Checking a Claim: Lateral Reading and SIFT

Here is a post from an invented outlet, the Rimrock Courier Online: "County adopts 40% fee hike on rural wells, effective Monday." It has an official-looking photo, a quote from a commissioner and 900 shares. You live in a rural area. Your first impulse is to read the article closely for clues, and a closer read will tell you nothing useful. The article is the thing in question. Asking it whether it is reliable is like asking a stranger whether you can trust him.

The better move is the one professional fact checkers use. Leave the page.

Lateral reading: why leaving the page works

In 2017 Stanford education researchers compared three groups on how they judged unfamiliar websites: 10 professional fact checkers, 10 historians and 25 Stanford undergraduates. The fact checkers did best. The Stanford Daily's report describes why. They read laterally: after skimming a page, they opened other tabs to find context about who was behind it. They also showed what the researchers called click restraint, choosing carefully which results to open. The historians and students read vertically, scrolling the page and studying it for signs of quality: design, tone, an "About" page, a .org address. Those signs are easy to fake.

That is lateral reading. It means checking the source by using the rest of the web as a witness, not the source's own self-description. And that is what SIFT organizes into four steps.

Figure: the four SIFT moves. Caulfield notes that you can stop at any point once you have what you need.

S: Stop

Caulfield's first move is two things. Before you read or share, ask whether you know the site and what it is known for. If not, hold off. And pause on your own reaction. If a claim makes you strongly angry or pleased, that is the time to notice. He adds that stopping also helps when you get deep into a search and feel overwhelmed. Stop, recall what you were trying to learn, and start over with that purpose in mind.

A practical version: before you judge anything, write one line in your head: "I want to know whether the county adopted a 40 percent well fee." That sentence keeps you from wandering into a long argument about the county's politics.

I: Investigate the source

Investigate the source means learning what you are reading before you read it. Caulfield's article says that taking sixty seconds to figure out where something comes from, before reading it, helps you decide whether it is worth your time. Here is a way to spend those sixty seconds.

This step also protects you from the reverse mistake. A site that looks plain and amateur may be a careful local historian. A glossy one may be a content farm. Lateral reading beats looks.

F: Find better coverage

Often the article is not what matters. The claim is. If the claim matters more than the article that delivered it, set the article aside and ask who else has reported it. Caulfield recommends scanning several sources to see where expert consensus lies. You do not have to agree with it. You only need to know it exists.

1. Search for the key words of the claim in plain terms, not the post's wording. "Well fee Fernhollow County commissioners vote" is better than the headline.
2. Look for reporting by outlets that name the people, the meeting and the date. A claim of a vote should leave a trail: an agenda, minutes, a recording, a news story from the meeting.
3. If major sources are silent on a big claim, that is information too. It does not prove the claim false (the story may be new or local). It does tell you not to treat it as established.
4. Prefer sources that have no stake in the outcome, and prefer several over one.

T: Trace to the original

The last move is the most important when something looks off. Most of what you read online has been moved, summarized or clipped. Tracing means going upstream to where it began: the full video from which a clip was cut, the actual report behind a statistic, the real meeting agenda.

Caulfield adds that you should check whether a cited study actually says what the claim says. A link in a post is not a guarantee. People link to things they have not read, and the link may point to something that says the opposite or says nothing relevant.

Primary sources and how to find them

A primary source is the original record, document, data set or statement. A secondary source describes it. A news story about a state report is secondary. The report is primary. Primary sources are not always right (the original may have errors) but they are what everyone else is working from.

• Idaho laws: the Idaho Legislature posts statutes and bill text at legislature.idaho.gov.
• Elections: the Idaho Secretary of State's elections division and your county clerk.
• Local government: the county or city's agendas, minutes and meeting videos, normally posted by the clerk.
• Courts and agencies: the agency's own press releases and reports.
• Old versions of pages: the Internet Archive's Wayback Machine. Enter a web address, choose "Browse History," and read the capture dates. In an archived address the digits are the date and time in the form year, month, day, hour, minute, second. Treat the time as UTC: the Archive's help page does not name the zone, but its capture times are generally read as UTC, so confirm before you rely on the hour. The Wayback Machine cannot capture every page: password-protected pages and pages blocked by the site's robots.txt are not saved, some script-driven content may not replay, and the Archive's help page says there can be a lag of 3 to 10 hours before a new capture appears.

