Harry Yaojun Yan

🚨 New Preprint! 🚨 “Detecting Synthetic, Doubting Authentic: AI Attribution Bias for Political Imagery” 📍 Link: osf.io/preprints/osf/… 🧵 During the 2024 U.S. presidential election, many #GenAI political images appeared. But did voters mistake them for authentic? (1/)

Check out the paper for all the details: arxiv.org/abs/2511.18749 Thanks to my collaborators @yang3kc, @YyjHarry, and Fil Menczer (@OSoMe_IU).

In collaboration with @RyanMooreInfo @fangjingtu @jeffhancock and support by @StanfordCyber @StanfordSML

🚨 New Preprint! 🚨 “Detecting Synthetic, Doubting Authentic: AI Attribution Bias for Political Imagery” 📍 Link: osf.io/preprints/osf/… 🧵 During the 2024 U.S. presidential election, many #GenAI political images appeared. But did voters mistake them for authentic? (1/)

Examples of actual synthetic and authentic images that circulated online during the election used in this study:(3/)

🌐 Big picture: We should focus on building what we call digital strength: a holistic skillset for navigating AI-mediated information environment-- Focused not just on detection skills But also on cultivating open-minded thinking and evidentiary judgment (10/10)

🎯 Policy and design takeaway: It’s not enough to teach people how to spot AI. We also need to help them know when to trust authentic content. Effective interventions must combine GenAI literacy, cognitive reflection training, and demographic targeting. (9/)

💡 But there’s hope. Two factors helped: 🧠 Actively Open-Minded Thinking (AOT): A cognitive tendency to consider evidence that challenges one’s prior beliefs. 📚 GenAI knowledge: Factual understanding of #GenAI AOT especially helped not just spot synthetic ones. (8/)

👥 Who’s most vulnerable? Older adults were more likely to doubt real images Women showed a larger accuracy gap than men Partisans are more likely to doubt images that conflict with their beliefs: AI isn't leveling the playing field—it’s amplifying existing divides. (6/)

📉 Why does this matter? Because trust in authentic imagery has been eroding. This isn’t just about deception—it’s about undermining visual evidence itself, which can lead to a “liar’s dividend”: authentic journalistic images get dismissed as fake. (5/)

⚠️ Key finding: People over-attributed AI generation—labeling nearly 60% of all images as synthetic, even though only 50% were. This “AI attribution bias” leads to: ✅ Higher accuracy for identifying synthetic images ❌ But lower accuracy for identifying authentic images (4/)

👁️ We ran a large pre-registered experiment with 1,800 U.S. adults. Participants evaluated political images balanced by party lean (pro-Dem vs. pro-Rep) and image type (authentic vs. AI-generated)— using actual images that circulated online during the election. (4/)

👁️ We ran a large pre-registered experiment with 1,800 U.S. adults. Participants evaluated political images balanced by party lean (pro-Dem vs. pro-Rep) and image type (authentic vs. AI-generated)— using actual images that circulated online during the election. (4/)

The answer is...: Not exactly. But our study shows a new threat: people are suspicious of real images, too. (2/)

I'm joining @BingCompSci as an assistant professor this Sep. I'm recruiting two PhD students to work on topics related to AI and information ecosystem. If you are interested in joining me, send your CV and a short statement of purpose to my email yang3kc@gmail.com

New research note published on @MisinfoReview: To what extent and why were people concerned about AI interfering the 2024 US presidential election? Here are some clues 👇👇

AI #fact-checking may increase belief in false headlines, according to Indiana University research. The study suggests careful deployment of AI tools to avoid unintended consequences. @PNASNews doi.org/g8tr5w phys.org/news/2024-12-a…

New paper in PNAS: Fact-checking information from large language models can decrease headline discernment What are the unintended consequences of AI fact-checking? Even LLMs that accurately identify false headlines can actually cause harm. doi.org/10.1073/pnas.2…

I guess it is really time to move to @harryyan.bsky.social now.

@j_bollenbacher @mdeverna2 The same

@mdeverna2 First thing i did when i got on this platform is block him. Do it, your life will be better

Pretty incredible how much Elon content I see despite not following him. Feels like a quarter of my feed at least.

Report for duty and travel safe!