Paper by Rebecca Crootof, Margot E. Kaminski and W. Nicholson Price II: “From lethal drones to cancer diagnostics, complex and artificially intelligent algorithms are increasingly integrated into decisionmaking that affects human lives, raising challenging questions about the proper allocation of decisional authority between humans and machines. Regulators commonly respond to these concerns by putting a “human in the loop”: using law to require or encourage including an individual within an algorithmic decisionmaking process.
Drawing on our distinctive areas of expertise with algorithmic systems, we take a bird’s eye view to make three generalizable contributions to the discourse. First, contrary to the popular narrative, the law is already profoundly (and problematically) involved in governing algorithmic systems. Law may explicitly require or prohibit human involvement and law may indirectly encourage or discourage human involvement, all without regard to what we know about the strengths and weaknesses of human and algorithmic decisionmakers and the particular quirks of hybrid human-machine systems. Second, we identify “the MABA-MABA trap,” wherein regulators are tempted to address a panoply of concerns by “slapping a human in it” based on presumptions about what humans and algorithms are respectively better at doing, often without realizing that the new hybrid system needs its own distinct regulatory interventions. Instead, we suggest that regulators should focus on what they want the human to do—what role the human is meant to play—and design regulations to allow humans to play these roles successfully. Third, borrowing concepts from systems engineering and existing law regulating railroads, nuclear reactors, and medical devices, we highlight lessons for regulating humans in the loop as well as alternative means of regulating human-machine systems going forward….(More)”.