Module Check: Think Before You Act
You have traveled through the full arc of intelligent planning — from the simple insight that big tasks cannot be done in one step, all the way through the architecture of a real planning agent. This lesson is your chance to consolidate everything: recall the key terms, test your understanding with fresh questions, and apply the module's ideas in one final synthesis activity.
Flashcards — click each card to reveal the answer
Module Check Questions
An AI agent is given the goal: organize and send a monthly report to 200 clients. It immediately begins writing the report body without any planning. What is the most likely consequence?
Which of the following correctly describes the relationship between decomposition and subgoals?
A planning agent backtracks to the junction where it chose Path A over Path B. What exactly does it preserve, and what does it discard?
A student estimates a research task at 2 hours and applies the planning fallacy defense by multiplying by 1.5. She records 3 hours as her estimate. The task actually takes 4.5 hours. What does this tell her about her baseline estimates?
An AI planning agent has four components: goal interpreter, planner, executor, and monitor. A web search returns an error that was not in the plan. Which component is responsible for detecting this and deciding whether to replan?
Why is contingency planning generally more efficient than repair planning after a failure occurs?
Synthesis: Trace an Agent Through a Full Plan
- In this final activity, you will trace an AI agent through an entire planning episode — all the way from goal to completion, including a failure and recovery.
- THE SCENARIO: An AI agent is asked to: Find the three highest-rated science fiction novels published in the last 5 years, write a 200-word summary of each, and compile them into a formatted document called sf-report.pdf.
- YOUR TASK:
- Part 1 — DECOMPOSE: Write the agent's decomposition tree for this goal. Include all major subgoals and leaf tasks. Aim for at least 12 leaf tasks.
- Part 2 — SEQUENCE: List the leaf tasks in the order the agent should execute them. Mark any that can run in parallel.
- Part 3 — ESTIMATE: Assign a time estimate to each leaf task. Apply a 1.5x planning fallacy correction to each estimate. Sum the total estimated time.
- Part 4 — SIMULATE A FAILURE: After completing the search for the second novel, the agent discovers the author's official website (which it planned to cite) is offline. Describe: What does the monitor detect? Is this a dead end or an obstacle? What does the agent do — backtrack, repair, or apply a contingency?
- Part 5 — REFLECT: Write a paragraph answering: What would have happened if this agent had been reactive instead of a planning agent? How would the failure have been handled differently — or not at all?