This study analyzes the formation process of bubbles in a social context, focusing on generative AI. Bubbles arise when an asset's price exceeds its fundamental value, driven by herd mentality and irrational behavior. The risk is particularly high for new technologies due to excessive expectations and optimism. However, bubbles not only have negative consequences but also foster innovation and industrial growth through investment and public support. Therefore, understanding and diagnosing the bubble formation process is crucial. This study proposes and empirically verifies a theoretical model of the bubble formation feedback loop(external value→driving value→behavioral intention→external value), incorporating the moderating effects of personal innovativeness and social interaction. Additionally, a framework for diagnosing the bubble formation process is presented. A survey was conducted in August 2024 among 300 adults aged 20 and older. The model fit was analyzed using the Lavaan package in R, while path analysis and multiple additive moderation effects were examined using SPSS 24.0 and SPSS Process Macro 2. The key findings are as follows: First, the generative AI bubble feedback loop was significantly confirmed. External value positively influenced driving value (β=.835, p < .001), driving value influenced behavioral intention (β=.403, p < .001), and behavioral intention affected external value (β=.532, p < .001). Second, personal innovativeness and social interaction acted as moderating variables, independently affecting each pathway. Personal innovativeness strengthened the link between external and driving values (β=.055, p < .1), while social interaction weakened it (β=-.082, p < .01). Additionally, social interaction reduced the effect of driving value on behavioral intention (β=-.078, p < .05) but strengthened the impact of external value on behavioral intention (β= .074, p < .05). This suggests that public perception at the time mitigated the bubble effect. Third, the explanatory power of the multiple moderation model (with both moderators) was significantly higher than the single-moderator model. Using the Johnson-Neyman technique, the effect of external value on driving value was strongest when personal innovativeness was high and social interaction was low (β=.661, p < .001). The effect of driving value on behavioral intention was strongest when both personal innovativeness and social interaction were low (β=.511, p < .001), while the effect of behavioral intention on external value was highest when personal innovativeness was low and social interaction was high (β=.706, p < .001). This study contributes by proposing a feedback loop model for diagnosing and predicting bubble formation and presenting a methodology for managing technology bubbles. It serves as a practical tool for balancing expectations and concerns about new technologies while understanding their social impact.