algorithmic accountability: computational journalism week 11
DESCRIPTION
Columbia University, Fall 2014 Syllabus at http://www.compjournalism.com/?p=113TRANSCRIPT
Fron%ers of Computa%onal Journalism
Columbia Journalism School
Week 11: Algorithmic Accountability November 21, 2014
Algorithms in our lives
• Personalized search • Poli%cal microtarge%ng • Credit score / loans / insurance • Predic%ve policing • Price discrimina%on • Algorithmic trading / markets • Terrorist threat predic%on • Hiring models
Title VII of Civil Rights Act, 1964 It shall be an unlawful employment prac%ce for an employer -‐ (1) to fail or refuse to hire or to discharge any individual, or otherwise to discriminate against any individual with respect to his compensa%on, terms, condi%ons, or privileges of employment, because of such individual’s race, color, religion, sex, or na%onal origin; or (2) to limit, segregate, or classify his employees or applicants for employment in any way which would deprive or tend to deprive any individual of employment opportuni%es or otherwise adversely affect his status as an employee, because of such individual’s race, color, religion, sex, or na%onal origin.
Learning from Facebook likes
From Kosinski et. al., Private traits and a4ributes are predictable from digital records of human behavior
Predic%ng gender from Twi_er
From Zamal et. al., Homophily and Latent A4ribute Inference: Inferring Latent A4ributes of Twi4er Users from Neighbors
Predic%ng race from Twi_er
From Pennacchioa and Popescu, A Machine Learning Approach to Twi4er User ClassificaFon
Even if two groups of the popula%on admit simple classifiers, the whole popula%on may not(from How Big Data is Unfair)
Websites Vary Prices, Deals Based on Users' InformaFon Valen%no-‐Devries, Singer-‐Vine and Soltani, WSJ, 2012