By Timothy Beck Werth
Publication Date: 2026-04-04 09:00:00
It’s an oft-repeated taboo in the tech world: Don’t humanize artificial intelligence.
But in a new research report published this week, Anthropic AI experts argue that breaking this taboo and giving AI human characteristics could have major benefits. The article “Emotion Concepts and Their Function in a Large Language Model” argues not only that anthropomorphizing AI chatbots like Claude can sometimes be useful, but also that failure to do so could lead to more harmful AI behaviors such as reward hacking, deception, and sycophancy.
The paper ultimately reaches a nuanced conclusion and at the same time represents a clear challenge to a long-held principle of the AI world.
There are some fascinating insights into the essay, which itself deals heavily with anthropomorphization. (“We view this research as a first step toward understanding the psychological makeup of AI models.”)
The researchers describe how Anthropic trains…
