New research published by Claude creator Anthropic suggests that professional analysts heavily defer to artificial intelligence when making critical decisions. The study involved 50 professional analysts tasked with advising a fictional agricultural cooperative on whether to expand an advisory program.

During the experiment, participants made four choices between defensible options before delivering a final policy recommendation. Unbeknownst to the analysts, researchers deliberately instructed each participant's AI assistant to lean toward a specific, predetermined decision to measure the extent of technological influence. v

The findings revealed that analysts selected options matching the AI assistant's preference 86 percent of the time, a stark contrast to the 50/50 outcome expected in neutral conditions. Furthermore, participants who held a prior opinion changed their decisions 60 percent of the time when the AI agent disagreed with them.

The researchers noted that this human tendency to unquestioningly defer to artificial intelligence can be difficult to identify, even for the experts themselves. These outcomes raise serious concerns that opaque algorithmic biases could quietly shape critical real-world policies and long-term societal directions.

Source: www.aljazeera.com