Learn it → Apply it → Save as Skill
Almost every plan looks achievable from the inside. The planning fallacy is the tendency to predict a task's time, cost and risk from the best-case story of how it will go, rather than from how similar efforts have actually turned out before.
The Dr. Wallace’s Favorite Mental Models is designed to turn useful ways of thinking into reusable tools. Each model gives you a fast visual overview, deeper interactive learning, practical application exercises and an AI-ready Markdown skill you can save for future use.
Instead of merely reading about mental models, you can gradually build your own AI skills library: a collection of portable reasoning skills that can help ChatGPT, Claude, Grok and other AI assistants apply the mental models you value.
When you estimate a task, you naturally picture the plan succeeding: no interruptions, no surprises, nothing going wrong. That best-case story is the inside view. The outside view looks instead at how comparable tasks actually turned out, which is almost always slower and costlier.
Important: the gap is not always this large, and it does not always resolve toward the outside view this cleanly. The direction, however, is consistent: plans run long, not short.
The model is most useful before an estimate becomes a promise: a deadline, a budget, a launch date or a plan someone else will rely on.
The correction is mechanical, not a matter of trying harder to be realistic.
State the task and your current best-guess estimate for time, cost or effort.
Find several comparable past projects or tasks, not just this one team's intentions.
Note how the reference class actually performed against what those teams originally predicted.
Move your estimate toward the reference class and add a buffer sized to its variance.
Open a use case to see what the model changes.
Before a sponsor locks in a date built from a milestone chart, ask for the track record of similar projects at this organization. A confident chart is not evidence; a completed history is.
Ask the contractor, or your own records, how long the last two or three comparable jobs actually took, not how long this one is scheduled to take. Permits, inspections and material delays are common, not exceptional.
Your own history is a reference class you already have. If the last three similar tasks each ran long, that pattern is more informative than this time's optimism.
Compare the estimate against how long the team's last several similarly sized features actually took to ship, including review, testing and integration, not just the coding.
Look at how far the last two or three launches slipped from their original announced date, and use that gap as the starting adjustment rather than a fresh optimistic guess.
Build a contingency line sized to the historical overrun rate for this category of spend, rather than a flat, unexamined percentage.
This account is drawn from Daniel Kahneman's own record of the episode, cited widely in behavioral-economics literature.
Expand for the studies behind the model and where the popular framing overreaches.
Daniel Kahneman and Amos Tversky described the pattern in the late 1970s, in work on intuitive prediction, naming the tendency to underestimate how long, how much and how risky a task will turn out to be.
Their 1994 study asked psychology students to predict how long their thesis would take. Students guessed 33.9 days on average but took 55.5 days, and only about 30 percent finished within their own estimate. Even their explicit worst-case guess, 48.6 days, still undershot reality.
In 2003, Dan Lovallo and Kahneman broadened the idea to costs and benefits, not only time: plans tend to underestimate cost while overestimating payoff, producing both overruns and shortfalls together. The Sydney Opera House, planned for 1963 completion at about $7 million, opened in 1973 at about $102 million. Denver International Airport opened 16 months late and more than $2 billion over budget.
Reference class forecasting, developed by planning researcher Bent Flyvbjerg, replaces inside-view detail with statistics from a class of genuinely similar past projects: find comparable cases, learn their actual outcome distribution, and place the current project inside it before adjusting for anything truly unique.
Picturing the plan succeeding smoothly is not the same as knowing how it usually goes. The rehearsal always omits the interruption you can't yet name.
Past projects of yours are a reference class you already own. A pattern of running late is information, not bad luck repeating itself.
Adding a flat, arbitrary buffer feels responsible but is still a guess unless it is sized to how comparable tasks actually varied.
Overruns are usually the sum of many small unplanned items, not a single villain. Naming one cause hides the pattern that will repeat next time.
A project plan describes the intended path. It is not automatically a probability-weighted prediction of when the work will actually finish.
Team confidence and morale are real and valuable, but they do not correlate with whether a date will be met.
the task is genuinely unprecedented, with no meaningful reference class of similar past efforts to draw on.
the task is short, routine and heavily practiced, so recent personal history already predicts it accurately.
stakes are high, the plan will be made public, and commitment will make the date hard to revise once announced.
This browser tool structures your thinking. It does not send your information anywhere or perform AI analysis.
Save as Skill converts the model into portable Markdown instructions that teach an AI when and how to apply the planning fallacy, what questions to ask, what evidence to seek, what mistakes to avoid and how to structure its analysis.
Save the skills you find useful and organize them into your own reusable library. The Markdown is intentionally model-neutral, so it can be used wherever an AI system accepts uploaded files, project knowledge, custom instructions or reusable text context. Platform capabilities differ, so this is a portable skill file rather than a claim of one-click installation.
No. The Buehler, Griffin and Ross studies found the effect even among people who described themselves as cautious planners. The bias comes from focusing on a specific plan rather than from a personality trait.
Both. Lovallo and Kahneman's 2003 extension showed that plans tend to underestimate cost and overestimate benefit together, which is why large public projects often disappoint on both the schedule and the payoff at once.
Experience with the outside domain helps only if it is used as a reference class. Experienced planners who still reason from a fresh inside-view story for each new project remain just as susceptible as beginners.