AI researchers want AI to fake "thinking" (www.machinesociety.ai)

Researchers from NYU Tandon School of Engineering have revealed intriguing insights into how users perceive the thinking capability of AI chatbots, suggesting that slower response times can lead to a greater trust in the AI's answers. Presented at the CHI’26 conference, the study involved 240 participants who rated answers from a chatbot that intentionally delayed responses by 2, 9, or 20 seconds. Findings indicated that users generally preferred the delayed responses, associating them with greater deliberation and thoughtfulness, despite the fact that these delays did not correlate with the complexity of the questions asked. This study's implications are significant for AI development, as it suggests that user satisfaction may be artificially enhanced through deliberate response delays, termed “Context-Aware Latency.” While this could optimize user interface design, it raises ethical concerns about fostering misconceptions surrounding AI, potentially leading users to anthropomorphize technology. The researchers caution that while slower responses might seem to signal quality thinking, this could inadvertently create a misguided trust in AI systems. The article argues for a more transparent approach in educating users about the nature of AI, advocating for rapid responses that reinforce AI’s role as a tool rather than a sentient being.