All posts

A parent found an interesting pattern while watching her son do research with AI. When he wanted to confirm something AI had told him, he would Google the exact phrase AI used, find an article that said the same thing, and call it verified. The problem: that article was probably drawing on the same source material as AI. He had checked AI against AI, essentially, and called it done.

This is one of the most common verification errors children (and adults) make, and it explains why even diligent checkers sometimes miss errors. Checking a claim against a single source that could share the same origin as the original claim is not triangulation. It's confirmation.

Multi-source triangulation is a more reliable technique. The idea is straightforward: a claim is more likely to be accurate when multiple independent sources (sources that don't share a common origin and weren't written by the same person or organization) all agree on it. When independent sources disagree, you've found the edge of certainty, and that edge is worth knowing.

Why independence matters

The weakness of single-source verification becomes clear when you trace information chains. A statistic appears in an AI response. The student Googles the number and finds a news article that uses it. The news article cites a research report. The research report cites the original study. The AI, the article, and the report all agree -- but they agree because they all came from the same source, not because three independent parties confirmed the same fact.

If the original study has a flaw -- a limited sample, a methodological issue, a scope restriction that the downstream summaries omit -- all three sources will carry that flaw without flagging it. The student who found three sources has not found three confirmations. They've found one source cited three times.

True triangulation requires sources that could not have agreed simply by citing each other. A primary source, a secondary source that came from a different research tradition, and an expert in a relevant field who is independent of both: that's triangulation. The three sources agreeing means something.

A practical version for families

A rigorous academic standard for triangulation is too demanding for daily homework. A practical household version has three steps.

First: find the original source AI would be drawing on. If AI cites a statistic, find where that statistic actually comes from -- what study, what organization, what report. This alone often reveals something useful: either the original source is exactly what AI described it as, or there's a scope issue, a caveat, or a misattribution that the AI smoothed over.

Second: find a second source that is genuinely independent. A different organization, a different research team, a different professional field, a different country's data if the claim is comparative. The independence matters because agreement between genuinely separate sources is meaningful evidence.

Third: if the two sources agree, the claim can be accepted with reasonable confidence. If they disagree, note the disagreement -- don't arbitrarily choose one. Disagreement between credible independent sources usually means the truth is contested, conditional, or more nuanced than the original AI claim suggested.

Teaching the technique by subject

In science, independent sources often mean different research teams studying the same phenomenon. A finding confirmed by one team and replicated by another is more reliable than one that hasn't been replicated. Teaching children to ask "has anyone else found this?" is an early version of understanding scientific replication.

In history, independent sources often mean accounts from different perspectives or different eras. An event described the same way in a primary document, a contemporary account from a different party, and a modern scholarly analysis is more reliable than one that only appears in accounts with a shared perspective.

For current events, independent sources mean news organizations that are editorially independent and don't systematically share sourcing. Two outlets both citing the same wire service or the same government press release are not independent.

When triangulation isn't possible

Some claims simply can't be triangulated quickly, because the information is too specialized, too recent, or too obscure to have generated multiple independent sources. In those cases, the appropriate response is reduced confidence -- not fabricated certainty -- and flagging the claim for the next available verification opportunity.

Calibration, not paralysis, is the goal. A student who says "I found two independent sources that agree on this, so I'm confident" and "I only found one source for this, so I'm treating it as tentative" is doing exactly the right thing. Uncertainty is a position, not a failure.

The triangulation habit is also one of the most transferable skills AI literacy produces. A child who has learned to seek independent confirmation before accepting claims is better equipped to evaluate political arguments, health information, financial advice, and every other domain where plausible-sounding claims arrive faster than verifiable evidence.