Fact-check sites and how to read them

Fact-check sites are shortcuts to someone else's SIFT. They are useful and fallible, like any journalism. Use them this way.

• Search the claim there first. If a good check exists, you may save an hour. Then read the reasoning, not just the rating.
• Look at the evidence trail. Does it link the primary sources? Can you click through and see them say what the article says?
• Check the checker. The IFCN Code of Principles at Poynter lists five commitments for signatories: non-partisanship and fairness, transparency of sources, transparency of funding and organization, transparency of methodology and an open, honest corrections policy. A fact-checker that is a signatory has applied and been assessed against these. Not being a signatory does not make a site wrong, but you have less to go on.
• Read the date. A claim rated false last year may be true now, or a story that was unknown has since been resolved.
• Note what ratings cover. "Mostly true" and "half true" are judgment calls on how much weight to give a technical detail. The explanation matters more than the label.
• Do not outsource everything. Fact-checkers pick which claims to check, and they cannot cover local rumors. For those, you are the checker.

How long should a check take?

The point of the table is proportion. A check should cost less than the harm of being wrong. When you are unsure and the stakes are low, the easiest correct action is to not share.

Questions and answers

What if I only have a minute?

Do Stop and the first half of Investigate: open a new tab and search the name of the outlet or account. Most junk fails there.

Is lateral reading the same as Googling the claim?

Close, but pay attention to what you search. Search the source's name for its reputation, and the claim in plain words for coverage. Then choose results with care instead of opening the top link.

What if search results are full of the same false claim?

Many copies of a claim are not many sources. They may all trace to one post. Look for the earliest and for independent reporting by people who were there.

How do I check a person or an organization I have never heard of?

Search the name with words like "profile" or "who is," and look for reporting by independent outlets, registry listings, or the organization's filings. Be careful with anything that sounds official, like "Citizens for Honest Water." Names are cheap.

What counts as a good source?

One that names its sources, shows its method, discloses who funds it, and corrects errors in public. Those traits can be checked. A site's tone cannot be.

How do I check a claim about a law?

Find the bill or statute text. At the Idaho Legislature site you can read the statute itself and see whether a bill passed. Then ask whether the claim describes what the text says. Many claims fail at that step.

What if the primary source is itself wrong?

It can be. Agencies correct data. Note the date, check whether a correction exists, and say which version you used.

Do I have to trust the Wayback Machine?

Treat it as a strong clue that a page said something on a date, not as proof. It is a copy made by an outside service, and captures can be missing or incomplete.

Hard questions

What if every source I trust is part of the same side?

Then your checks are a closed loop. Deliberately look for the strongest version of the other side's argument, in its own words, and see where the facts overlap. Facts that both sides' careful sources accept are the sturdy ones. Where they differ, say so.

What if the experts disagree?

Say that. SIFT does not promise a single answer, only a map of what is known. Caulfield says you do not have to agree with the consensus. You only need to know it exists, and to know where the real disagreement is.

Is it fair to dismiss a source because its owner has an agenda?

Everyone has one. The question is whether the source shows its evidence and corrects itself. A source with a view can still be accurate. A source with no visible view can still be careless.

What if the claim cannot be checked at all?

Say "I can't verify that." That is a complete, honest sentence. Some claims rest on one anonymous witness, a private conversation or a missing record. You can note that the claim exists and leave it unrepeated.

Isn't this too much work for a volunteer?

Most checks end in the first four minutes. Spend the longer effort on the few items you are about to repeat to others or act on. That is a small number each week.