[Congressional Bills 116th Congress]
[From the U.S. Government Publishing Office]
[S. 5052 Introduced in Senate (IS)]

<DOC>






116th CONGRESS
  2d Session
                                S. 5052

   To increase fairness and transparency in algorithmic eligibility 
                            determinations.


_______________________________________________________________________


                   IN THE SENATE OF THE UNITED STATES

                           December 17, 2020

   Mr. Coons introduced the following bill; which was read twice and 
   referred to the Committee on Commerce, Science, and Transportation

_______________________________________________________________________

                                 A BILL


 
   To increase fairness and transparency in algorithmic eligibility 
                            determinations.

    Be it enacted by the Senate and House of Representatives of the 
United States of America in Congress assembled,

SECTION 1. SHORT TITLE; TABLE OF CONTENTS.

    (a) Short Title.--This Act may be cited as the ``Algorithmic 
Fairness Act of 2020''.
    (b) Table of Contents.--The table of contents of this Act is as 
follows:

Sec. 1. Short title; table of contents.
Sec. 2. Definitions.
Sec. 3. FTC study and report on ways companies are developing and 
                            implementing algorithmic eligibility 
                            determinations.
Sec. 4. Substantive fairness.
Sec. 5. Procedural fairness.
Sec. 6. Enforcement.
Sec. 7. Research grants for the study of fair and transparent data 
                            analytics.
Sec. 8. Leadership program.
Sec. 9. Avoiding duplication.

SEC. 2. DEFINITIONS.

    In this Act:
            (1) Algorithmic eligibility determination.--The term 
        ``algorithmic eligibility determination'' means a determination 
        based in whole or in significant part on an algorithmic process 
        that utilizes methods of machine learning, advanced statistical 
        techniques, artificial intelligence, or similar techniques to 
        determine the eligibility for, the denial of access to, the 
        receipt of information about, the cost of acquiring, the 
        opportunity to access, or the revocation of important 
        opportunities including, but not limited to, education, 
        employment, credit, health care insurance, and housing.
            (2) Commission.--The term ``Commission'' means the Federal 
        Trade Commission.
            (3) Covered entity.--The term ``covered entity'' means an 
        entity that--
                    (A) makes an algorithmic eligibility determination; 
                and
                    (B) is--
                            (i) a person over which the Commission has 
                        authority pursuant to section 5(a)(2) of the 
                        Federal Trade Commission Act (15 U.S.C. 
                        45(a)(2)), but including banks, savings and 
                        loan institutions, and Federal credit unions 
                        that are otherwise excluded under such section;
                            (ii) a common carrier subject to the 
                        Communications Act of 1934 (47 U.S.C. 151 et 
                        seq.), notwithstanding the definition of the 
                        term ``Acts to regulate commerce'' in section 4 
                        of the Federal Trade Commission Act (15 U.S.C. 
                        44) and the exception provided by section 
                        5(a)(2) of the Federal Trade Commission Act (15 
                        U.S.C. 45(a)(2)) for such carriers; or
                            (iii) a nonprofit organization, including 
                        any organization described in section 501(c) of 
                        the Internal Revenue Code of 1986 that is 
                        exempt from taxation under section 501(a) of 
                        such Code, notwithstanding the definition of 
                        the term ``Acts to regulate commerce'' in 
                        section 4 of the Federal Trade Commission Act 
                        (15 U.S.C. 44) and the exception provided by 
                        section 5(a)(2) of the Federal Trade Commission 
                        Act (15 U.S.C. 45(a)(2)) for such 
                        organizations.
            (4) Online advertisement.--The term ``online 
        advertisement'' means an algorithmic eligibility determination 
        that determines whether an individual receives promotional 
        information from a covered entity through the use of paid 
        internet or paid digital communication.
            (5) Search result.--The term ``search result'' means an 
        algorithmic eligibility determination that determines the 
        information an individual is presented after querying an 
        internet search engine.

SEC. 3. FTC STUDY AND REPORT ON WAYS COMPANIES ARE DEVELOPING AND 
              IMPLEMENTING ALGORITHMIC ELIGIBILITY DETERMINATIONS.

