Algorithms Still Have a Bias Problem, and Big Tech Isn’t Doing Enough to Fix It – PCMag

Wondering about the health of the internet, both globally and in the US? Mozillathe company behind the Firefox browser and moretoday released its fourth Internet Health Report, which covers topics that impact the internet and big tech companies, such as accountability, labor, andas we'll discuss hereracial justice issues.

The hard numbers are above, pulled from tech companies' own reports, as charted by Wired. Among the biggest fourApple, Facebook, Google (Alphabet), and Microsoft (Amazon isn't listed)the stats on Black, Latinx, and Native representation among the workforce in each are practically unchanged after five years (see update below). The only real change is the minimal increase in the number of Asian employees, up 12% at Apple, 11% at Facebook, 9% at Google, and 4% at Microsoft. The report flatly states that even in that case, caste discrimination happens.

Gender diversity isn't making great strides, either. Women are still woefully underrepresented at those same four companies; increases have been very small since 2014.

The racial biases of having a primarily white workforce persist: for example, skewing perspectives on artificial intelligence, as evidenced by all the white plastic humanoids in a typical search.

The report includes a spotlight article titled "Decode the Default," which looks at the recent phenomenon of trying to call out the "racial inequities of data and algorithms," and the backlash and denials that arise when it happens. For example, the 2008 launch of a web browser called Blackbird (built on Mozilla code) as a "browser for Black people" faced accusations of segregationeven from the African-American community. Other examples include:

Over a decade ago, it was discovered that searching "black girls" on Google lead mostly to imagery from pornography.

Facial-recognition gets things woefully wrong for people who aren't white.

The "mainstream notion of a 'default' user" is typically believed to be white, cisgender, male, and American.

Many of these race-based problems and more in the tech community point back to the graphic at the top: The big tech companieson which the small tech companies are almost wholly dependent"fail to create work environments where people of color and women want to stay." Without that diversity internally, tech companies don't really know how to work and improve beyond their limited worldview.

There's much more to read in "Decode the Default" and the entire Internet Health Report, including sections on the labor rights movement in tech (in particular for "gig economy" workers) and the lack of transparency that prevents us from holding these companies accountable.

Update 1/30/2021: Microsoft says the Mozilla report pulls data from its 2019 report. It has since published anew diversity reportfor 2020, which cites "modest gains since 2019 including among women who now represent 28.6% of the global Microsoft workforce, an increase of 1.0 percentage point since last year.However, racial and ethnic minority communities have largely seen incremental progress and there is still much work to be done."

Further Reading

View original post here:

Algorithms Still Have a Bias Problem, and Big Tech Isn't Doing Enough to Fix It - PCMag

Related Posts

Comments are closed.