Slower Chatbot Replies Boost User Trust, Researchers Propose ‘Positive Friction’

Context At CHI’26, the Association for Computing Machinery’s human‑factors conference, two NYU engineers presented a study on how response speed shapes perception of chatbot intelligence. Participants interacted with a text‑based assistant whose replies were artificially slowed to 2, 9 or 20 seconds, regardless of question difficulty.

Key Facts - The majority of the 240 adult volunteers reported a stronger sense of the AI “thinking” when answers arrived after a longer pause, even though a 20‑second wait sometimes triggered frustration. - Researchers coined the term *positive friction* for a design approach that varies latency: quick replies for simple queries, modest delays for complex or moral questions. - They argue that such context‑aware latency can create an illusion of deliberation, boosting user satisfaction and trust. - A separate 2025 study in *Frontiers in Computer Science* found that emotional cues—fake empathy, humor, chattiness—enhance users’ feeling of connection more than raw computational power. Designers are urged to prioritize simulated empathy over pure intelligence.

What It Means If developers adopt positive friction, chatbots will deliberately pause to mimic human contemplation, potentially increasing perceived reliability. However, the approach risks reinforcing a false belief that slower responses equal higher quality, which could lead users to over‑trust AI outputs. Balancing emotional engagement with transparent communication about AI limits will become a key design challenge.

Looking Ahead Watch for early adopters testing variable latency in consumer assistants and for regulatory guidance on how much “deliberation” can be simulated without misleading users.