEAS description, designers may optimize the wrong objective or overlook important environmental factors.
The Performance measure defines what it means for the agent to succeed. It must be stated in terms of desirable world states, not the agent's internal behavior. For a taxi-driving agent, the performance measure might include safe arrival, minimal travel time, legal compliance, and passenger comfort — each weighted appropriately.
The Environment describes the external world the agent operates in. For the taxi agent this includes city roads, traffic signals, pedestrians, weather conditions, and other vehicles. Specifying the environment fully helps engineers identify what sensor data is needed.
Actuators are the mechanisms through which the agent affects the environment. The taxi agent's actuators include the steering wheel, accelerator, brake pedal, horn, and turn signals.
Sensors are the data collection mechanisms. The taxi agent uses GPS, cameras, radar, speedometer, and engine monitors.
A concrete example: consider an automated email-sorting agent. Its performance measure is the percentage of emails correctly categorized. Its environment is a corporate email server with users sending and receiving messages. Its actuators move emails into folders and flag urgent items. Its sensors read email headers, body text, and metadata such as sender history.
PEAS forces designers to think carefully about conflicting objectives. A medical robot's performance measure must balance speed, accuracy, and patient safety — optimizing for speed alone could be dangerous.
Exam tip: Practice writing PEAS descriptions for novel scenarios. Examiners often give a new agent scenario and ask you to fill in all four components correctly.