Page 55 - FIGI - Big data, machine learning, consumer protection and privacy

Should Learn About Machine Learning, Univ. of CA, Davis Law Review, 2017, available at https:// lawreview .law .ucdavis
.edu/ issues/ 51/ 2/ Symposium/ 51 -2 _Lehr _Ohm .pdf.
128 http:// patft .uspto .gov/ netacgi/ nph -Parser ?Sect1 = PTO1 & Sect2 = HITOFF & d = PALL & p = 1 & u = %2Fnetahtml %2FPTO
%2Fsrchnum .htm & r = 1 & f = G & l = 50 & s1 = 9100400 .PN . & OS = PN/ 9100400 & RS = PN/ 9100400
129 Jonathan Zim, The Use of Social Data Raises Issues for Consumer Lending, Miami Business Law Review, https:// business
-law -review .law .miami .edu/ social -data -raises -issues -consumer -lending/ .

130 Virginia Eubanks, Automating Inequality, St Martin’s Press (2018).
131 Hardt, Moritz, Price, Eric, and Srebro, Nathan. Equality of opportunity in supervised learning, NIPS, 2017; Chouldechova,
Alexandra, Fair prediction with disparate impact: A study of bias in recidivism prediction instruments, Corr, 2017.
132 Disparate impact has been defined using the “80% rule” such that, where a dataset has protected attribute X (e.g., race,
sex, religion, etc.) and a binary outcome to be predicted C (e.g., “will hire”), the dataset has disparate impact if:

for positive outcome class YES and majority protected attribute 1 where Pr(C = cjX = x) denotes the conditional
probability (evaluated over D) that the class outcome is c 2 C given protected attribute x 2 X. Feldman, Michael,
Friedler, Sorelle A., Moeller, John, Scheidegger, Carlos, and Venkatasubramanian, Suresh, Certifying and removing
disparate impact. In KDD, 2015. http:// sorelle .friedler .net/ papers/ kdd _disparate _impact .pdf.
133 Supreme Court of the United States. Griggs v. Duke Power Co. 401 U.S. 424, March 8, 1971.
134 The US Supreme Court found that Duke Power’s hiring decision was illegal if it resulted in “disparate impact” by
race even though it was not explicitly determined based on race. This prevented Duke Power from using intelligence
test scores and high school diplomas, qualifications largely correlated with race, to make hiring decisions. The legal
doctrine of disparate impact that was developed from this ruling is the main legal theory used to determine unintended
discrimination in the USA. Duke Power was unable to prove that the intelligence tests or diploma requirements were
relevant to the jobs for which they were hiring.
135 Texas Dep't of Housing and Community Affairs v. Inclusive Communities Project 135 S. Ct. 2507 (2015)
136 Solon Barocas and Andrew D. Selbst, Big Data’s Disparate Impact, 104 Calif. L. Rev. 671 (2016). http:// www
.californialawreview .org/ wp -content/ uploads/ 2016/ 06/ 2Barocas -Selbst .pdf.
137 Steve Lohr, Big Data Underwriting for Payday Loans, NY Times, January 19, 2015, https:// bits .blogs .nytimes .com/ 2015/
01/ 19/ big -data -underwriting -for -payday -loans/ .
138 See Accountable Algorithms, at footnote 129.
Paul Ohm and David Lehr, Playing with the Data: What Legal Scholars Should Learn About Machine Learning, Univ. of
CA, Davis Law Review, 2017, available at https:// lawreview .law .ucdavis .edu/ issues/ 51/ 2/ Symposium/ 51 -2 _Lehr _Ohm .pdf.
139 See Accountable Algorithms, at footnote 129.
Paul Ohm and David Lehr, Playing with the Data: What Legal Scholars Should Learn About Machine Learning, Univ. of
CA, Davis Law Review, 2017, available at https:// lawreview .law .ucdavis .edu/ issues/ 51/ 2/ Symposium/ 51 -2 _Lehr _Ohm .pdf.
140 Danielle Keats Citron, Technological Due Process, Washington University Law Review, Vol. 85, pp. 1249-1313, 2007,
https:// papers .ssrn .com/ sol3/ papers .cfm ?abstract _id = 1012360
141 Julia Angwin and Jeff Larson, The Tiger Mom Tax: Asians Are Nearly Twice as Likely to Get a Higher Price from
Princeton Review, ProPublica, Sept. 1, 2015.
142 See Karen Harris, Austin Kimson, and Andrew Schwedel, Labor 2030: The Collision of Demographics, Automation and
Inequality, Bain & Company Report, February 7, 2018 available at http:// www .bain .com/ publications/ articles/ labor -2030
-the -collision -of -demographics -automation -and -inequality .aspx.

143 See, e.g., Paul Ohm, Broken Promises of Privacy: Responding to the Surprising Failure of Anonymization, 57 UCLA L.
REV. 1701, 1716-27 (2010).
144 See FPF's Visual Guide to Practical Data.
145 See Cavoukian, Ann and El-Emam, Khaled, De-Identification Protocols: Essential for Protecting Privacy,
Information and Privacy Commissioner of Ontario, 2014; and Information Privacy Commissioner of Ontario, “De-
Identification Centre”, Information Privacy Commissioner of Ontario (https:// www .ipc .on .ca/ privacy/ de -identification
-centre/ ).

146 GDPR, Article 4(5).
147 Narayanan A, Felten EW. (Princeton). No silver bullet: De-identification still doesn't work 2014. Available at: http://
randomwalker .info/ publications/ no -silver -bullet -de -identification .pdf.

Big data, machine learning, consumer protection and privacy 53