Biased Programmers? Or Biased Training Data? A Field Experiment about Algorithmic Bias

How to cite

Dell'Acqua, Fabrizio; Cowgill, Bo. 2019. "Biased Programmers? Or Biased Training Data? A Field Experiment about Algorithmic Bias." AEA RCT Registry. AEARCTR-0003574. https://doi.org/10.1257/rct.3574-1.1

Download: BibTeX · RIS — for reference managers without a browser translator. Zotero and most others pick up this page's citation metadata directly.

Fabrizio Dell'Acqua* , Bo Cowgill

* Primary investigator

Registry ID
AEARCTR-0003574
DOI
https://doi.org/10.1257/rct.3574-1.1
First registered
Last updated
Status
On going
Trial dates
2019-05-03 – 2020-01-20
Countries
Private
JEL codes
J24, J15, J16, O38, D80, M50
Registry record
http://www.socialscienceregistry.org/trials/3574

Abstract

Why does “algorithmic bias” occur? The two most frequently cited reasons are “biased programmers” and “biased training data.” We quantify the effects of these using a field experiment.

Intervention

Participants will be assigned to one of four conditions, with either biased or unbiased data. We think of the biased data (with no other interventions) as being the control group.

Experimental design

We will recruit a diverse set of programmers into a data science contest. As participants register, they will receive a training dataset to utilize in predicting an outcome. The outcome will be related to an event that historically exhibited bias against a particular group.

Design details

Randomization method
The randomization will be done in office by a computer
Randomization unit
Individual programmers
Clusters
464 programmers
Observations
464 programmers
Arms
264 programmers in the ML class, equally divided between the four treatments.
Primary outcomes
- The performance of the algorithms developed for the data science competition. Performance will be evaluated based on accuracy of prediction (for both groups).

Cited by

No citation data available for this registration yet.

Keywords