Discussion: View Thread

  • 1.  Algorithmic Discrimination

    Posted 10-10-2022 14:26
    This draft of a new law may eventually have implications for statisticians working  in the D.C area and potentially for all other statisticians.
    The D.C. City council is preparing vote to ban "algorithmic discrimination". Basically computer algorithms used to make say, mortgage loan recommendations , medical treatment recommendations etc. have the potential or possibly have been shown to be biased
    A Description of the law 
    AG Racine Introduces Legislation to Stop Discrimination In Automated Decision-Making Tools That Impact Individuals' Daily Lives (dc.gov)

    And a new (to me) word to add to the statisticians pantheon "mathwashing"

    ...Second, algorithmic decision-making conceals discrimination behind a veneer of supposed scientific "objectivity." This is known as "mathwashing."..


    ------------------------------
    Chris Barker, Ph.D.
    2022 Statistical Consulting Section
    Chair-elect
    Consultant and
    Adjunct Associate Professor of Biostatistics
    www.barkerstats.com


    ---
    "In composition you have all the time you want to decide what to say in 15 seconds, in improvisation you have 15 seconds."
    -Steve Lacy
    ------------------------------


  • 2.  RE: Algorithmic Discrimination

    Posted 10-11-2022 00:42
    Whenever the topic of algorithmic bias comes up, I cannot help but wonder, what if the algorithm is correct, but we're offended by its results?

    ------------------------------
    Eric Siegel, MS
    Biostatistics Project Manager
    Department of Biostatistics
    Univ. Arkansas Medical Sciences
    ------------------------------



  • 3.  RE: Algorithmic Discrimination

    Posted 10-11-2022 12:46
    Hi Chris, thanks for sharing this. Here are some of my initial thoughts

    * One thing I don't get is: it's already been illegal to discriminate in general against protected classes eg fair lending laws, so how does banning "algorithmic discrimination" specifically actually change anything?  

    * The "private right of action" makes me concerned, maybe should get more insurance - not sure if standard policies cover that, we don't really have malpractice insurance for that like doctors would, although data breach policies are available.  I also wonder if this applies to the disclosure requirements as well, like could any random person living in DC now sue a company for $10k for not having a DC-compatible disclosure on their website (even if using a model that is /not/ "biased")?

    * The mandate for annual "algorithmic impact" audits seems really onerous, has the same potential downsides as all the environmental impact reviews that have been slowing down new housing development and public transportation in LA. 






  • 4.  RE: Algorithmic Discrimination

    Posted 10-24-2022 16:15
    An update on algorithmic discrimination .
    An Announcement from Emory and at least one  very-well-researched articles ( Science articles accessed through a paywall )
    On my possibly too quick read, ACLU (see below) refers to a problem with the training datasets used

    https://www.science.org/doi/10.1126/science.aax2342
    Obermeyer, Ziad, et al. "Dissecting racial bias in an algorithm used to manage the health of populations." Science 366.6464 (2019): 447-453.

    a public domain version of the above Science articles is available from the FTC
    https://www.ftc.gov/system/files/documents/public_events/1548288/privacycon-2020-ziad_obermeyer.pdf 


    Source: I found these links from an ACLU advert on FB. Posted here for reference only



    caveat: ACLU cites 2019 FDA guidance, though I  don't think the FDA guidance was specifically developed to address algorithm bias in -general-.
     FDA guidance refers to "clinical decision support " (CDS)

    https://www.fda.gov/media/109618/download



    ------------------------------
    Chris Barker, Ph.D.
    2022 Statistical Consulting Section
    Chair-elect
    Consultant and
    Adjunct Associate Professor of Biostatistics
    www.barkerstats.com


    ---
    "In composition you have all the time you want to decide what to say in 15 seconds, in improvisation you have 15 seconds."
    -Steve Lacy
    ------------------------------