PER: Bent Flyvbjerg

PUMA Context

Overview

Coined the “Uniqueness Trap” — the tendency of project managers to treat each project as absolutely unique, blocking statistical learning from historical data. This is one of PUMA’s three motivating problems.

Key Contribution

Reference Class Forecasting (RCF) as antidote to planning optimism bias. PUMA’s Stage 4 (RAG) can be seen as a computational implementation of RCF.

Relevance to H2: The “mean historical SP” baseline in Stage 2 is a simple form of reference class thinking. A well-designed LLM few-shot agent could do better by selecting a more specific reference class.

Related notes:

• PN-IssueTriage-StoryPoints — Story points estimation (H2)
• PN-KeyConcepts-Agents-Reproducibility-RedTeam — Uniqueness Trap concept
• PN-CoT-FewShot-Prompting — Few-shot as reference class
• EX-Hypotheses-H1-H2 — H2 rationale
• PR-PUMA-Ch1-Introduction — §1.1 problem 3
• Smart-PMO-Vision
• MOC-PUMA-Master