Intention mining through human-computer interaction (HCI) has been studied in various domains such as psychology, health, communication, and transportation. Unlike previous studies focusing solely on recognizing intentions, our work introduces a novel HCI-based approach to reshape recognized human intentions into desired ones defined by the computer. We address the psychological challenge of modifying human intentions, which depend on multiple psychobiological variables. To fill this research gap, we propose the Intention Risk and Stimulus factor Impact (IRSI) framework, which enables intention reshaping by considering the emotional effects of shape and color in proactive visuals, the risk status of intentions, and human habituation to stimuli. Our system operates in two phases: (1) intention recognition through deep learning-based mining, and (2) reshaping of these intentions into new, non-premeditated ones via robotic stimuli. As an experimental setup, we employ a computer-based bluff card game that allows both mining and reshaping of player intentions. The system analyzes recorded bluff sessions to generate intention matrices and trains a CNN-based model for recognizing bluff-related moves. During gameplay, the robotic interface delivers adaptive proactive stimuli based on risk levels to psychologically influence and reshape the player's intentions. Experimental results demonstrate the effectiveness of our IRSI-based HCI system in achieving successful intention reshaping performance.