Henry Farrell has a new article in the Annual Review of Political Science, and I cannot recommend it highly enough. The central idea is that:

Political scientists might engage AI by investigating its relationship to governance. How does AI affect existing forms of governance, such as markets, bureaucratic hierarchy, and democracy? Is AI itself a kind of governance?

However, governance too is a vexed term (Mayntz 2009). As Peters (2012, p. 19) tartly remarks, “the ambiguity of the concept…has been one of the reasons for its popularity.…It…obfuscates meaning at the same time that it perhaps enhances understanding.” Peters, however, notes that despite its variety of meanings, the word governance originally referred to steering a boat, like cybernetics (derived from the Attic Greek word for “steersmanship”). Governance, along with cybernetics and control (Beniger 1986, Wiener 2019, Yates 1993), is a catchall phrase for forms of social, political, and economic coordination that especially emphasize information processing.

Stealing from broadly analogous ideas presented by Simon [2019 (1968)], I treat governance as an umbrella term for the large-scale systems for processing information and social coordination that allow complex societies to work. A system of governance, then, has (a) an input, some large-scale source of complex information; (b) a technology for turning that information into useful, albeit lossy representations that can more readily be manipulated; and (c) outputs that can be used to coordinate on the basis of those representations.

In markets, the price mechanism summarizes tacit knowledge (Polanyi 1966) about relations of production, allowing widespread economic coordination (Hayek 1945, Lindblom 2002). In bureaucratic hierarchy, authority relations and classification systems turn diffuse social knowledge into tractable information that can enable government rulemaking (Scott 1998, Weber 1968). In democracy, mechanisms for representation and voice turn the desires and knowledge of citizens into tractable representations that enable feedback and control over their collective circumstances (Allen 2023, Dewey 1927).

These systems of governance are at best highly imperfect. The price mechanism, bureaucratic categories, and representations of the democratic public are “simulations” [Simon 2019 (1968)] or very lossy coarse-grainings (Flack 2017) of irreducibly complex underlying realities. But even harsh critics (Scott 1998) acknowledge that modern large-scale societies would be impossible without them.

They are deeply entangled with one another. Markets depend both on the external institutions of government (North 1990) and on the internal bureaucratic hierarchy of the firm (Coase 1937). Bureaucracies draw on markets and have regularly sought to import their logic (Dunleavy & Hood 1994). Democracy depends on bureaucratic hierarchy to implement decisions and, as Lindblom (2002) observes, seems practically conjoined to the market economy.

The value of this approach to governance is not that it provides precise definitions, let alone testable hypotheses, but rather broad heuristics. We can see how AI may not be a putative substitute for individual human intelligence but instead a means of collective information processing and coordination. Specifically, we can consider AI either (a) as an external technology, affecting how existing systems of governance coordinate and process information, or (b) as a possible form of governance in its own right, with its own particular form of coordination and information processing. How does AI affect the internal workings of existing forms of governance? Might it become its own form of governance, with associated pitfalls and possibilities (Farrell & Shalizi 2023)? These two broad questions motivate different but partially overlapping research agendas.

I strongly recommend reading the entire article (which will likely remain open access until the next edition of the Annual Review of Political Science comes out). It is a great example of scholarly engagement with the anxieties of our era.

My own field, where we focus a lot on how technological transformations reshape terrorist tactics and extremist politics, can only be strengthened by deep engagement with how these technologies may similarly reshape societal structures and mainstream politics.