Can’t Regulate ADS Algorithms? Let’s Try This Instead.
In his January 4th article for The Markup, Todd Feathers notes that “Governments increasingly use algorithms to do everything from assign benefits to dole out punishment—but attempts to regulate them have been unsuccessful.”
In the U.S. and perhaps elsewhere, government agencies are using unregulated automated decision systems (ADS) for distributing health benefits, criminal sentencing, handling child abuse cases, among other delicate issues.
Algorithms may introduce an element of objectivity, but who’s to be held accountable when an algorithm’s decisions negatively impact society?
I’m going to show that a very established method has been solving public accountability problems like this one for literally centuries, and that that method can be effectively applied to this accountability problem.
But first let’s first explore the kind of chaos ADS algorithms have been causing.
Lawmakers Have Been Unable to Regulate Algorithms Used by State Governments
In his article, Todd Feathers clearly detailed how these algorithms have been found to make decisions that have profoundly negative effects on society. For instance, the PredPol software used across the country by U.S. police departments to predict where crimes might occur has been found to disproportionately send police to Black and Hispanic neighborhoods. In some states, algorithms have automatically stopped aid for thousands of needy residents.
Who’s to be held accountable for these unfair decisions?
The concern is that attempts to regulate the ADS have been futile across various U.S. states. Lawmakers in Massachusetts, Maryland, Michigan, Washington state, New Jersey, California, and Vermont have taken a stab at regulating ADS algorithms, and none of them has resulted in anything significant. Bills proposing oversight over the ADS have either been thrown out or remain pending.
According to Feathers, the Lawmakers trying to regulate ADS tools struggle to understand how they work. They don’t know what algorithms are in use, and the extent to which sensitive decisions have been automated. We surely cannot expect them to understand all that.
How will they then be able to regulate these systems if they have no clue about how they work?
Well, lawmakers also tend not to be structural engineers and architects, or for that matter physicians or mortgage signing officer. Do we need a bunch of lawmakers to understand stress loading algorithms in order to regulate the habitability of buildings? Of course not.
Instead, legislators create a structural engineer’s professional license, for which a structural engineer must qualify in order to practice their profession. But most importantly, that professionally licensed structural engineer puts their license – and livelihood – on the line when they sign an application for an occupancy permit. Later, if the building develops structural defects, the structural engineer who attested that it was sound might find themselves looking for a new career.
Exactly that same principle of professional accountability can be applied to the functioning of ADS algorithms. An applicant for ADS Algorithmer Professional License would be expected to demonstrate that they understand all the nuances of invoking algorithms in applying public policy, and, once licensed, to take legal responsibility for the proper and fair effects of the application of those algorithms.
Professional Licensing
Legislators and those governed by these algorithms do not need to concern themselves how they are built and how they function, just as occupants of buildings need not know the detailed standards and procedures that govern the building's occupancy permit, even though those standards and procedures are a matter of public record.
It is not the business of legislators to determine how ADS algorithms should be built and work, or even regulate ADS auditing. They should simply advocate for the existence of professionally licensed ADS Algorithmer. The professional licenses must be issued by a source of duly constituted public authority (DCPA).
The professionally licensed ADS Algorithmers must have public authority to sign the code that embodies the ADS algorithms. Their obligation is to vet the algorithms and assure that code they’ve audited and signed will not negatively affect the public.
For examples of professional licenses that will bring authenticity to the digital world via professional accountability, see https://www.osmio.ch/professional-license.html. The City of Osmio was chartered on March 7, 2005 at the Geneva headquarters of the International Telecommunication Union, a United Nations agency, to serve as a source of duly constituted public authority for the digital world. Osmio’s Professional Licensing Department is set up to qualify applicants for licenses to practice the digital professions, after their expertise and identity and background has been vetted and proven.