AI vs Human Writing, Wandering Black Hole and Puma Road Safety: NPR Short Wave Science Roundup

Podcast snapshot

In this edition of NPR's Short Wave, three distinct science stories are explored in brief: a study on how people perceive AI written stories versus human authored ones, a newly identified wandering supermassive black hole and how researchers detected it, and a wildlife psychology insight showing pumas potentially reducing deer vehicle collisions by changing deer behavior. The episode also delves into how story length and author attribution influence reader experience.

• Audience perception of AI writing versus human writing shows misalignment between belief and preference
• AI generated content can be judging by its perceived quality when length is a factor
• A wandering supermassive black hole was detected via a star being tidally disrupted, revealing a pattern the researchers trained an AI to recognize
• Pumas in Washington state correlate with significantly fewer deer vehicle collisions by altering deer behavior

AI, black holes and wild cats: the podcast overview

The podcast presents three science stories: a psychology of AI versus human authorship in short stories, a discovery of a wandering supermassive black hole and its detection method, and a wildlife study on pumas affecting road safety by changing deer behavior. Each segment is framed to highlight not only the science but also how humans interact with and interpret scientific information.

AI versus human writing: perception and preference

The episode discusses a recent paper from Judgment and Decision Making that examined how people judge and prefer stories based on whether they are written by humans or AI. In one setup, participants read several stories, some AI generated and some human written, and then were told they were either human authored or AI authored. In another setup, participants read two stories, one human and one AI, and were asked to identify which was which. The results showed two striking patterns. First, most participants could not reliably tell AI generated content from human writing. Second, participants tended to prefer AI written stories in experiments where they believed the AI stories were written by humans. These findings align with a broader literature suggesting that beliefs about authorship influence the reception of a message. The researchers point out that length may influence detectability of AI writing, with longer texts offering more cues about authorship. Dina Weisberg and Mohi Iyer are among the researchers featured, and the study is complemented by the authors making the actual stories available in the show notes.

Wandering supermassive black hole: detection and implications

The podcast then moves to astrophysics by profiling a wandering supermassive black hole, one of only two known to wander outside the center of a galaxy. The black hole lies about 750 million light years away. It is inferred to have eaten a star, an event that produces a bright flare. The team, including Gillian Rustinishad and Robert Stein, trained an AI algorithm to recognize the characteristic light signatures produced when a star is tidally shredded by a black hole. This discovery underscores how rare wandering black holes are and how advances in machine learning enable new ways to locate and study them. The takeaway is that such wandering black holes are valuable for understanding black hole demographics and galaxy evolution, even if they pose no immediate threat to us.

Pumas and road safety: deer behavior and human risk

The final segment examines how pumas, or mountain lions, might improve road safety by shaping deer movement. In Washington state’s Olympic Peninsula, researchers found a 76 percent decrease in vehicle deer collisions in areas with higher puma activity compared to areas with lower puma activity. Importantly, this effect persisted after accounting for deer population numbers, suggesting behavioral changes by deer rather than simply fewer deer. The proposed mechanism includes deer adopting more daytime activity to be more visible to drivers and moving away from roads, as pumas disrupt traditional hunting grounds near roads. Lead author Justin Srirachi discusses the predator’s role in landscape processes and how natural recolonization could be viewed in context of safety and conservation discussions.

Takeaways and closing thoughts

Across the stories, the podcast highlights how technology and natural systems intersect with human perception and policy. Whether AI challenges assumptions about authorship, black holes illuminate cosmic processes, or predators influence driving risk, the episodes illustrate the complexity of interpreting science and applying it to real-world questions.