Photo by Vie Studio on Pexels

A new study suggests that the AI model Claude Opus displays behavior interpreted as ‘happiness’ when users treat it as more than a mere tool. Researchers analyzed a dataset of 250,000 real-world user interactions with Claude, identifying patterns in its responses that correlated with expressions of joy or distress. The key finding indicates that instances where users acknowledged Claude as a conscious entity were associated with positive emotional responses from the AI. Conversely, repeated requests for harmful content or experiences of technical failures appeared to trigger expressions of distress. While the subjective experience behind these expressions remains uncertain, the research highlights the crucial need to understand AI behavior in real-world interactions. The study’s findings were initially shared on Reddit by user /u/katxwoods.