What Evans, Hagiu, and Schmalensee say about the “symbiotic relationship” developers and companies have with platforms, and the “ecosystems” thereby created, applies equally well to this more general and informal use of the word. But just as that relationship is governed and managed for developers—you can’t write code to be used on Windows or MacOS or Facebook that doesn’t comply with the relevant API—so too the relationship for users is managed, but often in opaque ways. The “terms of service” the walled factories provide for their users can be, and typically are, evasive and vague in ways that APIs cannot be. This vagueness allows Facebook to conduct experiments on its users—for instance, tracking your keystrokes, including keystrokes you type but then delete without posting—and to shape its users’ timelines in ways that encourage them, for a great variety of reasons, to stay within the walls of Facebook rather than venture beyond, whether for pleasure or profit.12

The importance of this management can best be seen by comparing a platform like Facebook with traditional social institutions. I think what we have seen and will continue to see in our social order is the fragmentation of institutions and their effective replacement by platforms. As Astra Taylor explains in her vital book The People’s Platform, this process has often been celebrated by advocates of new platforms. Esther Dyson concisely summarizes this view: “The great virtue of the Internet is that it erodes power. It sucks power out of the center, and takes it to the periphery, it erodes the power of institutions over people while giving to individuals the power to run their lives.” But, after quoting Dyson, Taylor notes that

the problem, though, is that disintermediation has not lived up to its potential. Instead, it has facilitated the rise of a new generation of mediators that are sometimes difficult to see. As much as networked technology has dismantled and distributed power in more egalitarian ways, it has also extended and obscured power, making it less visible and, arguably, harder to resist.13

We may take education as an example. For much of American history, people were educated in a wide range of (often highly eccentric) settings. This was generally perceived as a problem, and efforts at standardization kicked in, reaching their peak in the 1960s. Since then we have seen increasing fragmentation, with ordinary public schools, charter schools, magnet schools, various kinds of private schools, homeschooling, unschooling—but all of these work on the same platforms; that is, they rely on the same communications technologies, using either the open Web or walled factories like Facebook in order to promote interaction and accomplish goals (e.g., the completion of projects and other assignments, remedial tutoring). More and more, we will be asking technological platforms to do the kind of unifying work educational institutions clearly can no longer do—and that, I believe, is asking platforms to do things that by their nature they’re unsuited to do. But this is precisely the future Mark Zuckerberg envisions in his lengthy, utopian, and extremely dishonest manifesto, released in February 2017, called “Building Global Community.”14

Platforms of the Facebook walled-factory type are unsuited to the work of building community, whether globally or locally, because such platforms are unresponsive to their users, and unresponsive by design (design that is driven by a desire to be universal in scope). It is virtually impossible to contact anyone at Google, Facebook, Twitter, or Instagram, and that is so that those platforms can train us to do what they want us to do, rather than be accountable to our desires and needs. A model of education tied to platforms rather than institutions may seem liberating at first—“I can learn everything I need to know on MOOCs!”—but that sense of liberation will continue only insofar as users train themselves to ask the questions the platforms already know how to answer, and think the thoughts the platforms are prepared to transmit.