Abstract:

The rapid advancement of generative artificial intelligence (GenAI) has fueled an unprecedented surge of AI-generated content (AIGC) within the digital content economy. Accompanying this expansion, the large-scale diffusion of low-quality AIGC has given rise to a form of informational “pollution,” posing potential threats to platform credibility, user trust, and ecosystem sustainability. Although this phenomenon has attracted growing attention, its underlying psychological mechanisms and governance implications remain insufficiently theorized. Drawing on the psychological processes embedded in human-GenAI interaction and adopting a nested Stimulus-Organism-Response (SOR) framework, this study develops a multi-level theoretical account of AIGC pollution through a structured “cause-effect-intervention” logic across three interrelated sub-studies.

The first sub-study theorizes how GenAI’s technological affordances may paradoxically contribute to low-quality content production. Rather than assuming that technological advancement inherently enhances output quality, this study argues that quality variance is primarily rooted in users’ psychological responses during the content creation process. In a content economy characterized by accelerated production cycles and traffic-driven monetization incentives, GenAI’s core affordances—ease of use, rapid generation, and iterative refinement—can be conceptualized as salient technological stimuli. These stimuli are theorized to reshape contributors’ cognitive evaluations and motivational orientations. Specifically, the study proposes that GenAI’s efficiency may intensify users’ pursuit of instant gratification while simultaneously elevating perceived self-efficacy. The heightened orientation toward immediacy may reduce willingness to invest additional effort in content refinement, whereas inflated self-efficacy may decrease perceived necessity for deep cognitive engagement. Together, these psychological dynamics are theorized to diminish contributors’ cognitive effort in content development, thereby increasing the likelihood of insufficient deliberation and low-quality AIGC production. This sub-study establishes the micro-foundational organism mechanism that is proposed to initiate the pollution cycle.

Building on this foundation, the second sub-study conceptualizes low-quality AIGC, once disseminated, as a new environmental stimulus within the platform ecosystem. The framework theorizes its effects on two key actors in the content economy: content audiences and other contributors. For audiences, exposure to low-quality AIGC may violate technological expectations regarding GenAI’s reliability as well as social expectations concerning creators’ responsibility and value orientation. Such expectation violations are theorized to generate cognitive dissonance, which may subsequently influence information processing patterns, engagement intentions, and consumption decisions. These psychological and behavioral adjustments are expected to shape downstream outcomes in the content economy, particularly in marketing performance. For other contributors, the presence of low-quality AIGC may function as a social signal that activates perceived identity threat and perceptions of platform unfairness. These threat perceptions are theorized to influence subsequent content creation strategies, potentially leading contributors either to increase effort to differentiate their work or to conform to prevailing low-effort norms. The model further proposes that these responses are contingent upon contributors’ motivational orientations, such that intrinsic motivation is likely to encourage differentiation through higher cognitive investment, whereas extrinsic motivation may increase susceptibility to conformity.

Recognizing the systemic implications of AIGC pollution, the third sub-study conceptualizes platform governance mechanisms as external stimuli capable of reshaping contributor cognition and behavior. First, official GenAI tools are theorized as guidance-based nudges. Compared with third-party tools primarily optimized for efficiency, official tools may be designed to scaffold structured creation and encourage more deliberate engagement. By embedding nudging mechanisms such as default quality templates, task simplification, and exemplary priming, platforms may enhance contributors’ cognitive investment and potentially improve content quality. Second, AIGC disclosure policies are conceptualized as social-norm nudges intended to strengthen accountability and creative responsibility. While disclosure may encourage contributors to invest greater effort to signal originality, it may also trigger audience bias that could dampen engagement incentives. To address the limitations of binary AI labels, this study proposes a multi-level disclosure framework that reflects varying degrees of human-AI collaboration (e.g., AI-assisted, AI-co-created, AI-dominated), supplemented by structured textual explanations. Such nuanced governance mechanisms are theorized to reinforce perceived creative ownership and responsibility, thereby increasing cognitive investment and ultimately contributing to higher-quality AIGC.

Overall, this study develops a comprehensive, multi-level framework for understanding and governing AIGC pollution. By tracing the psychological trajectory from technological stimuli to individual cognition, ecosystem reactions, and governance interventions, the research advances theory on the unintended consequences of generative technologies and offers actionable guidance for platform managers and policymakers seeking to foster a sustainable and high-quality content ecosystem in the era of GenAI.

Key words: content economy, AI-generated content, content quality, human-GenAI interaction