    (a) Study.--The Commission shall conduct a study on the ways 
covered entities are developing and implementing algorithmic 
eligibility determinations. Such study shall include an analysis of--
            (1) the industries that most commonly utilize algorithmic 
        eligibility determinations;
            (2) the types of algorithmic eligibility determinations 
        that covered entities make;
            (3) the data, data sources, pre-processing strategies and 
        other practices, data collection technologies, and 
        methodologies that covered entities use to make such 
        determinations;
            (4) the transparency of algorithmic eligibility 
        determinations, including the types of explanations that 
        covered entities disclose to--
                    (A) individuals or entities that are the subject of 
                an algorithmic eligibility determination;
                    (B) regulators; and
                    (C) the general public;
            (5) whether and how human decision makers are involved in 
        rendering algorithmic eligibility determinations;
            (6) whether covered entities have a mechanism to receive 
        complaints about algorithmic eligibility determinations, and 
        the number and nature of such complaints received from 
        individuals regarding such determinations;
            (7) whether covered entities have accountability procedures 
        in place should an algorithmic eligibility determination be 
        questioned or challenged, a description of such accountability 
        procedures, and what remedial steps covered entities have 
        taken, if any;
            (8) what information a covered entity could reasonably be 
        required to disclose such that an individual could understand, 
        question, or challenge an algorithmic eligibility 
        determination;
            (9) whether covered entities perform algorithmic audits to 
        understand whether algorithmic eligibility determinations are 
        fair, how covered entities define fairness for this purpose, 
        and whether any such audit techniques and outcomes of audits 
        are available to individuals or groups challenging an 
        algorithmic eligibility determination;
            (10) the extent to which, if any, algorithmic eligibility 
        determinations consider factors such as geographic information, 
        income, ethnicity, race, religion, national origin, age, sex, 
        sexual orientation, disability information, pregnancy, marital 
        status, physical or mental health status, criminal history 
        status, or proxies for such factors;
            (11) the extent to which algorithmic eligibility 
        determinations could result in negative or differential 
        treatment of individuals based on the factors described in 
        paragraph (10); and
            (12) the best way to address intellectual property concerns 
        of covered entities, like trade secret protections, that may 
        arise if the Federal Government requires companies to share 
        information related to their algorithmic eligibility 
        determination processes.
    (b) Reports.--
            (1) Initial report.--Not later than 1 year after the date 
        of the enactment of this Act, and once every 3 years thereafter 
        (until a total of 4 reports have been submitted), the 
        Commission shall submit to Congress a report on the study 
        conducted under subsection (a), together with recommendations 
        for such additional legislation and administrative action as 
        the Commission determines appropriate.
            (2) Special rule.--In the final report submitted in 
        accordance with paragraph (1), the Commission shall include a 
        recommendation as to whether to continue conducting the study 
        under subsection (a) (and submitting corresponding reports 
        under this subsection).
    (c) Consultation.--In conducting the study under subsection (a) and 
preparing the reports under subsection (b), the Commission shall 
consult with--
            (1) the National Institute of Standards and Technology;
            (2) the Department of Homeland Security;
            (3) the Consumer Financial Protection Bureau;
            (4) the Department of Housing and Urban Development;
            (5) the Department of Health and Human Services;
            (6) the Department of Veterans Affairs;
            (7) the Department of Education;
            (8) the Federal Communications Commission;
            (9) the Equal Employment Opportunity Commission; and
            (10) the Civil Rights Division of the Department of 
        Justice.

SEC. 4. SUBSTANTIVE FAIRNESS.

    (a) In General.--A covered entity shall not act on an unfair 
algorithmic eligibility determination in or affecting commerce.
    (b) Considerations.--In determining whether a particular 
algorithmic eligibility determination is unfair, the Commission--
            (1) shall consider the factors specified in section 5(n) of 
        the Federal Trade Commission Act (15 U.S.C. 45(n)); and
            (2) may consider--
                    (A) established public policies;
                    (B) emotional distress, bias on the basis of 
                protected class status, and other noneconomic injuries, 
                and may conclude that these injuries contribute to 
                substantial injury to consumers for the purposes of 
                section 5(n) of the Federal Trade Commission Act (15 
                U.S.C. 45(n)); and
                    (C) any code of conduct developed by industry 
                sectors, civil rights groups, consumer protection 
                groups, or academics, as long as the Commission has 
                first published such code of conduct in the Federal 
                Register and provided for a period of public comment in 
                accordance with section 553 of title 5, United States 
                Code.

SEC. 5. PROCEDURAL FAIRNESS.

    (a) In General.--Beginning 1 year after the date of the enactment 
of this Act, each covered entity shall, for each algorithmic 
eligibility determination--
            (1) retain for at least 5 years an audit trail that 
        records--
                    (A) the data used by the algorithm, as it existed 
                at the time the algorithm was deployed and rendered the 
                algorithmic eligibility determination;
                    (B) the source of pre-processing techniques, or any 
                other technique used to produce any such data described 
                in subparagraph (A);
                    (C) the methodology used by the entity to develop 
                the algorithm;
                    (D) the version of the design of the algorithm 
                utilized to make the determination;
                    (E) any data or sets of data used to train the 
                algorithm;
                    (F) any testing for model performance for 
                discriminatory effects across different subgroups and 
                the results of such testing;
                    (G) the methodology used to render the 
                determination; and
                    (H) the ultimate determination rendered;
            (2) notify the individual that they have been the subject 
        of an algorithmic eligibility determination; and
            (3) upon request, provide the individual with the 
        opportunity to--
                    (A) access the data pertaining to that individual 
                that the covered entity employed to make the 
                determination, in a human-readable format that a 
                reasonable individual can understand;
                    (B) submit corrections to the data pertaining to 
                that individual that the covered entity used in the 
                algorithmic eligibility determination; and
                    (C) request that the covered entity conduct a 
                reevaluation of the relevant algorithmic eligibility 
                determination based on the corrected data.
    (b) Review.--Upon the request of the Commission, a covered entity 
shall make available to the Commission the full audit trail described 
in subsection (a)(1). The Commission shall establish a secure and 
confidential process for reviewing the materials created and retained 
under such subsection.
    (c) Request to the Commission.--An individual who has been the 
subject of an algorithmic eligibility determination may request that 
the Commission conduct a review (and the Commission may conduct such a 
review) of--
            (1) the algorithmic eligibility determination to determine 
        if it was unfair (determined pursuant to section (4)); and
            (2) a covered entity's reevaluation of an algorithmic 
        eligibility determination where an individual has submitted 
        corrected data.
    (d) Rule of Construction.--Nothing in this section shall be 
construed to require a covered entity to provide an individual with the 
algorithm or model used to make the eligibility determination, or 
otherwise divulge legally protected trade secrets.
    (e) Scope.--Notwithstanding any of the preceding provisions of this 
section, subsection (a)(2) shall not apply to--
            (1) online advertisements;
            (2) search results; or
            (3) any other use case for which the Commission determines, 
        by rulemaking in accordance with section 553 of title 5, United 
        States Code, that the requirements of subsection (a)(2) would 
        impose an undue burden on or prove to be infeasible for that 
        general class of use case.

