“The Data Detective” by Tim Harford

Written by

in

  • Search your feelings
    • Excited, for or against, can be a sign of bad data
  • Ponder your personal experience
    • “Naïve reality” is when we think our perceptions are more accurate than they really are (most often because the news weights our perceptions to the novel)
    • We often substitute a hard question for an easier one, without knowing it
    • (Instead of “how many pregnant teenagers are there?” we ask “have I recently experienced a news story about pregnant teenagers”)
    • Be careful of how metrics are being leveraged
  • Avoid premature enumeration
    • Definitions can skew enumerations
    • Start by understanding what was meant
    • Once you learn something, it is difficult to remember that everyone else is not familiar
    • Inference is an easy way
  • Step back and enjoy the view
    • Carrying around reference numbers can help context
    • Good news tends to unfold slowly while bad news happens radiply
    • Reporting cadence can have a powerful effect on context
  • Get the backstory
    • Survivorship bias pervades… No one is interested is the expected
    • This is especially true with replication studies
    • Be wary of hypothesis made after the study is started
  • Ask who is missing
    • Females are almost always missing from studies
    • Almost every audience (especially “found” data) has response bias
    • “n=all” usually really means “n=all the people who do or have a thing
  • Demand transparency when the computer says “No”
    • Look for algorithm’s explanatory power or else they may be associating “winter trends” with flu cases
    • Algorithms are built by human and so can easily have bias built-in (hiring algorithms trained to hire men)
    • Transparency is what killed alchemy and brought science
    • We should question which algorithms we can trust
  • Don’t take statistical bedrock for granted
    • Basic understandings are important for building senses of the world
  • Remember that misinformation can be beautiful, too
    • Graphics lend authenticity, the more beautiful the more believable
    • Florence Nightingale was a statistician
  • Keep an open mind
    • We naturally fill in data blanks
    • “Things are going badly, so do something different”
    • Making public commitments makes it harder to take back
    • When information changes, change your conclusions
  • Be curious
    • Confirmation bias is real and powerful
    • Scientific curiosity is key (not scientific literacy)
    • Asking people to rate their knowledge, then explain how something works, can reduce stubbornness
    • “Please explain that” can do a lot to soften extreme stances

Comments

Leave a Reply