Exhibit 01Reflex Agent
Vacuum World
Two rooms, one robot, dirt that keeps coming back. The textbook reflex agent is two if-statements. Can a decision model behave like one?
In scopeA snap judgment inside a loop that code controls. TypeSafe docs, How to build: “Keep control flow, deterministic rules, and side effects in code.”
World
→State
→Decision
→Action
→New state
Dirt appears
dirt seed 101
Same dirt, two agents
Cleanliness = share of room-steps clean (AIMA's performance measure)
Classic rulesJev
Steps00
Cleanliness——
Moves00
Waits00
Wasted actions00
Jev decision
idlePress Step to let the model act once, or Auto run to watch it keep the rooms clean.
CLEAN—
MOVE RIGHT—
WAIT—
Lesson · Reflex
Observe → Decide → Act
The reflex agent is the simplest thing in the textbook: perceive, match a rule, act. Here it is two if-statements, and it keeps the rooms mostly clean for free. The comparison is not like for like: the textbook rule sees only the room it is in, while Jev is shown both rooms. That is why Jev can wait when both rooms are clean while the rule keeps shuttling, and a three-line rule given the same view of both rooms behaves the same way. Both agents face the same seeded dirt. What a decision model adds here is not better cleaning but a small, inspectable choice with a probability attached, doing work a rule already does.