Fear vs. Temptation: A Push-Pull-Mooring Perspective on Users’ Continuance Intention Toward AI Face-Swapping Technology

The rapid proliferation of AI face-swapping technology has created a complex user dilemma, characterized by both fascination and apprehension. This "technology paradox" has been the subject of a recent study, which utilized the Push-Pull-Mooring (PPM) framework to dissect the underlying mechanisms driving users’ continued engagement with these tools. The research, focusing on active users of the popular FacePlay application in China, reveals that while fears surrounding identity theft and privacy invasion (push effects) significantly influence user decisions, the allure of entertainment and social benefits (pull effects) exerts a far more dominant sway. Furthermore, individual factors like digital literacy and self-efficacy play a crucial role in mediating this delicate balance, acting as both buffers against fear and amplifiers of temptation.

The Rise of AI Face-Swapping and the Emergent Paradox

AI face-swapping technology, powered by advanced deep learning algorithms, allows for the highly realistic transfer of facial features between individuals. Its integration into the booming short-video industry, particularly in China, has led to widespread adoption. Data indicates that over 200 million users in China engaged with AI face-swapping applications in 2023, with platforms like FaceApp and Reface amassing substantial user bases. The technology’s low barrier to entry, high entertainment value, and inherent shareability have fueled its rapid growth.

However, this technological advancement is not without its shadows. Accompanying its creative appeal are escalating risks of identity theft, privacy invasion, and malicious misuse. This dichotomy – the coexistence of powerful temptation and significant fear – forms the bedrock of the "technology paradox," a phenomenon first identified by Mick and Fournier in their seminal work on consumer-technology relationships. Users, they argued, do not make simple binary choices but navigate a sustained tension between opposing forces. In the case of AI face-swapping, the direct manipulation of the face, the most defining biometric identifier, amplifies this tension, making users acutely aware of potential threats to their self-boundary even as they revel in the creative possibilities.

Decoding User Behavior: The Push-Pull-Mooring Framework

To unravel this complex interplay, the study adopted the Push-Pull-Mooring (PPM) model. Originally developed to explain population migration, the PPM framework is adept at simultaneously accounting for factors that drive individuals away from a behavior (push effects), attract them towards it (pull effects), and influence their ultimate decision-making (mooring effects).

In the context of AI face-swapping, the "push effects" are rooted in fear:

• Perceived Identity Threat: The fear that one’s facial features could be used to create deceptive content, damaging reputation or leading to identity theft. This is particularly potent given the irreversibility of facial data.
• Perceived Privacy Invasion: Concerns over the collection, processing, and potential leakage of sensitive facial biometric data, amplified by the opaque nature of AI algorithms and past privacy breaches.
• Technology Distrust: A general lack of confidence in the reliability and ethical deployment of AI face-swapping technology, fueled by algorithmic opacity and high-profile misuse cases.

Conversely, the "pull effects" are driven by temptation:

• Perceived Enjoyment: The intrinsic pleasure and amusement derived from the creative and transformative experience of face-swapping, a key driver for entertainment-oriented technologies.
• Perceived Usefulness: The practical value of AI face-swapping in content creation, social media engagement, and personal branding, lowering the barrier to producing high-quality visual content.
• Need for Social Identification: The desire for social recognition, group belonging, and affirmation through sharing engaging and trendy face-swapped content on social platforms.

Finally, "mooring effects" – individual-level factors – were examined through digital literacy and self-efficacy. Digital literacy, encompassing the ability to critically evaluate and use digital information, and self-efficacy, the belief in one’s capability to effectively use technology, were hypothesized to moderate the influence of push and pull factors.

Empirical Findings: Temptation Trumps Fear, Capabilities Matter

The study, involving 351 active users of the FacePlay app in China, employed Partial Least Squares Structural Equation Modeling (PLS-SEM) to analyze the collected data. The results painted a clear picture of user motivations:

• Dominance of Pull Effects: The pull effect exerted a significantly stronger positive influence on continuance intention compared to the negative influence of the push effect. This finding underscores that the allure of entertainment, usefulness, and social connection largely outweighs user anxieties, even when those anxieties are substantial. Perceived enjoyment emerged as a particularly strong driver within the pull dimension.
• Weight of Privacy Concerns: Within the push effect, perceived privacy invasion was found to have the highest weight. This highlights the unique psychological impact of facial data concerns, given their irreversible nature, surpassing even the fears associated with identity manipulation.
• Direct Impact of Mooring Factors: Both digital literacy and self-efficacy directly and positively influenced continuance intention, suggesting that individual capabilities are not just moderators but also independent drivers of continued engagement.
• Dual Moderating Mechanism: Crucially, digital literacy and self-efficacy demonstrated a dual moderating role. They acted as a "buffer" by weakening the inhibitory influence of the push effect (fear) and as an "amplifier" by strengthening the facilitating influence of the pull effect (temptation). This means users with higher digital competence and confidence are more likely to overcome fears and maximize the benefits of the technology.

However, the study also noted that the moderating effect sizes were relatively small, suggesting that the potent allure of temptation can partially override individual self-regulation. This observation carries significant implications for policy and platform design.

Broader Implications and Future Directions

The findings of this study offer substantial theoretical and practical insights into the complex dynamics of AI face-swapping technology adoption.

Theoretical Contributions:

• Integrated Framework: The study successfully integrates the Push-Pull-Mooring framework with the technology paradox to provide a holistic explanation for continuance intention in a context where fear and temptation coexist.
• Asymmetric Push-Pull Dynamics: It empirically demonstrates the dominance of pull factors (temptation) over push factors (fear) in the AI face-swapping context, refining our understanding of the technology paradox.
• Dual Moderating Role of Capabilities: The identification of a dual moderating mechanism for digital literacy and self-efficacy enriches the understanding of how individual capabilities shape user behavior in complex technological environments.

Practical Implications:

• For Platform Designers: The emphasis on privacy transparency and continuous enhancement of entertainment features is paramount. Platforms should clearly communicate data handling practices and invest in updating creative tools to sustain user engagement.
• For Differentiated User Operations: Recognizing that users with varying digital literacy levels respond differently to risks and rewards, platforms should tailor strategies. For less digitally literate users, simpler onboarding and clearer explanations are needed, while for more advanced users, opportunities for deeper engagement and creative expression should be provided.
• For Policymakers: The study underscores the need for robust regulatory frameworks that go beyond simple labeling. Mandatory data protection standards, third-party audits, and "safety-by-design" principles are crucial to mitigate risks, especially given the dominant influence of temptation factors. The findings suggest that individual self-regulation alone is insufficient to manage the potential downsides of such powerful technologies.

Limitations and Future Research:

The study acknowledges its limitations, including a cross-sectional design and a sample primarily composed of younger Chinese users. Future research could benefit from longitudinal studies to track evolving user behavior, incorporate objective usage data, and conduct cross-cultural comparisons to assess the generalizability of the findings across different regulatory and societal contexts. Expanding the investigation to include a wider range of mooring variables and applying the fear-vs.-temptation framework to other emerging AI technologies will further enrich our understanding of human-technology interaction in the age of artificial intelligence.