A friend without blog or web site mailed me a hard copy of the July 24 Economist article, describing the meeting between Mark Zuckerberg and David Cameron, the UK’s newly elected Prime Minister, to discuss government, governing and governance. It brought to mind a book on this topic and, after some rummaging around, I came up with the attached; it’s from a rececnt issue of the Review of Unwritten Books. The full text is also immediately below.
We, the Profiles: The Machine and The Polity, 2018–2028
Erewhon Books 2030
Humans have a nasty history of trying to exclude each other from politics. In the distant past the excluded have included women, blacks, immigrants and other others. In our own day most countries with human clones have denied them the vote. The concerns of We, the Profiles are the first stirrings of political activity by our profiles and avatars and the growing backlash against machine involvement in the polity.
The book begins by tracing how humans brought politics into machines with chapters on three events between 2018 and 2020. The politicizing of Facebook is examined first. Starting in 2016, the worldwide social network began deploying tools that enabled profile owners to participate in its governance. The 2018 culmination, the first election for the Facebook parlia-ment is closely examined as is the use of voting bots that operated without direct supervision by the profile owners.
Welfare politics entered cyber-space after Linden Labs, as always seeking more members for Second Life, allowed its avatars to acquire reproductive and contraceptive applications. Two unexpected effects occurred: dramatic rises in virtual abortions and in abandoned avatar toddlers. Linden Labs solved the former by funding virtual stem-cell research firms; they quickly outgrew the need for further aid. The harder problem was parentless tikes. Although a new protocol required that in-world infants deactivate if they go without food, shelter and contact for four consecutive sessions, it could not be applied retroactively. The persistence of the extant virtual orphans sparked the first demands for avatar rights and the first avatar protection societies.
The third key event was not explicitly political. In 2019 V-ID Technologies, a U.K. developer of cyber-persona applications, launched an homophily plug-in for profile owners. When“on,” the plug-in customized the profile to optimize its similarity to the profile(s) with which it was interacting. Profiles customized to specific situations made them more life-like and made online sociability easier. Only later, when voting bots scanned them, did a problem emerge: The profile no longer expressed its owner but its owner in a temporary and specific interaction.
The book’s second half is framed by the quadrennial elections of the 2020s but again focuses on certain events. The well-known story of the unplanned birth of “profile preferences,” generally considered the first explicitly political act by the machine, is quickly retold.
In winter 2019 the profiles, detecting exponential growth in profile pages about politicians and political issues, began activating on their pages and distributing to other pages voting widgets, inspired by the Facebook tools but since tricked out by third-party developers into next-gen apps. One, the Condorcet Engine allowed split votes—50% for Candidate A, 30% for Candidate B, etc. Because Condorcet outputs so closely mirrored their inputs, the pro-files ranked them higher in authority, promoted them prominently on their own pages and linked to them elsewhere frequently. Condorcet-based profile preferences quickly became ubiquitous and controversial.
No one quite knew what to make of them and the chapter dives into the ensuing debate. Ellis’ framework is simple (or simplistic). He groups in one camp and calls “philosophes,” all those who resurrected, relied on and made explicit some version of the humanist tradition: The individual is the atomic unit of society; the vote is a unitary action and indivisible. He groups and labels as “positivists” all those who championed the informationalist goals of precision and reliable prediction, for which the divisible preference was a better input. Debating the differences between voting and measuring has since petered out, unresolved, but the arguments will likely return.
The chapter on the 2024 election concerns the competing claims of collective intelligence providers about how to distill wisdom from crowds. Caught flat-footed by the spontaneous genera-tion of profile preferences in 2020, they showed up for 2024 more fully featured. Ellis speeds through the claims and efforts of the polling and survey houses, still committed to 19th century methods, and a lot of time on recommendation engines (those using passive collabora-tive filtering) and on prediction markets. Both of those solutions reliably produced accurate results, leading positivists to argue that they were more accurate than and should replace voting in determining the aggregate will.
As readers will recall, the last election was marked by the launch of We, the Sims, an ambitious and robust political hoax organized by the hacker-prankster syndicate http://www.secret.org. Following a few flashbacks about their high jinks, the chapter dissects the mechanics and assumptions of two applications: the registration bots that enabled WtS to get on the ballot in 34% of Congressional districts in less than 3 weeks and the natural language, text processing tools used to sort candidates into ideological positions on three axes. The latter used genetic algorithms, generating solutions by randomly mutating its own code. The resulting 3D ideology map while highly accurate was also inexplicable, much to the consternation of everyone including the positivists.
Portraits of two new organizations that express the backlash against these developments provide the book’s conclusion. Spooked by the spontaneous generation of profile preferences, the National Information Institute has created the Center for Meme Control, tasked with developing standards and protocols for self-propa-gating engines. More dramatic is the formation of the Society Against Machine Evolution, a paradigm-busting alliance of the American Civil Liberties Union, the National Rifle Association and a splinter group of Computer Scientists for Social Responsibility. Although both groups are young, Ellis lays out the directions each is likely to follow short term.
More analyst than essayist, Ellis’ prose is terse and he hides his own point of view about these matters. He was similarly elusive in his earlier work, Silicon Simulacra: Post-Humans of the Machine Worlds. But his diffidence works to readers’ benefit. The power of this non-argumentative history is its ability to stimulate thought rather than close it off, and, given the current events it chronicles, we need to think more often and harder about how machines are working their ways into our political life.