Task force, commission want government intervention with AI; especially facial recognition
Reports pushing recommendations for more-careful use of public algorithms continue to get more direct and succinct: Communicate, get stakeholders involved early, monitor and evaluate.
It is basic B school — except that many accountability groups continue to add a call for heavy shackles on face biometrics.
A task force created by University of Pittsburgh, for instance, has published a report examining the use of AI algorithms, including those behind facial recognition systems, in its home city and county, Allegheny.
Meanwhile, a Long Beach, California, commission has handed a similar, but more focused, report to that city’s leaders.
The Pittsburgh group’s stated goal was to help AI “flourish” where needed and appropriate, on one end of a spectrum and, on the other, gate it from any area where misapplication could threaten individuals’ rights.
To summarize, the task force felt most work that is based on conventional administrative roles could be augmented or fully automated so long as government was open, transparent, careful and responsive to problems and mistakes related to their deployment.
But agencies should “avoid” face biometrics for the “foreseeable future.”
Algorithms have biases and can make mistakes that particularly affect people who are not middle-aged, male and white, the report states.
Even if bias and agency misconduct can be eliminated and errors made all-but non-existent, according to the authors, blanketing areas with unblinking eyes could redefine privacy, and not for the better.
And last week, Long Beach’s Technology and Innovation Commission passed on to the city council its recommendation that a moratorium be imposed on all facial recognition systems.
That was one of three steps commissioners want the council to take. No date has been set for considering the report.
The other two points call for an independent commission with “authority and oversight” over AI algorithm use and protocols for vetting and monitoring technologies that harvest biometric identifiers.