Machine Learning Newsletter

Short Notes on Thinking with Data

Introduction

  • Arguments are how we convince ourselves that something is true.
  • Argument theory provides a useful set of mental models for understanding arguments.
  • Data work is held together by arguments.

Data

  • Data are just (sadly) observations.
  • What we want is knowledge and representation, maybe even understanding and insight.
  • By doing so, arguments move from what is unknown/not agreed upon yet.

Arguments are how we turn observations into knowledge.

Arguments

  • Claim => Audience does not believe it(yet).
  • Prior Knowledge => Things you/your audience believes already before the case is made.
  • Evidence => Where data enters into an argument. It becomes part of the argument, not just "data" anymore. It gains context in that sense.
  • Justification => Reasoning why evidence should cause vs to believe the claim.
  • Rebuttal => Any of the reasons why the justification might not hold in this particular case.

Patterns of Arguments

Template for Claims

  • Categories of Dispute
  • Causal Analysis

Causal Analysis

  • Large set of patterns are necessary.
  • Goal is to claim to have reasonably accounted for alternative explanations.
  • People will interpret our work causally, we should try our best to make it reasonable to do so.

Conclusions

  • Mental models make our lives and our work better.
  • Arguments are a deep part of working with data.
comments powered by Disqus