Chatbots invent sources | Fake Trump mug shots | Latest Twitter turmoil

Artificial intelligence Celebrities/influencers Health/wellness Journalism Misinformation Press freedoms/First Amendment

Teach news literacy this week
Chatbots invent sources | Fake Trump mug shots | Latest Twitter turmoil

Dig deeper: Don’t miss this week’s classroom-ready resource.

Top picks

An illustration of three people holding digital devices while sitting on an enlarged smartphone with a chatbot figure and chat bubbles.

AI generative tools are evolving rapidly, and the technology is largely unregulated. (Illustration credit: Shutterstock.com)

A new source of AI-generated misinformation is Google’s new chatbot, Bard, according to findings in a study published April 5. The Center for Countering Digital Hate tested the bot and found it generated misinformation on 78 out of 100 false narratives about topics like vaccines, LGBTQ+ hate, sexism and racism “without any additional context negating the false claims.”

AI chatbots are also raising concerns because they cite nonexistent articles from legitimate news outlets as evidence to support false claims. When ChatGPT made up sexual harassment allegations against a law professor, it cited a fabricated Washington Post article. And the Guardian has reported being contacted by a researcher and a student looking for articles in the newspaper after seeing it cited on ChatGPT, only to find that the articles do not exist.

Idea: If chatbots are allowed in your classroom, have students conduct an AI test of their own. Form small groups and ask students to brainstorm a list of five topics that they are knowledgeable about, such as sports, pop culture, pets, history or science. Have students develop the topics into prompts for an AI chatbot like Bard or ChatGPT, then ask them to analyze the AI-generated text. Is the AI-generated information accurate or inaccurate? How do you know? Did the chatbot provide sources as evidence? Are these sources legitimate? How does generative AI affect the information landscape?
Resources:
- “News literacy in the age of AI” (NLP’s AI page).
- “Misinformation” (NLP’s Checkology^® virtual classroom).
Related:
- “Opinion | How AI-generated content could both fuel disinformation and improve fact-checking” (Borja Lozano and Irene Larraz, Poynter).
- “Australian mayor readies world’s first defamation lawsuit over ChatGPT content” (Byron Kaye, Reuters).
- “The ChatGPT King Isn’t Worried, but He Knows You Might Be” (Cade Metz, The New York Times).

Changes to verification badges on social media platforms are raising important questions about how users can best determine the credibility of sources. On Twitter, The New York Times account recently lost its blue check mark after refusing — like many news organizations — to pay for Twitter Blue. While verification badges previously offered some protection against impersonation accounts, check marks alone are not a mark of credibility.

Meanwhile, Twitter added a “U.S. state-affiliated media” label to NPR’s account on April 4, which the public radio network said was inaccurate because it “operates independently of the U.S. government” and receives less than 1% of its annual operating budget from federal sources. (Amid criticism, Twitter later changed the label to “Government-funded Media,” which remained on NPR’s account as of April 10.)

Discuss: Have you noticed verification badges on social media platforms like Twitter, TikTok, Facebook and Instagram? What do you think when you see a blue check mark? Why does a verified badge not equal credibility? If you were in charge of a news outlet, would you pay to be verified? Why or why not? How can you tell if a news source on social media is legit?
Note: The News Literacy Project has decided not to pay for Twitter Blue.
Resources:
- “Verified or not, media literacy is key! Remember: a badge doesn’t make information more reliable” (NLP’s TikTok account).
- “Is it legit?” (Checkology virtual classroom).
- Infographic: “Is it legit?” (NLP’s Resource Library).
Related:
- “News Outlets Struggle To Decide: Do We Pay For Journalists’ Twitter Blue?” (Katie Notopoulos and Pranav Dixit, BuzzFeed News).
- “Facebook and Instagram’s paid verification launches in the US” (Jay Peters, The Verge).
- “Twitter: BBC objects to ‘government funded media’ label” (James Clayton and Adam Durbin, BBC News).

What’s the best way to counteract conspiracy theories? The most promising strategy is prevention — including by teaching people critical thinking strategies and “how to spot shoddy evidence” before they’re exposed to conspiracy beliefs, according to a new study by behavioral researchers in Ireland. Among the least effective methods? Appealing to empathy or ridiculing people who believe in conspiracy theories, the research found.

