A Response to "How AI Destroys Institutions"

Abstract

In “How AI Destroys Institutions,” Professors Woodrow Hartzog and Jessica Silbey argue that Generative AI (GenAI) systems—by their very design—undermine expertise, short-circuit decision-making, and isolate humans from one another.  These aren't bugs to be fixed with better governance. They're features, baked into the technology's core affordances.  The conclusion is stark: GenAI constitutes "a death sentence for civic institutions," like higher education, the rule of law, journalism, and democracy itself.

Hartzog and Silbey offer a provocative argument. In this response, I focus on one methodological issue that would sharpen their analysis and make the likely institutional effects appear more contingent: the baseline question—relative to what?

Every claim about GenAI's destructive effects implicitly invokes a baseline. GenAI destroys institutions compared to... what? They never quite say, and the ambiguity does a lot of work.

The unstated comparison seems to be institutions as they ought to function: repositories of legitimate knowledge, sites of transparent deliberation, communities bound by shared purpose. The rule of law as guarantor of accountability. Universities as engines of free inquiry. Journalism as democracy's watchdog. These are noble aspirations. They're also not how these institutions operate in practice.

This isn't a minor oversight. The "relative to what" problem pervades the analysis of each institution, with the flagged risks either overstated or failing to recognize potential benefits to these institutions from the use of GenAI. Although Hartzog and Silbey do recognize that our institutions are currently fragile and sometimes ineffective, their analysis would be strengthened by engaging more directly with the extent of those problems—and by analyzing when GenAI might mitigate rather than worsen them.

The reality is that any predictions about GenAI’s institutional impact—whether a “death sentence” or utopia—are simplifications. The future is much likely to unfold in ways that we cannot even begin to imagine.