Twitter | Search | |
Philip E. Tetlock
Annenberg University Professor, Wharton & School of Arts and Sciences (Psychology & Political Science)
1,499
Tweets
216
Following
24,961
Followers
Tweets
Philip E. Tetlock Jan 29
I'm reasonably confident that Ian's "reasonable confidence" means roughly 60-80%, which is less vague than a lot of vague-verbiage forecasts. The clearer you make your forecasts, the easier it is to spot mistakes & become better calibrated
Reply Retweet Like
Philip E. Tetlock Jan 26
Let’s lock in a mutual-admiration triad: I think very highly of both Tim’s and Matthew’s work. And we are not only complimentary. We are complementary. Read us to discover why
Reply Retweet Like
Philip E. Tetlock retweeted
Matthew Syed Jan 25
Totally agree with . It’s a superb book
Reply Retweet Like
Philip E. Tetlock Jan 23
Feuds between rival methods can be productive. But I see polling & prediction markets as each adding forecasting value. And here is some evidence of how these methods can complement each other:
Reply Retweet Like
Philip E. Tetlock Jan 21
This superb article illustrates critical-reasoning skills that are also at the heart of the “superforecasting” research program. We’re all aiming at the same cognitive target:
Reply Retweet Like
Philip E. Tetlock Jan 20
Business as usual at Davos: 1.The higher the status of forecasters, the vaguer their forecasts; 2.The more politicized the topic, the vaguer the forecasts; 3.The higher the stakes, the fewer the chances to learn from big shots' forecasting mistakes
Reply Retweet Like
Philip E. Tetlock Jan 16
in that case, it's also "official" that the World Economic Forum needs a remedial course in Research Methods. Correlation doesn't "mean" causality--and that is even true for claims that are pitch perfect for corporate PR purposes.
Reply Retweet Like
Philip E. Tetlock Jan 13
Memory lane: my first grad-school publication (1977) & basic finding holds up fairly well. Rising or falling integrative complexity of messaging predicts which crises escalate into violence. Credit: my first mentor, Peter Suedfeld.
Reply Retweet Like
Philip E. Tetlock Jan 8
“Superforecasters” did so well in tournaments because our scientific competitors were allergic to the cognitive elitism of tracking above-average people into superteams. Tracking had exactly the inequality-amplifying effect that egalitarians fear tracking in schools will have
Reply Retweet Like
Philip E. Tetlock Jan 6
Here’s a robust psychological effect that does not wilt under replication scrutiny. Kurt Lewin noticed it in the 1930s: making public commitments “freezes” attitudes in place. So saying something dumb makes you a bit dumber. It becomes harder to correct yourself. Tweeters beware.
Reply Retweet Like
Philip E. Tetlock Jan 2
How to save lives and money: Define forecasting accuracy as skill at achieving both a high Hit rate & low False-Positive rate. It’s trivially easy to claim you can predict every recession, war,… when no one is tracking your False-Positive rate.
Reply Retweet Like
Philip E. Tetlock Dec 29
We love turning points. Would China have grown faster, slower, or about the same in a world in which the Politburo liberals had prevailed? No one knows for sure but we can be pretty darn sure what current Politburo thinks
Reply Retweet Like
Philip E. Tetlock Dec 29
A thoughtful effort to anticipate key themes in the 2020s. Unfortunately, its first falsifiable forecast is false. People often over-predict change (e.g., Ch. 2, Expert Political Judgment). One reason: we get more credit for correctly predicting change than the boring status quo
Reply Retweet Like
Philip E. Tetlock Dec 28
Giant oaks from tiny acorns grow—& we still haven’t figured out how to spot which acorns will dominate our future (and we probably never will) The tweet below nails it
Reply Retweet Like
Philip E. Tetlock Dec 22
Very readable article on a very tricky topic: when do we have good grounds for believing something?
Reply Retweet Like
Philip E. Tetlock Dec 21
Vague-verbiage forecasts plant a jumble of probabilities in our minds that easily sum to < or > 1.0. So 2020 “could be” 1.Xi’s worst year (say 20%-80% range) year (say 10-40%) 3.a mix of successes & setbacks (say10-70%) Why not just give your best guess?
Reply Retweet Like
Philip E. Tetlock Dec 20
And not just right but "brilliant." Do you want a scoring system in which pundits can compensate for being far off on a hard forecasting problem (a year off) by scoring a bull's eye on an much easier one (extrapolate the polls a month out)?
Reply Retweet Like
Philip E. Tetlock Dec 20
It looks like cost-free tomfoolery—but it's really costly. It means alpha-pundits have no incentive to become more accurate—& lower-status challengers, who got it right sooner, have no chance to be noticed
Reply Retweet Like
Philip E. Tetlock Dec 19
"Superforecasters" are better at tuning in signals—& tuning out noise. It’s not the millions of words had no effect. They had lots of effects, mostly tiny offsetting ones
Reply Retweet Like
Philip E. Tetlock Dec 18
So refreshing: i like the modesty--and eclecticism. Would like to explore possible joint work with IARPA program known as FOCUS (feel free to e-mail).
Reply Retweet Like