How to Avoid Misleading Summaries

A summary misleads when a reader who sees only your post walks away believing something the article doesn't say. It rarely takes a lie. It usually takes one cut word, one rounded number or one sentence pulled out of the paragraph that limited it. The good news is that the failure modes are predictable, so you can check for them in under a minute.

Here are the seven we see most often, roughly in order of how often they happen.

1. The dropped hedge

"Coffee linked to longer life in large study" becomes "Coffee makes you live longer." The first is what the research found. The second is a causal claim nobody made.

Hedges are short words that carry a lot of meaning: may, could, linked to, in mice, in a small trial, early results, proposed, so far. When you're over the character limit, they look like easy cuts. They aren't. Cut adjectives, the outlet's name in the sentence (move it to the end), or the second detail instead.

The check: circle every hedge in the article's main claim. Is each one in your post, or replaced with an equivalent?

2. The swapped number

Numbers get distorted in a few standard ways. A relative change becomes an absolute one ("risk rose 50%" when it went from 2 in 1,000 to 3 in 1,000). A projection becomes a fact ("will cost $2 billion" when the article said "could cost up to"). A range becomes its top end. A percentage point becomes a percent.

The check: find the number in the article and read the full sentence it sits in. Make sure your version has the same unit, the same base and the same tense.

3. The missing "but"

Many articles state a finding and then qualify it in the next sentence: "The program cut absences by 12%. But the effect disappeared after the first year." Summaries that take only the first sentence are accurate word for word and wrong in meaning.

Extractive summarizers, including TweetSum, are especially prone to this because they score sentences one at a time. TweetSum lowers the score of sentences that open with "But", "However" and similar words, and strips those openers when it uses them, but it can't know what the following sentence takes back.

The check: read the sentence after the one you're summarizing. If it starts with but, however, still or yet, that's probably part of the claim.

4. The wrong speaker

Articles quote people who disagree. A line from a critic, a lawyer or a company spokesperson can be the most striking sentence in the piece, and when it lands in your post without attribution, it reads as the article's conclusion.

The check: for every quoted phrase in your summary, name the person who said it. If it's an advocate for one side, say so.

5. The headline swap

Headlines are written for clicks and space, often by someone other than the reporter. They overstate more often than the body does. Summarizing a headline instead of the article spreads that overstatement further.

The check: don't write the post until you've read past the third paragraph. If the body is more cautious than the headline, follow the body.

6. The scope creep

A survey of 1,000 adults in one country becomes "people". A pilot in three schools becomes "schools". A finding about one species, one city or one year becomes a general rule.

The check: find who or what the article actually studied or reported on. Put that in your post. "In a pilot at three Ohio schools" is 31 characters and keeps you honest.

7. The stale story

An article from four years ago, shared as if it were news, misleads even if every word in the summary is accurate. So does a story that has since been corrected or updated, if you summarize the original version from a screenshot or a cached copy.

The check: look at the date and any correction notice at the top or bottom of the article. If the story isn't recent, say when it's from.

Why this matters more for short posts

A long article has room to qualify. A 280-character post doesn't, and most people who see it won't click. A 2016 study by researchers at Columbia University and the French research institute Inria found that 59% of the news links shared on Twitter in their sample were never clicked at all. For most of the people who see your post, your post is the article.

Treat the summary as a small piece of reporting and give it the same care.

A one-minute test

Before you post, imagine a reader who sees only your post and then meets the author of the article. They say, "I read that you found X." Would the author agree, or would they wince and say "well, not exactly"?

If they'd wince, find the word you dropped. It's usually one of the seven above.

Fixing a summary after you post

Everyone gets one wrong eventually. When you do, correct it where people will see it. On most platforms that means replying to your own post with the correction, or deleting it and posting a new version that says what changed. "Correction: the study was in mice, not people" is short and earns more trust than the original post ever would have.

Quietly deleting a post that has already been shared leaves the screenshots and reposts out there with nothing to correct them. A visible correction travels with the conversation.

Questions