Back to LessWrong

Outside view

From Lesswrongwiki

Revision as of 06:22, 21 July 2013 by Evan Erickson (Talk | contribs)

(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to: navigation, search

Taking the outside view (another name for reference class forecasting) means using an estimate based on a class of roughly similar previous cases, rather than trying to visualize the details of a process. For example, estimating the completion time of a programming project based on how long similar projects have taken in the past, rather than by drawing up a graph of tasks and their expected completion times. The planning fallacy is that people tend to be hugely optimistic when visualizing the details of a case, and become even more optimistic as they visualize more details.

Example 1: Japanese students expected to finish their essays an average of 10 days before deadline. The average completion time was actually 1 day before deadline. When asked when they'd completed similar, previous tasks, the average reply was 1 day before deadline. (Buehler, R., Griffin, D., & Ross, M. 2002. Inside the planning fallacy: The causes and consequences of optimistic time predictions. Heuristics and biases: The psychology of intuitive judgment, 250-270. Cambridge, UK: Cambridge University Press.)

Example 2: Students instructed to visualize how, where, and when they would perform their Christmas shopping, expected to finish shopping more than a week before Christmas. A control group asked when they expected their Christmas shopping to be finished, expected it to be done 4 days before Christmas. Both groups finished 3 days before Christmas. (Buehler, R., Griffin, D. and Ross, M. 1995. It's about time: Optimistic predictions in work and love. European Review of Social Psychology, Volume 6, eds. W. Stroebe and M. Hewstone. Chichester: John Wiley & Sons.)

It is controversial how far the lesson of these experiments can be extended. Robin Hanson argues that this implies that, in futurism, forecasts should be made by trying to find a reference class of similar cases, rather than by trying to visualize outcomes. Eliezer Yudkowsky responds that this leads to "reference class tennis" wherein people feel that the same event 'obviously' belongs to two different reference classes, and that the above experiments were performed in cases where the new example was highly similar to past examples. I.e., this year's Christmas shopping optimism and last year's Christmas shopping optimism are much more similar to one another, than the invention of the Internet is to the invention of agriculture. If someone else then feels that the invention of the Internet is more like the category 'recent communications innovations' and should be forecast by reference to television instead of agriculture, both sides pleading the outside view has no resolution except "I'm taking my reference class and going home!"

Blog posts

See also