[Congressional Record Volume 165, Number 62 (Wednesday, April 10, 2019)]
[Senate]
[Page S2389]
From the Congressional Record Online through the Government Publishing Office [www.gpo.gov]
By Mr. WYDEN (for himself and Mr. Booker):
S. 1108. A bill to direct the Federal Trade Commission to require
entities that use, store, or share personal information to conduct
automated decision system impact assessments and data protection impact
assessments; to the Committee on Commerce, Science, and Transportation.
Mr. WYDEN. Mr. President, today I, along with my colleague Senator
Booker of New Jersey, are introducing the Algorithmic Accountability
Act. This bill is a critical first step to address the use of biased or
discriminatory algorithmic decisions impacting American consumers. It
is a bicameral effort, led in the House of Representatives by
Congresswoman Yvette Clarke, vice chair of the Energy and Commerce
Committee.
Today's biggest companies are increasingly using algorithms to make
decisions about consumers. The use by companies of algorithms can often
benefit consumers, and these technologies have been critical in the
creation of thousands of American companies. Alongside this beneficial
proliferation, algorithms have become entrenched in the most life-
changing of decisions. Algorithms can now determine whether Americans
are hired for a dream job, are approved for a home mortgage, or even
sent to jail.
But, as history has shown, a win for the corporation is not always a
win for the consumer. And, in this case, when consumers lose, all too
often they are a woman or an American of color. Though an innovation
critical for future growth, algorithms can be as discriminatory as the
humans they have begun to replace.
The issue is a simple one: While algorithms come to conclusions based
on calculations, these calculations are created by humans or use data
collected and supplied by humans. And, unfortunately, we humans can be
biased, whether we know it or not, or we can created algorithms that,
in time, create biases of their own.
Thanks to a flood of news reports and investigations detailing
algorithms-gone-wrong, these issues are coming to light. Yet, American
companies and the U.S. government are doing far too little to assess
whether their own algorithms depend on biased assumptions, have created
biases assumptions, and have the effect of increasing discrimination in
the U.S.
Senator Booker and I intend to change that by ensuring that today's
racial, social, and gender biases do not become entrenched in the
automation of tomorrow.
Our bill has four main components.
First, it authorizes the Federal Trade Commission to create
regulations requiring companies under its jurisdiction to conduct
impact assessments of highly sensitive algorithms. This requirement
would apply not only to new algorithmic systems, but also those that
are both new and already in existence.
Second, it requires companies to assess their use of algorithms--
including any relevant training data--for impacts on accuracy,
fairness, bias, discrimination, privacy, and security.
Third, it requires companies to evaluate how their information
systems protect the privacy and security of consumers' personal
information.
And, finally, it requires companies to correct any issues they
discover during the impact assessments.
This legislation is in no way intended to hinder the adoption by
American companies of advanced technologies like algorithms. Automated
decision systems are out there, and they are being adopted into
commercial decision-making processes.
What we are seeking to do with this bill is to ensure that companies
take a hard look at their own technologies to ensure that they address
any unintended side effects.
Mr. President, it is time for Congress to get involved by requiring
companies to address biases and unintended discriminatory effects in
their automated decision systems.
I thank my colleague Senator Booker for his efforts on this bill, and
I hope the Senate will promptly consider and pass this critical
legislation.
______