Abstract: Accompanying the rapid economic development and reforms in income distribution， Chinese residents have enjoyed a continuous growth in earnings， and the middle-income group has expanded. However， subjective class identification continues to be characterized by a downward bias， with the phenomenon of downward identification among the middle-income group being particularly pronounced. The gap between objective and subjective status remains a mystery. In parallel with the expansion of the middle-income group, the rapid development of artificial intelligence technology in recent years has triggered a change in productivity that has propelled us into the AI age. As AI progresses， the work tasks of many occupations are at risk of being replaced by the technology， which in turn has impacted the social mindset of the population. This paper adopts a risk perspective to reveal the mechanisms through which occupational risks， in the context of AI， affect subjective class identification. Our research finds that occupational risks have a significant downshift effect on class identification. Further analysis indicates that unemployment and income risk are two dual transmission mechanisms through which occupational risks influence subjective class identification. In addition， personal assets moderate the effect of occupational risks on the downward class identification. Workers with more recourses are less impacted by occupational risks on their class identification. The analysis of the middle- income group shows that， compared to low-income and high -income groups， the group faces higher occupational risks， resulting in a greater degree of downshift bias in class identification. This study explains the downward bias in subjective class identification from a risk perspective， offering a significant contribution to the traditional resource-based theories of class identification. Furthermore， our research addresses the question of the “power of era” that causes the lower subjective class identification of middle-income groups， providing important insights into understanding the trends in social stratification in the AI age.

Key words: AI age, occupational risks, subjective class identification, risk perspective, middle-income group