The Emerging Law of Algorithms, Robots, and Predictive Analytics

In 1897, Holmes famously pronounced, “For the rational study of the law the blackletter man may be the man of the present, but the man of the future is the man of statistics and the master of economics.” He could scarcely envision at the time the rise of cost-benefit analysis, and comparative devaluation of legal process and non-economic values, in the administrative state. Nor could he have foreseen the surveillance-driven tools of today’s predictive policing and homeland security apparatus. Nevertheless, I think Holmes’s empiricism and pragmatism still animate dominant legal responses to new technologies. Three conferences this Spring show the importance of “statistics and economics” in future tools of social order, and the fundamental public values that must constrain those tools.

Tyranny of the Algorithm? Predictive Analytics and Human Rights

As the conference call states

Advances in information and communications technology and the “datafication” of broadening fields of human endeavor are generating unparalleled quantities and kinds of data about individual and group behavior, much of which is now being deployed to assess risk by governments worldwide. For example, law enforcement personnel are expected to prevent terrorism through data-informed policing aimed at curbing extremism before it expresses itself as violence. And police are deployed to predicted “hot spots” based on data related to past crime. Judges are turning to data-driven metrics to help them assess the risk that an individual will act violently and should be detained before trial. 


Where some analysts celebrate these developments as advancing “evidence-based” policing and objective decision-making, others decry the discriminatory impact of reliance on data sets tainted by disproportionate policing in communities of color. Still others insist on a bright line between policing for community safety in countries with democratic traditions and credible institutions, and policing for social control in authoritarian settings. The 2016 annual conference will . . . consider the human rights implications of the varied uses of predictive analytics by state actors. As a core part of this endeavor, the conference will examine—and seek to advance—the capacity of human rights practitioners to access, evaluate, and challenge risk assessments made through predictive analytics by governments worldwide. 

This focus on the violence targeted and legitimated by algorithmic tools is a welcome chance to discuss the future of law enforcement. As Dan McQuillan has argued, these “crime-fighting” tools are both logical extensions of extant technologies of ranking, sorting, and evaluating, and raise fundamental challenges to the rule of law: 

According to Agamben, the signature of a state of exception is ‘force-of’; actions that have the force of law even when not of the law. Software is being used to predict which people on parole or probation are most likely to commit murder or other crimes. The algorithms developed by university researchers uses a dataset of 60,000 crimes and some dozens of variables about the individuals to help determine how much supervision the parolees should have. While having discriminatory potential, this algorithm is being invoked within a legal context. 

[T]he steep rise in the rate of drone attacks during the Obama administration has been ascribed to the algorithmic identification of ‘risky subjects’ via the disposition matrix. According to interviews with US national security officials the disposition matrix contains the names of terrorism suspects arrayed against other factors derived from data in ‘a single, continually evolving database in which biographies, locations, known associates and affiliated organizations are all catalogued.’ Seen through the lens of states of exception, we cannot assume that the impact of algorithmic force-of will be constrained because we do not live in a dictatorship. . . .What we need to be alert for, according to Agamben, is not a confusion of legislative and executive powers but separation of law and force of law. . . [P]redictive algorithms increasingly manifest as a force-of which cannot be restrained by invoking privacy or data protection. 

The ultimate logic of the algorithmic state of exception may be a homeland of “smart cities,” and force projection against an external world divided into “kill boxes.” 


We Robot 2016: Conference on Legal and Policy Issues Relating to Robotics

As the “kill box” example suggests above, software is not just an important tool for humans planning interventions. It is also animating features of our environment, ranging from drones to vending machines. Ryan Calo has argued that the increasing role of robotics in our lives merits “systematic changes to law, institutions, and the legal academy,” and has proposed a Federal Robotics Commission. (I hope it gets further than proposals for a Federal Search Commission have so far!)


Calo, Michael Froomkin, and other luminaries of robotics law will be at We Robot 2016 this April at the University of Miami. Panels like “Will #BlackLivesMatter to RoboCop?” and “How to Engage the Public on the Ethics and Governance of Lethal Autonomous Weapons” raise fascinating, difficult issues for the future management of violence, power, and force.


Unlocking the Black Box: The Promise and Limits of Algorithmic Accountability in the Professions


Finally, I want to highlight a conference I am co-organizing with Valerie Belair-Gagnon and Caitlin Petre at the Yale ISP. As Jack Balkin observed in his response to Calo’s “Robotics and the Lessons of Cyberlaw,” technology concerns not only “the relationship of persons to things but rather the social relationships between people that are mediated by things.” Social relationships are also mediated by professionals: doctors and nurses in the medical field, journalists in the media, attorneys in disputes and transactions.


For many techno-utopians, the professions are quaint, an organizational form to be flattened by the rapid advance of software. But if there is anything the examples above (and my book) illustrate, it is the repeated, even disastrous failures of many computational systems to respect basic norms of due process, anti-discrimination, transparency, and accountability. These systems need professional guidance as much as professionals need these systems. We will explore how professionals–both within and outside the technology sector–can contribute to a community of inquiry devoted to accountability as a principle of research, investigation, and action. 


Some may claim that software-driven business and government practices are too complex to regulate. Others will question the value of the professions in responding to this technological change. I hope that the three conferences discussed above will help assuage those concerns, continuing the dialogue started at NYU in 2013 about “accountable algorithms,” and building new communities of inquiry. 


And one final reflection on Holmes: the repetition of “man” in his quote above should not go unremarked. Nicole Dewandre has observed the following regarding modern concerns about life online: 

To some extent, the fears of men in a hyperconnected era reflect all-too-familiar experiences of women. Being objects of surveillance and control, exhausting laboring without rewards and being lost through the holes of the meritocracy net, being constrained in a specular posture of other’s deeds: all these stances have been the fate of women’s lives for centuries, if not millennia. What men fear from the State or from “Big (br)Other”, they have experienced with men. So, welcome to world of women….

Dewandre’s voice complements that of US scholars (like Danielle Citron and Mary Ann Franks) on systematic disadvantages to women posed by opaque or distant technological infrastructure. I think one of the many valuable goals of the conferences above will be to promote truly inclusive technologies, permeable to input from all of society, not just top investors and managers.

X-Posted: Balkinization.

You may also like...