Etiquette Podcast 2: Bot-ched Statistics?
Published:
🎥: “Listen on YouTube”
In this episode, we stress-test large language models on one everyday, high stakes question: how social media shapes relationships. We design the prompt in Gemini, then compare outputs from ChatGPT and DeepSeek, and discover that ChatGPT’s five “sources” included only one working link, with three empty citations and one 404. We talk about why citation hallucinations happen, how they can quietly skew conversations about trust, intimacy, and conflict online, and what “human in the loop” really looks like for students and researchers. Expect a friendly, evidence, first vibe, practical verification tips, and a reminder that confidence ≠accuracy.
Listen: