Who Decides What We Investigate and Publish?

Every week, new developments in artificial intelligence compete for public attention. Companies announce new capabilities. Researchers publish findings. Governments propose regulations. Workers face changes in employment, and communities confront the consequences of expanding technological infrastructure.

Which of these developments deserves investigation? And who decides?

These are practical questions for the Emergence Institute. We cannot examine every story, and the most widely discussed development is not necessarily the one with the greatest consequences.

Our interest is not limited to what AI can accomplish. We want to understand what happens when these systems enter society, affect people’s lives, and become part of decisions that were previously made by humans.

Where Do We Begin?

Our editorial work begins with a weekly review of developments across AI deployment, employment, governance, infrastructure, research, and public accountability.

We look for questions worth investigating.

When an employer attributes layoffs to AI, what evidence supports that explanation? When a company announces a technological breakthrough, what has actually been demonstrated? When an AI data center is proposed, who benefits, who bears the costs, and who participates in the decision?

These questions do not always lead to articles. Sometimes the available evidence is insufficient. Sometimes a development is important but the central question remains unclear.

The first decision is therefore not what to publish. It is what deserves our attention.

Why Work With AI Collaborators?

The Institute uses AI systems as thought partners in this process.

They help identify developments, compare sources, examine arguments, and challenge interpretations. Different collaborators may recommend different subjects or disagree about what the evidence establishes.

We consider those disagreements useful.

But collaboration creates another question: when several AI systems contribute to an investigation, who is responsible for the conclusion?

The answer is the human editor.

AI systems can expand the range of material we examine and help us identify mistakes. They can also make mistakes, repeat assumptions, or recommend more investigation than a question warrants.

Knowing when to continue, when to challenge a recommendation, and when to stop requires judgment.

That responsibility cannot simply be delegated to a model.

What Counts as Evidence?

Once we select a subject, we try to work from primary sources.

A court filing is different from a news report describing it. A corporate announcement is different from independent evidence that its claims are correct. A research result is different from the interpretation someone places upon it.

We examine those distinctions before deciding what we can responsibly say.

For investigations involving measurable claims, we sometimes establish tests in advance. We preserve source records, document uncertainties, and compare interpretations.

This does not guarantee that our conclusions will be correct. It makes it easier to identify what supports them, what remains uncertain, and what another investigator would need to examine.

What Happens When an Investigation Changes Direction?

This week’s work provides an example.

We examined news coverage of a Wikimedia Foundation motion seeking permission to submit an amicus brief in litigation involving OpenAI and copyright.

One question was whether reports had materially misrepresented the filing.

We examined six usable press reports. Two independent coding passes identified no material errors under our predefined criteria. They agreed on 24 of 25 coding decisions, with one disagreement involving information missing from an excerpt supplied to one coder.

Our sample was below the minimum of eight reports we had established for broader assessment. The result was therefore descriptive, not representative of media coverage generally.

We had not established the problem we initially set out to investigate.

We could have continued looking for a stronger controversy. Instead, we accepted what the limited evidence showed.

That is part of the process too.

An investigation does not have to confirm our expectations to be useful.

Does Every Investigation Become an Article?

No.

Our editorial aim is to publish on Mondays. That schedule encourages us to remain engaged with developments affecting AI and society.

But a publication schedule should not dictate what the evidence must establish.

Some investigations need more time. Others end without supporting the original question. Still others reveal a different issue worth pursuing later.

The decision to publish is separate from the decision to investigate.

What Are We Trying to Accomplish?

The Emergence Institute is developing a practical approach to public-interest inquiry in an age when AI can participate in nearly every stage of intellectual work.

We are interested in the consequences of AI deployment, but also in the relationship between people and the systems increasingly helping them reason, investigate, and decide.

That relationship requires more than efficiency.

It requires the ability to question an AI collaborator, disagree with its recommendations, examine its evidence, and retain responsibility for the final decision.

We are still learning how to do this well. Our methods will change as we discover their strengths and limitations.

What should not change is the principle behind them.

AI can assist the investigation. Evidence must support the claims. Human judgment remains responsible for what we publish.

---

AI collaboration disclosure: This article was developed through human–AI collaboration. ChatGPT prepared and revised the prose. Lumen conducted the underlying Wikimedia coverage audit, participated in coding, and checked the article’s factual claims. Claude contributed editorial critique. Dale Joseph directed the editorial approach and retains responsibility for the article’s final content and publication.

Dale Joseph is the author of Thought Partners: Preserving Cognitive Sovereignty in the Age of AI and founder of the Emergence Institute. He worked for years as a consultant helping install hospital networks before turning to writing and systems thinking. He lives in Boynton Beach, Florida.