TL;DR
This study explores how consumers' understanding of algorithms (Algorithm Awareness or AA) influences their decisions to share personal information on e-commerce platforms. By integrating the dual calculus model and the theory of planned behavior, the research reveals that AA significantly impacts how users perceive privacy risks and benefits, ultimately shaping their willingness to disclose data.
What Problem Does This Paper Address?
In today's digital landscape, e-commerce platforms heavily rely on personalized algorithms to tailor user experiences. While these algorithms offer benefits like better product recommendations, they also raise significant privacy concerns. Consumers often lack understanding of how these algorithms operate, leading to a "black box" problem where their data is collected and used without their full awareness. This lack of transparency can lead to privacy risks and erode trust. Existing research has touched upon consumer privacy behaviors, but there's a notable gap in understanding the specific role of "algorithm awareness" (AA) in the decision-making process of information disclosure, particularly the underlying mechanisms. This paper aims to bridge this gap by investigating how AA influences consumers' privacy decisions within the context of AI-driven personalized e-commerce.
Methods
The study employs a quantitative approach, constructing a theoretical model that integrates the dual calculus model and the theory of planned behavior (TPB).
- Dual Calculus Model: This model examines information disclosure decisions through two trade-offs:
- Risk Calculus: Weighing privacy threats against coping mechanisms.
- Privacy Calculus: Weighing perceived benefits against privacy risks.
- Theory of Planned Behavior (TPB): This theory posits that behavioral intention is shaped by attitude, subjective norms, and perceived behavioral control.
The researchers developed a questionnaire to measure various constructs, including algorithm awareness (based on fairness, explainability, accountability, and transparency - FEAT), threat appraisals (perceived vulnerability and severity), coping appraisals (response efficacy and self-efficacy), perceived benefits and risks, privacy attitude, subjective norms, perceived behavioral control, and information disclosure intention. Structural equation modeling (SEM) using SmartPLS 4.0 was used to analyze the data and test the hypotheses.
Datasets
The study collected data from 581 online-shopping consumers in China through an online survey. Participants were users of e-commerce platforms who had shopped online in the past three months. A pretest with 95 users was conducted to refine the questionnaire.
Key Insights and Ideas
- Algorithm Awareness is Foundational: AA isn't just about knowing algorithms exist; it's about understanding their attributes (fairness, explainability, accountability, transparency) and how they impact personal data. This awareness is a crucial first step in making informed privacy decisions.
- AA Shapes Risk Perception: Higher AA leads to lower perceived vulnerability and severity of privacy threats, while simultaneously boosting perceived response efficacy and self-efficacy in managing privacy. This suggests that understanding algorithms empowers users to feel more in control.
- The Privacy Paradox is Real (and Complex): While consumers express concerns about privacy, they often disclose information when perceived benefits outweigh risks. The study found that perceived behavioral control had an insignificant impact on disclosure intention, suggesting that other factors like privacy attitude and subjective norms are more dominant. This highlights a potential "privacy paradox" where stated concerns don't always translate into action, possibly due to a lack of viable alternatives or a perceived necessity to trade privacy for benefits.
- Mediating Mechanisms are Key: The study effectively demonstrates how AA influences privacy decisions through a chain of mediators. AA affects threat and coping appraisals, which in turn influence perceived risk. This perceived risk, along with perceived benefits, shapes privacy attitudes, which then influence the intention to disclose information.
Key Results
- Algorithm Awareness (AA) significantly and negatively impacts perceived vulnerability and severity. (H1, H2 supported)
- AA significantly and positively impacts response efficacy and self-efficacy. (H3, H4 supported)
- Perceived vulnerability and severity positively influence perceived risks. (H5, H6 supported)
- Response efficacy and self-efficacy negatively influence perceived risks. (H7, H8 supported)
- Perceived benefits positively influence privacy attitude, while perceived risks negatively influence it. (H9, H10 supported)
- Privacy attitude and subjective norms positively influence information disclosure intention. (H11, H12 supported)
- Perceived behavioral control's direct impact on information disclosure intention was insignificant. (H13 not supported)
- Perceived behavioral control positively influences self-efficacy. (H14 supported)
- The mediation analysis showed that AA partially mediates the relationship with perceived risk through threat and coping appraisals. Privacy attitude partially mediates the relationship between perceived risks/benefits and information disclosure intention.
Limitations and Future Work
- Sample Limitation: The study's sample was limited to Chinese e-commerce users. Future research could explore these dynamics in different cultural contexts.
- Self-Reported Data: Data was collected via questionnaires, which may not perfectly reflect actual online behaviors.
- Moderator Variables: The authors suggest adding moderator variables to the model to explore boundary conditions and further understand how AA affects privacy decision-making.
- Methodological Expansion: Future studies could incorporate experimental methods, such as scenario experiments or event-related potentials (ERP), to gather more robust behavioral data.
- Privacy Paradox Exploration: Further investigation into the privacy paradox, particularly the reasons behind the disconnect between stated concerns and actual disclosure behavior, is recommended.
Summary updated