CS Table 4/25: Algorithmic Accountability

petersos — Tue, 25 Apr 2017 12:55:36 +0000

Algorithms are essential to computer science, and increasingly they are essential to modern decision making on all levels. But are they unbiased? The emerging field of ‘Algorithm Accountability’ is beginning to identify cases in which inherent bias is imbedded in the inference structure of algorithms. The articles for this week’s CS Table run the gamut from popular press to general audience special interest to technical position papers to ACM recommendations for basic principles. The question is no longer ‘are algorithms biased’, but how to determine whether they are, and if so how to prevent them from being so.

Linda Oyolu, Ruth Wu, and Ursula Wolz will be leading our discussion on April 25, 2017. The following readings will give you a good sense of the area; please do your best to read at least some subset of these articles before our CS Table discussion:

This NY Times overview is a good place to start: Make Algorithms Accountable (01-Aug-2016)
An MIT Press summary of the emerging field: How to Hold Algorithms Accountable (17-Nov-2016)
An ACM statement about principles for creating accountable algorithms: (PDF) Statement on Algorithmic Transparency and Accountability (12-Jan-2017)
One of the seminal articles - a good place to get into the serious study: (PDF) Algorithmic Accountability, Journalistic investigation of computational power structures (07-Nov-2016)

Computer science table (CS Table) is a weekly meeting of Grinnell College community members (students, faculty, staff, etc.) interested in discussing topics related to computing and computer science. CS Table meets Tuesdays from 12:00-1:00pm in JRC 224B. Contact the CS faculty for the weekly reading. Students on meal plans, faculty, and staff are expected to cover the cost of their meals. Visitors to the College and students not on meal plans can charge their meals to the department.

CS Table 11/15: Algorithmic Bias

petersos — Mon, 14 Nov 2016 22:14:58 +0000

As computation plays a larger role in society, we are beginning to see cases where algorithms encode the biases of the past. While this phenomenon is widespread, there is a particularly interesting case where computer programs make bail recommendations for suspects in criminal cases.

There are two short readings suggested for the discussion on November 15:

Battling Algorithmic Bias. Keith Kirkbatrick, Communications of the ACM.
When Algorithms Discriminate. Claire Cain Miller, The New York Times.
Printed readings are available at Noyce 3827.

Computer Science - bias

CS Table 4/25: Algorithmic Accountability

CS Table 11/15: Algorithmic Bias