Dartmouth logo Dartmouth College Computer Science
Technical Report series
CS home
TR home
TR search TR listserv
By author: A B C D E F G H I J K L M N O P Q R S T U V W X Y Z
By number: 2017, 2016, 2015, 2014, 2013, 2012, 2011, 2010, 2009, 2008, 2007, 2006, 2005, 2004, 2003, 2002, 2001, 2000, 1999, 1998, 1997, 1996, 1995, 1994, 1993, 1992, 1991, 1990, 1989, 1988, 1987, 1986

Accuracy and Racial Biases of Recidivism Prediction Instruments
Julia J. Dressel
Dartmouth TR2017-822

Abstract: Algorithms have recently become prevalent in the criminal justice system. Tools known as recidivism prediction instruments (RPIs) are being used all over the country to assess the likelihood that a criminal defendant will reoffend at some point in the future. In June of 2016, researchers at ProPublica published an analysis claiming an RPI called COMPAS was biased against black defendants. This claim sparked a nation-wide debate as to how fairness of an algorithm should be measured, and exposed the many ways that algorithms are not necessarily fair. Algorithms are used in the criminal justice system because they are regarded as more accurate and less biased than human predictions; however, there does not exist a contemporary comparison of the performance of human and algorithmic recidivism predictions. To address this, we set out to determine if COMPAS is more accurate than human prediction, and to identify how the racial biases of human recidivism predictions compare to the racial biases of the COMPAS algorithm. After establishing a baseline performance of human prediction, we explore whether incorporating human judgment into algorithms can enhance prediction accuracy.

Note: Senior Honors Thesis. Advisor: Hany Farid.

PDF PDF (2252KB)

Bibliographic citation for this report: [plain text] [BIB] [BibTeX] [Refer]

Or copy and paste:
   Julia J. Dressel, "Accuracy and Racial Biases of Recidivism Prediction Instruments." Dartmouth Computer Science Technical Report TR2017-822, June 2017.

Notify me about new tech reports.

Search the technical reports.

To receive paper copy of a report, by mail, send your address and the TR number to reports AT cs.dartmouth.edu

Copyright notice: The documents contained in this server are included by the contributing authors as a means to ensure timely dissemination of scholarly and technical work on a non-commercial basis. Copyright and all rights therein are maintained by the authors or by other copyright holders, notwithstanding that they have offered their works here electronically. It is understood that all persons copying this information will adhere to the terms and constraints invoked by each author's copyright. These works may not be reposted without the explicit permission of the copyright holder.

Technical reports collection maintained by David Kotz.