people are overconfident
once you know that, you do not think it applies to you. so all you learn is that others are over-confident.
linear models are better than experts. so interviews are not good predictors. they are good for predicting future interview performance. e.g kellog’s MBA started interviews. students became taller and prettier. and got higher salaries after MBA.
simple linear (improper) models do very well. better than committees.
so, why will we not change? why are we skeptical.
‘broken leg’ argument. robyn dawes. you recruit a player to a team without seeing them and they have a broken leg. this also can be introduced to the model. how attractive is a firm that uses a model. you are weird if you do not. feel-good, loyalty, advantage in competitive scenario, but not for selection. model predicts cultural compatibility? lacks collegiality. a lunch interview is a good predictor of a future lunches.
it is tough to get others to agree and to use the model. do you need variables from experts? experts want control.
if the model is different than intuition, may want to ask what would the model have to look like?
what decisions would you let a model make.
so, a model may be a good alphabet for group decisions.
Cain says: trust your gut when
- it may pick on something that is not in the data – something that makes sense. is there a story why it may make sense?
- for simple decisions
- when there is no time
- when deciding about your gut, especially if there is ‘gut data’ over time
how to do well on interviews:
claim: memory is recreated from start, end, and extremes. with huge recentcy effect
- firm handshake
- first impressions
- a few high points
- minimum low points
- great ending
on recentcy. if the end is not as bad as the previous experience, you are more likely to proceed than if the last experience is not so bad but without comparison.