Through Time

Checking a claim: text, image, voice

How to check in minutes what a model told you, what a photo or video shows and whose voice that is, and when to stop checking.

A course for anyone who has to decide whether to believe what is in front of them: a chatbot’s answer, a quotation, a photo, a recording. You will learn the moves fact-checkers use, how to reach the original source and when to stop. Taking data and charts apart belongs to the course on numbers; here it is always one claim.

So far there is only this table of contents. Lessons appear one at a time, and we promise no dates. Feed of new lessons (Atom)

18 lessons, 4 parts

Part 1 What a model says

A model writes with the same confidence whether it knows or invents, so its tone tells you nothing.

  1. Smooth is not the same as true

    Hallucinations in practice: which parts of an answer to check first, and why a model’s polite confidence is no evidence.

    coming soon

  2. A citation to nothing

    Models invent sources, and people have already paid for trusting them: how to check in a minute whether a citation exists.

    coming soon

  3. A summary distorts

    A model’s summary of a story or a paper can distort the point: how to compare it quickly with the original.

    coming soon

  4. Can you ask the model itself

    What asking for a source really gives you, and why two machines agreeing is not a check.

    coming soon

Part 2 A fact-checker’s moves

Experienced checkers read a page less and look at what others say about it, and that can be learned.

  1. Reading sideways

    Fact-checkers quickly leave a page and check it in other tabs: how to learn to do the same.

    coming soon

  2. The four moves of SIFT

    Mike Caulfield’s method: stop, investigate the source, find better coverage and trace the claim to its origin, all in a couple of minutes.

    coming soon

  3. Who really said that

    How to get from a fine quotation to the text where it first appeared, and spot a clipped context or words pinned on someone famous.

    coming soon

  4. The number in the headline and the one in the source

    How to find the report behind a figure and compare it with the headline: same measure, same people; the data itself is for the course on numbers.

    coming soon

Part 3 Images, video, voice

First find out where a file came from, and only then look closely at it: that is faster and more reliable.

  1. A real photo with a false caption

    Old footage under a new caption: a reverse image search checks the context before you start examining pixels.

    coming soon

  2. The picture that moved the market

    A fake image of an explosion at the Pentagon spread from verified accounts: a lesson in where to look first.

    coming soon

  3. A detector is not a judge

    Why detectors of generated content get it wrong, and how to read their verdict when there is no other evidence.

    coming soon

  4. Where a file came from: C2PA

    What a signature on a file proves, why it gets lost in sharing, and why its absence proves nothing.

    coming soon

  5. A voice on a recording

    How to check a public recording: who posted it first and whether a full version exists; a scammer’s cloned call belongs to the course on security.

    coming soon

  6. The liar’s dividend

    Once anything can be faked, a real recording is easy to dismiss as a fake too, so checking has to run both ways.

    coming soon

Part 4 When to stop

Nobody can check everything, so deciding what to check and how deeply is a skill too.

  1. How much checking is enough

    The cost of being wrong sets the depth: reading and forgetting is one thing, forwarding to family or deciding about money or treatment another.

    coming soon

  2. The skill of not reading

    Attention is limited,, not every claim deserves to be checked: how to decide what you can simply not read.

    coming soon

  3. “I do not know” as a result

    How to note what you checked and what you did not, and how to correct yourself when a falsehood has already gone out from you.

    coming soon

  4. A look ahead: checking versus generating

    What changes if signed files become normal and fakes get cheaper still, and which habits you will need whichever way it goes.

    coming soon