As artificial intelligence becomes a tool ubiquitous in daily life—and one that distributes access to information more readily—knowledge commons may be in grave danger.
As a young person, I often see my peers using AI tools to gather information. This tendency extends beyond young people, though, from higher education students to employees to the elderly. Generative AI systems frequently draw on resources from knowledge commons like Wikipedia. The communities upholding open information, however, have little say in how and where this information is presented.
This article will explore AI’s effects on knowledge commons and public information. As open information remains freely available, artificial intelligence companies have the ability to consolidate information and interpret it. A knowledge commons could therefore flourish as a source of information yet weaken as an institution: its communities would continue producing knowledge while depending on private corporations to spread that information to a larger audience.
To evaluate AI’s impact, we must consider the crux of knowledge commons. At its root, a knowledge commons brings together a shared body of information with a community that maintains and expands it. Crucially, knowledge has a certain property: one person’s use of knowledge does not diminish what remains available to another. A scientific finding, for instance, can inform thousands of subsequent investigations without being ‘used up’. Indeed, when a finding is accessible, others can test it, challenge it, and apply it to derivative questions. Thus, broad access to knowledge serves a practical purpose—when a community actively engages with knowledge, knowledge develops accordingly.
Developing knowledge, though, requires sustained analysis and improvement. Readers and contributors must evaluate which evidence is sound or how to represent competing accounts of a source. Leading scholars of knowledge commons emphasize community governance—the system a community uses to maintain shared resources. Responsibilities in community governance are a large part of what makes knowledge commons powerful. On Wikipedia, for example, editors work toward consensus on information through constant revisions and discussions with each other. Readers can consume articles but also contribute to how articles should be written, or even submit corrections themselves.
An underlying principle of knowledge commons, then, is that every piece of information and assumption of authority remains open to scrutiny. Such scrutiny can evaluate and correct a claim, giving the public a beacon of information that can change and be challenged. Importantly, scrutiny also occurs across generations. Young people encounter inherited knowledge through circumstances their predecessors may not have experienced. Sociologist Karl Mannheim described this as “fresh contact,” which happens when previously established understandings become available for reconsideration. Within a knowledge commons, this can mean questioning whose experiences a source represents or editing a longstanding article. We might understand this as intergenerational review: people of varying ages examine inherited knowledge together, bring their experiences to the table, and decide what warrants preservation or revision. Young people should have a stake in this process because the accounts they inherit will eventually inform the decisions they make, along with the knowledge they pass on to future generations.
We might understand this as intergenerational review: people of varying ages examine inherited knowledge together, bring their experiences to the table, and decide what warrants preservation or revision.
But what role does AI play, specifically, in this system?
When shared knowledge enters a commercial AI system, it becomes part of a separate, monetized product governed by interests distinct from the knowledge commons. Suppose a community produces information that a corporation uses to train a large language model (LLM). The corporation may combine this work with computations and technical expertise, then sell access to its AI. As a result, the value of this AI is drawn from both corporate investment and the investment (in time and work) of a community independent of the corporation.
These contributions, though, engender different levels of control—the public with less control, and private with more. A company’s monetary investment gives it control over the AI service: it decides what exactly to develop with valuable information, and under what conditions people can access their service. Yet contributing knowledge from the communal end does not give a community similar influence over how information is used. Thus, in the case above, the public may supply an essential resource, simultaneously remaining outside the institutions that dictate the commercial access to the resource.
In other words, a community supplies knowledge, but a company decides how to turn that knowledge into something people will pay to use. Though this may not seem inherently detrimental, contemplate the following. A community may prioritize documenting a subject thoroughly, including disagreements between the community that remain unresolved; at the same time, a company may prioritize helping users complete a task quickly. Both of these groups draw on the same information, but they organize and deploy the information differently.
It would be dubious to suggest that AI solely has an adverse impact on knowledge commons. Chatbots can translate articles and explain complex terms to laypeople, or synthesize and relate dense material scattered in different sources. Those who would struggle to interpret a heavy research paper may understand its findings through a well-worded summary. In fact, Wikimedia itself acknowledges that information written by its volunteers now reaches people who never typically visit Wikipedia, and welcomes these new ways of information access. Wider circulation almost certainly helps advance the commons’ public purpose, since a precursor to sustaining information is access to information. So what gives?