SEC. 6. ENFORCEMENT.

    (a) Enforcement by the Federal Trade Commission.--
            (1) Unfair act or practice.--Any covered entity that 
        violates section 4 or section 5 shall be deemed to have 
        committed an unfair act or practice under section 5(a)(1) of 
        the Federal Trade Commission Act (15 U.S.C. 45(a)(1)).
            (2) Powers of the commission.--
                    (A) In general.--The Commission shall enforce 
                sections 4 and 5 in the same manner, by the same means, 
                and with the same jurisdiction, powers, and duties as 
                though all applicable terms and provisions of the 
                Federal Trade Commission Act (15 U.S.C. 41 et seq.) 
                were incorporated into and made a part of this section.
                    (B) Civil penalty authority.--When enforcing 
                section 4 or 5, the Commission may, depending on the 
                nature and severity of the violation, include an 
                assessment of a civil penalty in the cease and desist 
                order provided for under section 5(b) of the Federal 
                Trade Commission Act (15 U.S.C. 45).
                    (C) Privileges and immunities.--Any covered entity 
                that violates section 4 or 5 shall be subject to the 
                penalties and entitled to the privileges and immunities 
                provided in the Federal Trade Commission Act (15 U.S.C. 
                41 et seq.).
                    (D) Authority preserved.--Nothing in this 
                subsection shall be construed to limit the authority of 
                the Commission under any other provision of law.
    (b) Regulations.--
            (1) In general.--The Commission may promulgate, in 
        accordance with section 553 of title 5, United States Code, 
        regulations to--
                    (A) carry out section 4, including by describing 
                which algorithmic eligibility determinations are unfair 
                for the purposes of section 5 of the Federal Trade 
                Commission Act (15 U.S.C. 45); and
                    (B) carry out section 5.
            (2) No requirement for regulations.--Nothing in this Act 
        shall require the Commission to promulgate regulations before 
        enforcing violations of section 4 or 5.
    (c) Referral Regarding a Potential Violation of a Federal Anti-
Discrimination Law.--If the Commission finds that an algorithmic 
eligibility determination may be in violation of a Federal anti-
discrimination law, the Commission shall refer the matter to the 
appropriate Federal or State agency with authority to initiate 
proceedings relating to such violation.

SEC. 7. RESEARCH GRANTS FOR THE STUDY OF FAIR AND TRANSPARENT DATA 
              ANALYTICS.

    (a) Grant Authority.--The Director of the National Institute of 
Standards and Technology may award grants for research into fair, 
accountable, and transparent data analytics and machine learning, 
particularly with respect to algorithmic eligibility determinations.
    (b) Funding.--There are authorized to be appropriated $10,000,000 
for each of fiscal years 2021 through 2026, to carry out this section. 
Amounts appropriated under the preceding sentence shall remain 
available until expended.

SEC. 8. LEADERSHIP PROGRAM.

    (a) In General.--By not later than 1 year after the date of the 
enactment of this Act, the Commission shall establish a leadership 
program under which the Commission recognizes covered entities that 
display excellence in fair, accountable, and transparent data science 
or machine learning, particularly with respect to algorithmic 
eligibility determinations.
    (b) Process.--Such leadership program shall include a process for--
            (1) covered entities to apply to the Commission for such 
        recognition;
            (2) the Commission, in consultation with the National 
        Institute of Standards and Technology, to evaluate the merits 
        of applications in accordance with standards that the 
        Commission shall promulgate as a rule under section 553 of 
        title 5, United States Code; and
            (3) recognizing covered entities that the Commission 
        determines have achieved excellence in fair, accountable, and 
        transparent data science or machine learning, particularly with 
        respect to algorithmic eligibility determinations.

SEC. 9. AVOIDING DUPLICATION.

    In implementing the provisions of this Act, the Commission shall 
work with other Federal agencies to avoid the burden on a covered 
entity from having to comply with requirements under this Act that may 
be duplicative of requirements under other provisions of Federal law.
                                 <all>