Algorithms as politics
Building principles-based digital institutions in the 21st century, in fiction and reality
Hello World: Being Human in the Age of Algorithms, Hannah Fry, W. W. Norton & Company, 272 pages, $25.95
In the year 2052, a series of “Machines” manage the world economy. These advanced robot brains allocate resources and ensure humanity’s welfare. Although people formally control the Machines, they have begun to operate at a level that evades human comprehension. Still, they remain beholden to mankind’s interests through the Three Laws of Robotics. By hard-coding these inviolable rules in each successive generation, humans ensured that robots would continue to operate on the principle of “do no harm.”
This scenario provides the backdrop to Isaac Asimov’s “The Evitable Conflict.” Written in 1950, Asimov’s short story concludes the narrative arc of the I, Robot series, which chronicles the transformation of autonomous machines from mute domestic servants to hyper-rational guardians of the species.
Travel a short distance and two decades to the alternate universe of Robert Heinlein’s The Moon is a Harsh Mistress. In 2075, the supercomputer that runs a lunar penal colony’s life-support systems gains self-awareness. An engineer discovers the fact, befriends the computer, and uses its intelligence and computational power to lead an insurrection against the Earth’s rule, causing a brief war with several thousand casualties. Moon depicts the alternative scenario to I, Robot, where a machine programmed without safeguards evolves unexpectedly and turns against its builders. The computer need not be malevolent or anti-human for its erratic behavior to cause harm or disruption.
Asimov and Heinlein’s imagination neatly foreshadowed the quandaries that we face today as we grapple to understand and control algorithms. In Hannah Fry’s Hello World, these dilemmas come to life in a study of these tools in crime, justice, medicine, transportation, and culture. Fry, a mathematician and Associate Professor at University College London, ably links their human effects — where algorithms touch individuals with a name and a face — with their societal repercussions.
Hello World helps explain how algorithms work in a way that is non-technical and accessible to lay readers. This also makes the book at times too superficial for those with some background in algorithms, who will also be familiar with many of the real-life stories that Fry discusses. The mechanics and (mis)application of algorithms are not the most crucial or thought-provoking elements of the book, however. The more fundamental dilemmas center around the values, or lack thereof, with which their human makers endow these mathematical creatures.
To explore this thought, it is helpful to compare algorithms to analog institutions. Talking about the software that powers autonomous vehicles, Fry notes that “At the heart of this new technology — as with almost all algorithms — are questions about power, expectation, control, and delegation of responsibility.” The same applies to structures that humans have been building for millennia. What we call a “parliament” is, in effect, an algorithm that computes decisions based on a deliberate set of values: majoritarian, proportional, or other. Similarly, a tribunal is an algorithm that combines principles like preponderance of the evidence with case facts to issue a verdict. No one in their right mind would conceive to create an institution without defined operating values. To do so for an algorithm should be equally dubious.
Special collisions occur when we insert algorithms in analog institutions. Fry drives at this in her analysis of software in justice and policing. Why should decisions made by code have no right to appeal? Why should police officers target individuals based on a program’s risk assessment that may conflict with standard requirements for probable cause? Programs open themselves to criticism when they do not embody the core principles of the institutions they serve — in government especially.
At the same time, algorithms can be a means through which we improve institutions where these are failing to uphold their principles. Fry notes that “[t]he choice isn’t between a flawed algorithm and some imaginary perfect system. The only fair comparison to make is between the algorithm and what we’d be left with in its absence.” When we plan to introduce an algorithm in, say, a courtroom, we also have the space to re-envision the courtroom itself. By calibrating an algorithm’s values, we can rethink those that underpin analog institutions, too. The result is a more effective whole.
Such a process is evident in the work that is being done on the ethics of self-driving cars, where developers and engineers are tackling issues such as the trade-offs between preserving the lives of passengers and those of pedestrians during an accident. This is perhaps the best example of a thoughtful, relatively transparent discussion about the moral path that an invention should take. As Fry highlights time and again, this approach is happening far less than it should. Too often, decisions on principles are made behind closed doors, or not at all.
Algorithms, even ones developed by the private sector, have politics. Wittingly or not, they contain decisions about right and wrong, about who wins and who loses; questions of “power, expectation, control, and delegation of responsibility.” Hello World pushes us to rethink our relationship with algorithms and their makers. It urges us to abandon the engineer’s belief that the mathematics should prevail; to interrogate the entrepreneur’s sales pitch on “artificial intelligence”; and to doubt the technocrat’s suggestion that a little code is all it takes to fix a broken government system.
We are experiencing a revolutionary moment in digital technology as firms and governments unlock astounding capabilities in computing and software. Without a parallel revolution in the politics of software, such progress could backfire. Algorithms are not demi-gods at whose altar we must worship. They are fallible human creations; institutions like others, requiring judicious development and constant scrutiny. Hello World illustrates the challenges ahead: the gap between Asimov and Heinlein’s futures.
