Rethinking Human-Centric AI Responses: Evaluating Support versus Constructive Challenge for Sustainable Human-AI Interaction

Human-AI interactions have become an integral part of daily life, influencing how people solve problems and make decisions. However, AI’s tendency to blindly agree with users can reinforce misconceptions, biases, and discourage critical thinking. While previous studies have focused mainly on AI performance and user satisfaction, less attention has been given to how constant AI support affects human thinking. This study examines whether human-centric AI should always support users’ perceptions or respectfully challenge them when necessary. A mixed-methods approach was used, combining prompt-response analysis with a survey measuring trust, perceived helpfulness, and critical thinking. The findings show that constant agreement can create an “echo chamber,” reinforcing personal biases, while respectful challenges encourage users to reconsider their beliefs and think more deeply. Three response types were identified: general agreement, constant support, and polite counterargument. The study concludes that human-centric AI should balance emotional support with constructive challenges to promote critical thinking and better decision-making.