ºÝºÝߣ

ºÝºÝߣShare a Scribd company logo

In Praise of Epistemic Irresponsibility

?Download as PPTX, PDF?
?
Boris Yakubchik
Boris Yakubchik

The document discusses the idea of epistemic irresponsibility and how being lazy and ignorant can sometimes lead to better results than relying on experts. It argues that simple linear models can often outperform experts in domains like college admissions, hiring decisions, and medical diagnosis. The document recommends some strategies for developing simple linear models with little effort, such as bootstrapping off one expert or using random weights. It also suggests averaging multiple estimates or predictions as a way to improve results over relying on a single estimate.

1 of 12
Download to read offline
In praise of epistemic irresponsibility: How lazy and ignorant can you be? Boris Yakubchik @Forbes June 5th, 2018
Epistemology ~ theory of knowledge Experimental philosophy ~ philosophy + empirical inquiry
We need: ¡ñ reliable methods ¡ñ accessible to regular people for making conclusions Traditional epistemology sucks
What we need: Reliably good results Traditional virtue in epistemology: Hard work
Experts can really suck Wide range of domains: ¡ñ College admissions ¡ñ Parole officers ¡ñ Hiring decisions (unstructured interviews) ¡ñ Medical professionals ¡ñ etc Linear Models rule
Experts can really suck Linear Models rule Why? pourquoi? No feedback loops
Linear model ~ weight x variable + weight x variable + ¡­ + weight x variable = value 0.4 x 0.5 + -0.6 x 0.7 + ¡­ + 0.1 x 1.3 = 0.8 Variables Outcomes Good statistics Linear model
Time to get lazy Himouto! Umaru-chan
How lazy can we get? ¡ñ Bootstrap your model ¡ð A linear model off just one expert ¡ñ Random linear model ¡ð Random weights but non-random sign (+/-) ¡ñ Ignore most variables ¡ð Simply add a few normalized scores Experts hate him! Pay attention for one weird trick:
Simple strategy for better results: 1. Estimate / make prediction 2. Consider the opposite / assume you are wrong 3. Make second estimate / prediction 4. Average the two
Don¡¯t think about probabilities, think about frequencies 1. Patient thinks she has a rare disease (1 in 1000) 2. Test is 99% accurate (1% chance it is wrong each administration) 3. Test is positive 4. What is the probability the patient has disease? hint: only 18% of Harvard Medical School faculty & staff got an equivalent problem right hint: average diagnosis was 28 times too high
Solution: 99% correct 1% incorrect 999 people disease free 1 person has disease 99 people 999 people 99 99 + 999 9%=

More Related Content

In Praise of Epistemic Irresponsibility

  • 1. In praise of epistemic irresponsibility: How lazy and ignorant can you be? Boris Yakubchik @Forbes June 5th, 2018
  • 2. Epistemology ~ theory of knowledge Experimental philosophy ~ philosophy + empirical inquiry
  • 3. We need: ¡ñ reliable methods ¡ñ accessible to regular people for making conclusions Traditional epistemology sucks
  • 4. What we need: Reliably good results Traditional virtue in epistemology: Hard work
  • 5. Experts can really suck Wide range of domains: ¡ñ College admissions ¡ñ Parole officers ¡ñ Hiring decisions (unstructured interviews) ¡ñ Medical professionals ¡ñ etc Linear Models rule
  • 6. Experts can really suck Linear Models rule Why? pourquoi? No feedback loops
  • 7. Linear model ~ weight x variable + weight x variable + ¡­ + weight x variable = value 0.4 x 0.5 + -0.6 x 0.7 + ¡­ + 0.1 x 1.3 = 0.8 Variables Outcomes Good statistics Linear model
  • 8. Time to get lazy Himouto! Umaru-chan
  • 9. How lazy can we get? ¡ñ Bootstrap your model ¡ð A linear model off just one expert ¡ñ Random linear model ¡ð Random weights but non-random sign (+/-) ¡ñ Ignore most variables ¡ð Simply add a few normalized scores Experts hate him! Pay attention for one weird trick:
  • 10. Simple strategy for better results: 1. Estimate / make prediction 2. Consider the opposite / assume you are wrong 3. Make second estimate / prediction 4. Average the two
  • 11. Don¡¯t think about probabilities, think about frequencies 1. Patient thinks she has a rare disease (1 in 1000) 2. Test is 99% accurate (1% chance it is wrong each administration) 3. Test is positive 4. What is the probability the patient has disease? hint: only 18% of Harvard Medical School faculty & staff got an equivalent problem right hint: average diagnosis was 28 times too high
  • 12. Solution: 99% correct 1% incorrect 999 people disease free 1 person has disease 99 people 999 people 99 99 + 999 9%=