Upgrade to Pro — share decks privately, control downloads, hide ads and more …

Escaping the Cult of Prediction (StrataConf NYC 2019)

Escaping the Cult of Prediction (StrataConf NYC 2019)

We’re living in a cultural moment is obsessed with making predictions. In politics and in business, we’re constantly coming up with ways to collect more data for a singular purpose: to predict what will happen next.

This overwhelming desire for prescience shapes the way we design, measure, and understand everything from products and marketing to politics and movements. Good predictions demand both precision and accuracy. Farrah Bostic walks you through how, in the quest to get more and more granular about how people will behave in the future, in the hopes that we can anticipate or manipulate that behavior, businesses are often tempted to rely on emerging or untested technologies—and sometimes pseudoscience—to get the “data” that fuels those predictions.

While this moment seems to be particularly defined by prediction, the practice goes back to (at least) the first lie detectors and has come to encompass practices like hypnosis, technology like medical imaging, and encoded anthropological approaches like microexpressions. But the implications are worse than wasting money and time. Businesses and brands are sacrificing the opportunity to understand things deeply and are simultaneously creating social negative externalities, like normalizing surveillance and misinformation, undermining public trust and values, and dehumanizing the very people whose behavior we want to predict.

The Difference Engine

September 25, 2019
Tweet

More Decks by The Difference Engine

Other Decks in Business

Transcript

  1. UNDERSTANDING 

    ESCAPING THE CULT OF PREDICTION
    25 September 2019, StrataConf NYC #stratadata
    Farrah Bostic, The Difference Engine

    View Slide

  2. View Slide

  3. OK, LET’S
    GET STARTED

    View Slide

  4. @farrahbostic
    Head of Strategy & Research, Founder, 

    The Difference Engine
    20 year in advertising & market research
    JD, Benjamin N Cardozo School of Law
    BA, Journalism & Communications, 

    University of Oregon

    View Slide

  5. 1992: 

    MR. MORTON’S ADVANCED ENGLISH
    ALL* YOU HAVE TO DO IN LIFE 

    IS MAKE CHOICES.
    *AND DIE, EVENTUALLY.

    View Slide

  6. 1995: 

    PS 492 DECISION MAKING 

    WITH PROF. JOHN ORBELL
    “[Hamlet’s] problem is making a judgment 

    about the facts surrounding a choice he must make, 

    doing so when some judgment cannot be avoided, 

    when the consequences of error are unthinkable, 

    and when the odds one way or the other are unknown.”

    View Slide

  7. 1999: 

    BELLAGIO HOTEL & CASINO
    “I follow the system. 

    Sometimes I win, sometimes I lose. 

    I leave half my winnings in the casino bank. 

    That’s how I started my business.”

    View Slide

  8. 2002: 

    FIRST YEAR OF LAW SCHOOL
    1066, the Norman Invasion, 

    “the King’s Peace” & the Rule of Law

    View Slide

  9. 2006: 

    MEETING WITH MY VODKA CLIENT
    “Which one of these campaigns will actually work?”

    View Slide

  10. The Difference
    Engine

    View Slide

  11. HOW DO 

    WE MAKE
    DECISIONS?

    View Slide

  12. IT’D BE NICE IF DECISION-MAKING WAS THIS EASY
    Profit!

    View Slide

  13. THERE ARE TWO MAIN KINDS OF DECISION-MAKING
    Outcome-oriented decisions: 

    Good outcomes = 

    good decisions
    Process-oriented decisions:

    Good processes = 

    good decisions

    View Slide

  14. FOR EXAMPLE, THE RULE OF LAW
    Formal characteristics: generality, publicity,
    prospectivity, intelligibility, consistency,
    practicability, congruence, stability
    Procedural characteristics: impartial &
    independent tribunal, right of
    representation, right to be present &
    participate in your defense, right to know
    why a judgment was made
    Values: a bond of reciprocity and a mutuality
    of constraint between ruler & ruled,
    predictability & reliability, liberty, dignity

    View Slide

  15. BUT OUTCOMES ARE WHAT MATTERS, RIGHT?
    If the focus of your decision-making is achievement - getting a particular outcome -
    then you have a lot riding on your ability to predict whether that outcome will occur.
    We should probably think about this probabilistically. But we’re not very good at that.
    0% 100%
    v.

