Classification and its Discontents


Scott H. Hawley, Ph.D.
Professor of Physics
Belmont University


  • is a fundamental process in science, philsophy, sociology, medicine, law, industry,...and more.
  • simplifies the intractable complexity of the world around us
    • by grouping similar things together
    • drawing boundaries to demarcate territories of interest
    • to help us decide which policy is applicable

Classification is...

  • Infrastructure
  • Policy
  • Politics
  • Power
  • the lifeblood of society
  • increasingly being automated.

Classification Words

clustering - grouping similar things
taxonomy - setting up classes/categories/bins
labeling - what to name the classes
assignment - to which class does a new thing belong?
threshold - boundary/criterion between classes

sort, decide, choose, file, discern, judge, codify, name, peg, divide, segregate, arrange, grade, count as, identify, typecast, distinguish, pigeonhole

Classification in different Fields


Linneus' taxonomy
Haeckel's Tree of Life:
Guides, e.g.
Bible: Is this food kosher?
What is this thing

Library Science

Organizing recorded knowledge for the purpose of retrieval
Subject headings, Keywords
...can be politically charged.


Diagnosis: which disease is it?
Is it cancer? (Whitney et al., 2019)

IEEE "AI vs. Doctors"


Was the contract boken?
What are appropriate consequences?
Criminal justice: What risk level is this defendant?


"Genre" is

Cover song identification
(e.g. YouTube)

Image: Chris Traile (Duke U)

Why is a physicist interested?

  • Compared to other areas of science, physicists do much more regression* than classification.
    *fitting a curve to data points
  • But I got involved in..
    • Machine Learning (ML)
    • "Bridging the Two Cultures II" (B2C2) grant (Struthers,Bishop): "Christian Responses to AI"
      • Questions of aesthetics, ethics, policy: "How would one automate this?"
    • "AI Ethics" often involves societal impact due to classifications

Binary Classification

Binary Classification in Song

Binary classification is a common & important human activity. Questions of identity, purpose, even reality itself...

Choice AChoice BSongArtist/Band
Is this the real lifeIs this just fantasyBohemian RhapsodyQueen
Are we humanOr are we dancerHumanThe Killers
Can't tell if you're a boyOr a girlRebel RebelDavid Bowie
Is you isOr is you ain't my babyIs You Is or Is You Ain't My BabyLouis Jordan
Was it something I saidOr something I didEvery Rose Has its ThornPoison
Should I stayOr should I goShould I Stay or Should I GoThe Clash

Binary classification example

Machine Learning
is Moving In

Math/CS Methods

(Press Reload if interactive graphs don't appear)

Logistic Regression

For binary classification

Feature selection

Can you divide these?

Feature Engineering

Nearest Neighbor
Object Detection Vid:

Labeling Issues

Labeling Issues

In ImageNet dataset:

Kate Crawford and Trevor Paglen, “Excavating AI: The Politics of Training Sets for Machine Learning (Sept. 19, 2019)

Classifiying people

  • Nazis' use of Hollerith machines
  • Phrenology (China: updated with ML)
  • Facebook quizzes -> Targeted Ads -> "The Great Hack"
Much work is needed: e.g., ACM Conference on Fairness, Accountability, and Transparency (FAT)

Classifications as Interventions

Regression Discontinuity Design (RDD)

When you have a threshold or date at which a classification is implemented.

RDD Illustrated

Hawley's Conjecture

Release Rates

(Stevenson, 2017)

Schooling in Kenya

(Ozier, 2018)

Recent Headlines

Content Moderation: "Pinterest says AI reduced reported self-harm content by 88%" (Oct 10, 2019)

Classification and its Discontinuities

Scott H. Hawley, @drscotthawley
Belmont University
This work was supported Sponsored by a grant given by Bridging the Two Cultures of Science and the Humanities II, a project run by Scholarship and Christianity in Oxford (SCIO), the UK subsidiary of the Council for Christian Colleges and Universities, with funding by Templeton Religion Trust and the Blankemeyer Foundation.