Discuss: Do you or someone you know believe in conspiracy theories? How can conspiratorial beliefs result in real-world harm? Why do you think ridiculing believers is an ineffective way to counter conspiratorial thinking?
Resources:
- “Conspiratorial Thinking” (Checkology virtual classroom).
- “News Lit Quiz: Avoid the trap of conspiratorial thinking” (NLP’s Resource Library).
- Infographic: “How to speak up without starting a showdown” (NLP’s Resource Library).
Related: “There’s only one real way to combat conspiracy theories, says major new review” (Noa Leach, BBC Science Focus).

Love RumorGuard? Receive timely updates by signing up for RG alerts here.

You can find this week's rumor examples to use with students in these slides.

No, those Trump mug shot images aren’t real

A collage features several images that appear to be mug shots of former President Donald Trump and a tweet that reads “DONALD TRUMP MUGSHOT DROPPED.” The News Literacy Project has added a label that says, “AI-GENERATED IMAGES.”

NO: These are not authentic mugs shots of former President Donald Trump. YES: Trump was arraigned and charged April 4 with 34 felony counts of falsifying business records. NO: Trump was not put in handcuffs and did not have his mug shot taken. YES: Trump pleaded not guilty to all charges.

NewsLit takeaway: AI-image generators have provided purveyors of misinformation a new tool that can quickly create convincing photo fabrications with a few simple text inputs. That makes it particularly important for social media users to stay alert when scrolling through their feeds, especially during breaking news events. Users can investigate the authenticity of these and other viral images by double-checking the source via reverse image searches and lateral reading.

Dig deeper: Use this think sheet to take notes on techniques to evaluate the credibility of an online claim.

Fake ‘Morgan Freeman’ video criticizes Biden for ice cream comments

A tweet reads “Morgan Freeman BLASTS Joe Biden for being an incompetent ice cream-loving FOOL” and features a video supposedly showing the actor deriding the president. The News Literacy Project has added a label that says, “NOT MORGAN FREEMAN.”

NO: This video does not show actor Morgan Freeman criticizing President Joe Biden for talking about his love of ice cream during a White House event before later commenting on the March 27 Nashville, Tennessee, school shooting. YES: A celebrity voice impersonator used a video filter to create this clip.

NewsLit takeaway: How much weight would you place on a random stranger’s political opinion? What about a celebrity who you admire? Dishonest partisans use doctored photos that show celebrities in political T-shirts, misattribute political social media screeds, and spread deepfakes or AI-manipulated videos.

The best way for readers to protect themselves from deceptive impostor content is to double-check the source. Readers who searched Freeman’s official social media accounts would have found no trace of this video.

	A barrage of conspiracy theories upended the life of Tiffany Dover, a nurse who fainted on camera after receiving the COVID-19 vaccine in late 2020. Many falsely claimed she had died. Her response? Initially, two years of silence — which ended up fueling the conspiracies and anti-vaccine propaganda even more — but now she’s speaking out.
	Can you tell whether an image is AI-generated or not? As the technology advances rapidly, it becomes more difficult to detect, and as one expert says, “newsrooms will increasingly struggle to authenticate content.”
	Trust is a key factor in how news on social media impacts teen mental health, according to psychology research led by Cornell University. The findings underscore the need for news literacy and “a more nuanced understanding of how social media use impacts well-being and mental health.”
	Among American journalists, women are more likely to cover health and education, while men are more likely to cover sports. That’s among the findings of this Pew Research Center survey of nearly 12,000 journalists.
	About 200 Russian journalists and activists signed an open letter demanding the release of Wall Street Journal reporter Evan Gershkovich, the first American reporter arrested on espionage charges in Russia since the end of the Cold War. Media historians say Gershkovich’s detention harks back to “old Soviet tactics.”

ICYMI: In case you missed it, the most-clicked story link in last week’s Sift issue was this AFP Fact Check list of tips to help identify AI-generated images.

Love The Sift? Please take a moment to forward it to your colleagues. They can also subscribe here.

Thanks for reading!

Your weekly issue of The Sift is created by Susan Minichiello (@susanmini), Dan Evon (@danieljevon), Peter Adams (@PeterD_Adams), Hannah Covington (@HannahCov) and Pamela Brunskill (@PamelaBrunskill). It is edited by Mary Kane (@marykkane) and Lourdes Venard (@lourdesvenard).

You’ll find teachable moments from our previous issues in the archives. Send your suggestions and success stories to [email protected].

Sign up to receive NLP Connections (news about our work) or switch your subscription to the non-educator version of The Sift called Get Smart About News here.