    View Slide

  16. Known Knowns Known Unknowns
    Unknown Knowns Unknown Unknowns

    View Slide

  17. MAKING PREDICTIONS INSTEAD
    OF MAKING DECISIONS

    View Slide

  18. HOW DO WE MAKE PREDICTIONS?
    If the prediction comes
    true, it’s a good prediction.

    View Slide

  19. PREDICTIONS DON’T
    LIVE IN A VACUUM
    PEOPLE TEND TO ACT IN RELIANCE ON THEM.

    View Slide

  20. SOME FOLKS HAVE TO MAKE PREDICTIONS, THOUGH.
    Being reliably precise & accurate is the key to credibility.

    View Slide

  21. PREDICTIONS =
    HYPOTHESES?
    NOPE.
    A GOOD PREDICTION IS RELIABLE, ACCURATE & PRECISE.
    A GOOD HYPOTHESIS IS DISPROVABLE.

    View Slide

  22. HOW DO WE FORM HYPOTHESES?
    Hypotheses are explanations
    of observed phenomena
    that can be disproved.

    View Slide

  23. HYPOTHESES > PREDICTIONS
    Outcome-oriented decisions: 

    Good outcomes = 

    good decisions
    Process-oriented decisions:

    Good processes = 

    good decisions
    PREDICTIONS HYPOTHESES

    View Slide

  24. PREDICTIONS TRAFFIC IN CERTAINTY.
    PEOPLE WOULD RATHER BE 

    CERTAINLY WRONG 

    THAN A LITTLE UNCERTAIN.

    View Slide

  25. USING ANALYTICS 

    AS A DRUNK 

    USES A LAMP POST
    FOR SUPPORT, RATHER THAN ILLUMINATION.

    View Slide

  26. PEOPLE LIKE 

    MAKING PREDICTIONS 

    MORE THAN THEY LIKE 

    MAKING DECISIONS

    View Slide

  27. MAKING DECISIONS IS HARD AND
    HIGH STAKES.
    YOU’RE NOT EXPLAINING OR
    PREDICTING ANYMORE.
    YOU’RE CHOOSING.
    SO NOW IT'S ON YOU.

    View Slide

  28. SATISFICING & RESULTING
    We have “decision defaults” - the
    choice that’s good enough, what we’d
    do if we had no additional information,
    and what we'd do if we had quite a bit
    (but not “enough”) new information.
    Cassie Kozyrkov
    Chief Decision Scientist, 

    Google
    We equate the quality of a decision
    with the quality of its outcome, and
    succumb to the temptation to change
    our strategy just because it didn’t pay
    off immediately.
    Annie Duke
    Poker Player & 

    Decision Strategist

    View Slide

  29. WE ATTRIBUTE GOOD OUTCOMES
    TO GOOD DECISIONS.
    AND BAD OUTCOMES TO 

    “POOR RISK MANAGEMENT”.
    AND WE “MANAGE RISK” BY COLLECTING DATA, TO
    SUBSTITUTE PREDICTION FOR PROCESS, CERTAINTY
    FOR STRATEGY.

    View Slide

  30. THIS IS MOTIVATED
    REASONING.
    WE PAY ATTENTION TO CONFIRMING EVIDENCE AND
    DISCREDIT DISCONFIRMING EVIDENCE.

    View Slide

  31. BY THE WAY, PEOPLE TEND TO BELIEVE
    THAT QUALITATIVE SOURCES PRESENT A
    MUCH GREATER RISK OF PRODUCING
    DISCONFIRMING EVIDENCE.
    BUT THEY DON’T.

    View Slide

  32. WE ALSO THINK 

    PEOPLE ARE LIARS.
    WE’RE NOT, USUALLY.

