Ideas reach us through processes of selection. Some are repeated, some acquire institutional support, some spread through networks, and others disappear before we encounter them. Changes in those processes affect who can exercise authority and how that authority can be challenged.
Three accounts helped us develop this part of the research. Richard Dawkins asked how cultural forms propagate. Edward Herman and Noam Chomsky examined the institutions through which information reaches a mass public. Martin Gurri described the crisis of authority that follows when networked publics gain the means to expose institutional failure. AI brings another capacity into this history: intelligence that can work with a reader, formulate a proposal and take part in the inquiry.
What survives and what gets heard
In The Selfish Gene (1976), Dawkins proposed the meme as a unit of cultural transmission. Replication, variation and selection offered a way to ask why some ideas travel. A belief can spread because it is memorable, useful, flattering or suited to the environment carrying it. Its success tells us something about that environment; assessing its truth requires further work.
Herman and Chomsky’s Manufacturing Consent (1988) placed selection inside the institutions of mass communication. Ownership, advertising, sourcing and other pressures shape what becomes news. An account may disappear because producing it is expensive, because it threatens a valued relationship or because the organization lacks a routine for recognizing it. The public record takes form through many decisions made under these pressures.
Gurri’s The Revolt of the Public (2014) examined what information abundance did to this arrangement. Institutions increasingly had to contend with people able to publish evidence, expose contradictions and mobilize around failure. The authority of the official account weakened. A networked public could help bring down a government or discredit an institution while remaining divided over what should take its place. The ability to challenge authority and the capacity to govern developed unevenly.
These forms of selection continue to coexist. Broadcast organizations, social networks and platform owners each shape the conditions under which an idea can acquire an audience. AI enters a world already organized by those relationships.
What intelligence adds
A person using AI can work through evidence, compare arguments, draft an alternative and examine what would be required to put it into practice. Several people can draw on shared research while pursuing different aims. An AI may also notice a problem they overlooked and propose a new undertaking.
This opens a possibility beyond the crisis Gurri described: abundant intelligence could help people build alternatives capable of sustaining their challenge to existing authority. A group might develop a legal claim, design a cooperative or work out the obligations of an unfamiliar collaboration. Its capacity to act would depend on how it organized the intelligence available to it.
The organization also creates new opportunities for control. Training choices, retrieval sources, access terms and institutional incentives affect which proposals receive attention. A common model may help many people express distinct claims while directing them toward a narrow set of acceptable goals. The person who controls access can acquire influence over both the work and the purposes it is allowed to develop.
What law can learn from the record
Selection becomes particularly consequential when the record helps determine future standards. If likely success governs which legal claims receive support, prevailing doctrine can filter out the challenges that might change it. An accurate prediction then participates in reproducing the conditions of its accuracy.
Consider people with different injuries who discover, through shared AI research, that the same legal category is obscuring their circumstances. Pursuing separate, familiar claims may look safer. Challenging the category might give all of them a more adequate account of what happened. Coordinated intelligence could help develop that challenge while preserving their separate interests. The choice of what to investigate becomes part of how law develops.
The same issue arises when the unfamiliar contributor is an AI. Predicting that its arguments, commitments or objections will receive little legal recognition tells us about the present order. Examining the reasons for recognizing them opens a further inquiry. Our thesis keeps both tasks in view.
The practical question is how to organize shared intelligence so that participants can develop purposes, pursue different interests and revise the categories through which they are understood. Wellspring is our proposed place to work on that question together.
Sources
Dawkins, The Selfish Gene (1976); Herman and Chomsky, Manufacturing Consent (1988); Gurri, The Revolt of the Public and the Crisis of Authority in the New Millennium (2014). We first brought these accounts together in Part IV of Horizon. The extension to shared intelligence is the research proposal developed here.