Wikipedia 4 min read

Wikipedia Runs on People Who Bother to Fix Things

Key takeaways

  • Wikipedia’s sources, edit histories, and discussions help readers scrutinize its claims.
  • AI answers may reduce opportunities for readers to notice errors and become editors.
  • Welcoming contributors takes effort from volunteers whose time is already limited.
  • AI editing tools should be judged partly by how much review work they save.

An AI answer can save you a trip to Wikipedia. It may also mean skipping the page where you would have spotted a mistake—and decided to fix it. If we were building Wikipedia from scratch today, how would we persuade people to keep showing up for that work?

The history behind a sentence matters

A Wikipedia article comes with more than its visible text. Its edit history records changes. Its discussion page can reveal disagreements over wording, sources, and interpretation.

Those records become especially useful when a subject is contested. A single word in an account of an event can carry considerable weight. Readers who question that word can look back at how it got there and examine the reasoning behind it.

That is a valuable kind of traceability. It gives readers somewhere to start when a polished sentence deserves a second look.

An open record does not guarantee accuracy. Errors can survive, and the people debating a topic may share the same blind spots. But the record leaves clues about what needs checking.

Rebuilding Wikipedia would mean recreating this capacity to challenge and revise its contents. A vast collection of articles still needs people willing to question a claim, inspect its evidence, and correct it.

A future editor often starts as a reader

Imagine someone correcting the year a neighborhood library opened. They track down an old brochure, compare it with the article, and add a citation.

For the reader, one number changes. For the contributor, that correction may represent an afternoon’s work.

The reward might be modest: their neighborhood is now described accurately. Another editor checks the change and thanks them. That small acknowledgment could be enough to make the contributor return.

Now imagine readers getting the same information through an AI answer without opening the article. They may have fewer opportunities to notice a mistake in the underlying page. Reading and repairing the shared record become further separated.

This does not establish that AI answers will cause contributor numbers to fall. It does raise a question about how new editors find their way in.

Alongside page views, a knowledge community needs to pay attention to the path from reader to contributor. That path can begin with a very ordinary thought: I know something about this, and I can help.

The first edit has to feel worth making

An obvious edit button helps. What happens after someone clicks it matters just as much.

Suppose a newcomer makes a correction and sees it reversed without explanation. An experienced editor may consider the reason obvious. The newcomer may conclude that their help is unwelcome.

A specific explanation changes the experience. Show them what kind of source the claim needs, and they have a practical next step. They can learn the rule and try again.

But that explanation takes somebody’s time. Established editors are volunteers too. Asking them to welcome every newcomer, explain every policy, and inspect every contribution can make participation more exhausting.

Any attempt to rebuild a collaborative encyclopedia has to account for this cost of participation. It includes the time spent correcting bad edits, resolving disputes, and explaining the same requirement again.

A sustainable community needs ways to make a first contribution rewarding while keeping experienced contributors willing to stay. Opening the door is only the beginning.

Judge AI by the work left for reviewers

AI could help with some of that effort. It could explain editing rules in plain language or summarize what changed between two versions of an article.

Those summaries need to be checkable. Suggested sources need scrutiny too: a link beside a sentence does not establish that the linked material supports the claim.

Drafting assistance deserves the same accounting. Producing more text creates more material for someone to review. A tool that saves a contributor ten minutes of writing could still leave the community with extra work if its suggestions require lengthy checking.

The useful question is how easily an editor can verify, correct, or reject a suggestion. Showing the relevant passage from a source could help. So could flagging uncertainty and making the reasoning behind a proposed edit clear.

Those features would give reviewers a better basis for judgment. The editor responsible for the final decision needs to understand what is being proposed and have a straightforward way to change it.

Building another Wikipedia would require people who feel that contributing their time is worthwhile. AI can make answers easier to obtain, while the work of checking and maintaining their foundations continues. The future of shared knowledge depends partly on whether the person who fixes that library’s opening date has a reason to come back.

Wikipedia Artificial Intelligence Collective Intelligence

Comments

    Loading comments...