The problem is that wider access to knowledge does not necessarily entail wider examination and challenging of knowledge. Knowledge commons need not dictate how others interpret their work, including private companies. They do, however, provide consumers of knowledge the opportunity to dispute, assess, or revise information that others review. In reality, pursuing wider circulation of information alone can be problematic, creating an echo chamber in which information is distributed more but disputed less.
In a paper published in 2024, philosopher Nora Freya Lindemann examined a Bing chatbot’s response to a request for ten famous philosophers. The response named no living philosophers, women, or philosophers from Africa or the Americas. When prompted on a conventional search engine, though, the results returned far broader lists of philosophers, including from a wide range of backgrounds and cultures. While this single example does not establish how all chatbots respond, it demonstrates how an answer, though not demonstrably false in listing philosophers, offers a narrow and oversimplified account of the knowledge available.
Reductive answers are inevitable, including in knowledge commons. But within a functioning commons, this information can be contested, and the voice of the few can be recognized.
Reductive answers are inevitable, including in knowledge commons. But within a functioning commons, this information can be contested, and the voice of the few can be recognized. A chatbot user may request an alternate perspective, but due to the sheer volume of responses, an AI revising its answer in one conversation does not create a public correction. This means that later readers may tend to see the original response if a dispute is not significant enough to catalyze change in an AI’s internal decision-making.
Given this, whether readers visit the underlying sources of information is clearly important. Pew Research Center’s analysis of browsing data from 900 US adults found that users clicked a conventional search result on 8% of Google visits featuring an AI summary, compared with 15% without one. Though correlation is not causation, the study shows how rarely users follow the sources attached to information, and how much less often users clicked on conventional search results when granted an AI summary.
A source citation can credit a knowledge commons without bringing a user into contact with its discussions and primary community.
A source citation can credit a knowledge commons without bringing a user into contact with its discussions and primary community. If readers increasingly encounter a community’s knowledge production through AI, contributors may need to do more than broaden their shared resources if AI is to be the primary distributor of a commons. The commons would also have to persuade an external company to investigate the resulting answers. Their ability to resolve a misrepresentation of their work, then, depends partly on a process in which they have no say.
To mitigate AI’s adverse impacts on the knowledge commons, several changes could be made at both the institutional and organizational levels. Initiatives should primarily focus on slowly tipping the power balance to give communities meaningful tools to negotiate with companies and pursue their goals collectively.
AI providers could be required to establish a public review process if information on their chatbots repeatedly misrepresents its source. A community would submit empirical examples that show a service overlooking relevant evidence or excluding a documented point of view. The provider would then publish its response to the request, explaining whether it adjusted the information correctly or otherwise. Unlike a singular submission with isolated feedback from a user, this process would allow others to examine the evidence directly. However, a community complaint should not automatically bring about change. This is because a well-funded organization could otherwise overwhelm the process or use it to suppress alternate interpretations that it opposes. Thus, independent review and input from smaller communities would be required to make this process sound.
Additionally, commons organizations should collectively negotiate the terms of any partnerships with private corporations, especially those centered on AI. Representatives of commons organizations could seek funding to maintain their platform and a formal commitment to investigate recurring chatbot problems. Negotiating together across multiple organizations could give smaller communities more leverage and access to expertise they could not obtain individually.
With that said, companies may have little incentive to negotiate if they can obtain information without accepting the obligations of communities. Organizing collectively does not, by itself, give communities more bargaining power than mega corporations. Negotiations would have to concern something that goes beyond a community’s permission to reuse open material. In order to enforce responsibilities that companies must assume when they become major distributors of material, a legal duty to negotiate with public information communities may be required (if the company is a major distributor of that information). Communities could then negotiate enforceable commitments on how large corporations should investigate misrepresentations of their information. In doing so, the bargaining position that communities would have would derive from the recognized right to interact with institutions handling their information, rather than the threat of withholding information that is intended to remain open to all.
The future of knowledge commons depends on communities, youth, and ordinary people’s commitment to question and contribute to information. AI can make information more accessible to audiences—ones that knowledge commons may not have otherwise reached—but the value of this expanded access is also related to whether that information is challenged or engaged with in the spirit of commons.
For young people like myself, inheriting knowledge should include the opportunity to reconsider and ponder what we consume.
For young people like myself, inheriting knowledge should include the opportunity to reconsider and ponder what we consume. The state of a commons rests on us being able to contribute to knowledge that we could not otherwise produce alone. As AI expands evermore, we must preserve the opportunity to question, challenge, and wonder.
