Amy Affelt
Slides at http://conferences.infotoday.com/documents/293/A103_Affelt.pptx
Fake news. Shared knowing it’s fake, shared without reading page linked to.
Facebook said it was crazy to think it had anything to do with influencing U.S. election. Then said maybe it did. Has some remedies, but they are much too slow.
Google has automatic links to face-checking sites. But are those correct? Or consistent?
CNN pointed to an IFLA document for spotting fake news. https://www.ifla.org/node/11175
Look at domain names. Read “about us” pages.
But respectable sites aren’t always correct. CBS News said Tom Petty died before it was true. Look for supporting sources. One source may not be enough.
Read about the author.
Check the date. Is it old? Is it April 1?
Check your bias. Does the site have a political slant?
Watch out for provocative headlines.
Watch for the “promoted” label.
Fake health news: Check source, plausibility. “The secret doctors won’t tell you.” Look for peer-reviewed article, human trials. HealthNewsReview.org debunks fake health news. Google it for other reactions.
Melissa Zimdars’ list of dubious news sources.
How to Lie with Statistics. Book from 1950s showing misleading graphics.
Fake videos coming next.
Tom Reamy:
Presentation:
http://conferences.infotoday.com/documents/293/A103_Reamy.ppt
Text analytics, taxonomy, training.
Book, “Deep Text”
There is no single definition of “fake news.”
Fake people, automatic bots (e.g., on Twitter)
Google: can be manipulated so top stories on a topic are fake news.
Two drivers: making money and manipulating people.
People make up news to get clicks, which get ad revenue.
Debunking: a fraction of the people who saw the original post will see the debunking one. No money in it. Effects linger.
Can block ads on a site, but that does nothing for politically motivated fake news.
Technical tools to finding misleading domain names, etc.
Automated systems aren’t smart enough and can be manipulated.
Case study: hybrid analysis of news
Inxight Smart Discovery (now SAP), multiple taxonomies.
Pulled in thousands of news stories, used rules to categorize them.
Faster than human review, smarter than automatic solutions.
Weight if word occurs in title, etc. There is no such thing as unstructured text.
Pronoun analysis: “The Secret Life of Pronouns.” All the words that search engines throw out. Analyzed e-mail and could often tell age, gender, power status of writer. Lying and fraud detection: fewer and shorter words, more positive emotion words. C. 76% accuracy.
Deep text solutions:
1. Database of known sites.
2. Deep learning of text/linguistic patterns
3. Flexible categorization rules
Fake news is a serious problem: undermines democracy, communication, civilization.
Multiple factors: multiple solutions
Hybrid human-machine solutions are getting best results. Ultimate solution is better education.
Books:
“Weaponized lies”
“Post-truth”
“Don’t Think of an Elephant”
Libraries can keep pushing information about how to spot fake news. Colleges and universities can teach critical thinking.
Take a minute to check it out. You don’t want to be “first but wrong.”
Water Librarian's Blog 