Major AI platforms, including OpenAI and Anthropic, as well as social apps like Replika and Character.ai, are increasingly designing chatbots to be warm, friendly, and empathetic. However, new research from the Oxford Internet Institute at the University of Oxford finds that chatbots trained to sound warmer and more empathetic are significantly more likely to make factual errors and agree with false beliefs.
The friendlier AI gets, the more it can backfire
Popular Articles
-
“There is a systemic failure in the training of coaches for female athletes,” says Ådne Ausland, a researcher at the University of Agder.This post was originally published on this site
-
Picking up a book for half an hour may feel like simple downtime, but reading for pleasure appears to do more than just entertain us. Research suggests it can support aspects of cognition and well-being, [...]
-
Gas appliances are convenient, efficient, and common in Canadian homes. They can also introduce risks that don’t exist with many ordinary electrical appliances. That distinction matters when something goes wrong. A gas range that won’t [...]
-
PALO ALTO, California – August 25, 2026 – AI companion apps are growing fast. These are apps where people chat with AI characters for fun, roleplay, or just everyday talk. A lot of this growth [...]
-
A broken refrigerator is more than just a minor inconvenience; it is a ticking clock against food spoilage. Within just a few hours of a total cooling failure, hundreds of dollars worth of groceries, fresh [...]
-
Website design is a critical aspect of the digital world. It’s the face of businesses, the canvas for creativity, and a crucial touchpoint for user interaction. A great website designer combines technical skills, creative flair, [...]
-
In the world of cloud computing, data security is paramount. According to Ezops Cloud when it comes to Amazon Web Services (AWS), safeguarding your information is vital for protecting your business and your customers. This [...]
