关键词: 领导力, 人机协作, 归因过程, 伦理, 绩效

Abstract: Human-AI collaboration is increasingly initiated from the bottom up, as employees independently incorporate artificial intelligence into everyday work before organizations fully institutionalize its use. This makes leaders’ interpretations and responses especially consequential: leaders can legitimize, support, constrain, or even undermine subordinates’ human-AI collaboration. Yet existing research has largely adopted an actor-centered perspective and examined how collaboration affects employees themselves, leaving limited understanding of how leaders, as salient observers and decision makers, react to subordinates’ collaboration with AI. To address this issue, this research develops a leader-centered framework grounded in dual-system processing theory and examines why subordinates’ human-AI collaboration can elicit qualitatively different leader responses and how these responses shape the sustainability of collaboration over time.
The central theoretical argument is that leaders process subordinates’ human-AI collaboration through two distinct but potentially coexisting routes. The first is a deliberative, “cold” processing route in which leaders engage in relatively effortful causal reasoning and future-oriented evaluation. The second is an automatic, “hot” processing route in which collaboration activates intuitive threat perceptions and affective reactions. By integrating these two routes, this study moves beyond explanations that assume leaders respond to AI-enabled employees in a uniformly favorable or unfavorable manner. Instead, it proposes that the same collaborative behavior can generate constructive developmental responses, constructive change-oriented responses, or destructive retaliatory responses depending on the psychological process activated.
Study 1 investigates the cold-processing route through two complementary mechanisms. Study 1a focuses on leaders’ causal attributions regarding why subordinates collaborate with AI. Leaders may interpret such collaboration as driven by performance-improvement motives or, alternatively, by impression-management motives. These contrasting attributions are expected to shape leaders’ willingness to invest in employees’ future development, including behaviors such as empowerment, developmental support, and promotion-oriented decisions. Study 1b shifts from motive attribution to future-oriented appraisal. Leaders may evaluate whether subordinates’ human-AI collaboration enhances the team’s future competitive position and adaptive capacity. When collaboration is viewed as strengthening the team’s long-term standing, leaders should be more likely to display change-oriented behaviors, such as supporting experimentation, innovation, and work-process transformation. Together, these studies show that leaders do not merely judge current performance; they also infer motives and anticipate longer-term consequences.
Study 2 examines the hot-processing route by theorizing that AI-enabled subordinates can alter perceived status, influence, and control relationships between leaders and followers. When subordinates become more capable through AI, leaders may experience relative deprivation because their own comparative advantage or managerial influence appears diminished. This perceived loss can trigger a defensive stress reaction and motivate retaliatory incivility toward the collaborating employee. This mechanism extends current human-AI collaboration research by showing that leader resistance may arise not only from rational concerns about technology, accuracy, or governance, but also from self-relevant social comparison and affective threat.
Study 3 integrates the constructive and destructive leader responses identified in the first two studies and adopts a dynamic perspective. Rather than treating human-AI collaboration as a one-shot behavior, it examines whether leaders’ responses reinforce or erode employees’ willingness and ability to sustain such collaboration over time. Development-oriented and change-oriented leader behaviors are expected to strengthen the continuity and adaptive quality of human-AI collaboration, whereas retaliatory incivility should undermine its sustainability. This longitudinal comparison directly addresses calls for research on the dynamic trajectory of human-AI collaboration and connects leaders’ immediate reactions to longer-term organizational digital transformation.
This research offers three main contributions. First, it shifts the level of analysis from employees as AI users to leaders as observers, evaluators, and resource allocators, thereby identifying leadership as a critical social mechanism through which bottom-up AI use becomes either institutionalized or suppressed. Second, it introduces a dual-process explanation for leaders’ heterogeneous reactions, integrating rational attribution and future-oriented cognition with intuitive, threat-based affective processing within one framework. Third, it advances a dynamic view of human-AI collaboration by linking distinct leader responses to the long-term sustainability of employee collaboration with AI. Overall, the research explains not simply whether leaders support or resist human-AI collaboration, but why different responses emerge and which responses are more likely to enable organizations to convert employee-initiated AI use into sustained capability development and digital transformation.

Key words: leadership, human-AI collaboration, attribution, ethics, performance