    View Slide

  33. IF YOU GENUINELY BELIEVE THAT
    MOST PEOPLE ARE LIARS, WELL,
    FIRST OF ALL…
    IS EVERYTHING OKAY?

    View Slide

  34. SECOND, 

    ‘SHALLOW’ PSYCHOLOGY TELLS US:
    “THERE IS SOME GOOD REASON
    FOR MOST THINGS PEOPLE DO”

    View Slide

  35. SOME PROBLEMS CAN BE SO HARD THAT
    ANYONE WILL BE DRIVEN TO DISTRACTION -
    IT’S NOT THEIR PSYCHE, IT’S THE SITUATION.

    View Slide

  36. THIRD,
    [AHEM]
    WHAT MAKES YOU SO SPECIAL?

    View Slide


  37. A decision maker confronts risk when he or she
    can attach probabilities to alternative states of
    the world with confidence… Under uncertainty,
    not only can one still lose but one does not
    know the odds.
    - Professor John Orbell, 

    University of Oregon

    View Slide

  38. IN THIS ERA, WE COPE WITH
    UNCERTAINTY BY DEMANDING
    MORE DATA.
    IT’S BECOMING AN ADDICTION.

    View Slide

  39. OOH! MYSTICISM…

    View Slide

  40. DATA ADDICTION 

    LEADS TO 

    PROBLEMATIC DATA
    COLLECTION

    View Slide

  41. TECH & SCIENCE HAVE BEEN
    USED TO SCAM PREDICTION
    ADDICTS FOR DECADES

    View Slide

  42. DATA CAN BE 

    FRAUDULENT.
    DATA CAN BE 

    UNETHICAL.

    View Slide

  43. ANYBODY USED THESE TOOLS?
    ➤ Lie detectors
    ➤ Body language experts
    ➤ Hypnotists
    ➤ EEG
    ➤ fMRI
    ➤ Hidden cameras
    ➤ Sensors
    ➤ Computer vision
    ➤ Sentiment analysis

    View Slide

  44. THE FOUR HORSEMEN
    STALKERS
    LIE DETECTORS
    MIND READERS
    MENTALISTS

    View Slide

  45. STALKERS

    View Slide

  46. STALKERS AREN'T JUST
    FOLLOWING YOU, THEY'RE
    JUDGING & MANIPULATING YOU

    View Slide

  47. LIE DETECTORS

    View Slide

  48. MIND-READERS

    View Slide

  49. MENTALISTS

    View Slide

  50. SO WHAT DO
    WE DO?

    View Slide

  51. HAVE YOU TRIED
    ENGAGING WITH
    HUMANS?

    View Slide

  52. COLLECT DATA
    ETHICALLY

    View Slide

  53. INTERROGATE SOURCES AND METHODS
    ➤ Is the technology peer reviewed?
    ➤ Has the vendor published its results?
    ➤ Are the results replicable?
    ➤ Is the method in keeping with your values?

    View Slide

  54. INTERROGATE THE NATURE OF THE DATA
    ➤ What do they claim the data will tell you? 

    Reducing uncertainty is great, removing risk is a lie.
    ➤ Do people understand that this data is being collected and did they opt in? 

    Would you opt in if it were collecting on you?
    ➤ Who else wants this data, and why?

    Are you comfortable with that?
    ➤ What are the unintended consequences of collecting this data?

    Who could get hurt?

    View Slide

  55. MAKE DECISIONS
    INTENTIONALLY

    View Slide

  56. what do you
    w
    ant to do?
    what’s the
    desired outcome?
    what are your
    constraints?
    What are your
    v
    alues?
    what are your
    constraints?
    What is a good
    process?
    what dat
    a is
    necessary?
    act.
    identify options.
    what sources
    are best?
    what will you
    do with it?
    strategy process dat
    a decide

    View Slide

  57. THANK YOU!
    LET’S KEEP IN TOUCH.
    @FARRAHBOSTIC
    [email protected]

    View Slide