Understanding Social Influence in Collective Product Ratings Using Behavioral and Cognitive Metrics

Title of the Paper

Understanding Social Influence in Collective Product Ratings Using Behavioral and Cognitive Metrics

Paper Information

• Research Domain: Human-Computer Interaction and Social Influence Studies
• Keywords: Social Influence, Social Conformity, Online Ratings, Brain-Computer Interface, EEG, Consumer Behavior, Emotion and Decision-Making, Social Computing

Research Background and Problem

• Research Questions and Challenges:
  • Online platforms often assist users' decision-making by displaying user-generated information, but such displays may induce social influence, affecting individual preferences.
  • Existing research has revealed the impact of social ratings on user preference formation, but there is a lack of in-depth understanding of the specific cognitive and emotional mechanisms involved.
  • There is insufficient research on the effects of rating presentation formats and the combination of ratings with textual reviews.
• Significance of the Research:
  • Understanding the cognitive and behavioral mechanisms of social influence is crucial for designing fair and effective online recommendation and rating systems.
  • This understanding can mitigate the negative impact of potential biases on user decisions and enhance the reliability of decision-support systems.
• Motivation and Related Work:
  • Existing studies mainly focus on user rating behaviors but overlook the combined effects of rating and textual review formats on users' attention, memory, and emotions.
  • Integrating neuroscience tools (e.g., EEG) with cognitive assessments may uncover the underlying mechanisms of social influence, providing a scientific basis for interface design improvements.

Solution

• Research Methods:
  • Design a two-phase experiment to quantify users' behavioral and cognitive changes after viewing others' evaluations (rating distributions and/or review titles).
  • Use EEG (electroencephalography) to record participants' attention, working memory, and emotional changes.
• Innovations:
  • Combine behavioral data and neurophysiological data to study the impact of presenting collective information (rating distributions and review titles) on users' social conformity.
  • Propose a new social conformity rating metric to evaluate how users' ratings shift toward collective preferences.
  • Apply Linear Discriminant Analysis (LDA) to demonstrate the high discriminative power of EEG data under different presentation conditions.
• Implementation Steps:
  1. Collect data on 90 smart home products (from the Amazon platform). Display content includes product images, titles, rating distributions, and review titles.
  2. Design presentation conditions: no display (no social information), rating distribution only (limited display), and rating distribution with reviews (comprehensive display).
  3. Conduct a two-phase experiment: the first phase collects initial preference ratings, and the second phase collects ratings under different presentation conditions while recording EEG signals.
  4. Extract EEG-related metrics, including attention (alpha/theta ratio), working memory (upper alpha power), and emotional valence (SASI index).
  5. Quantify the conformity effect in users' rating shifts and validate the predictive value of EEG data through LDA and regression analysis.

Research Findings

• Specific Findings:
  • Behavioral Level:
    • Users had the longest reaction times under the comprehensive display condition, indicating that additional review content significantly increased decision complexity.
    • Users were more likely to adjust their initial preferences to align with social preferences under the "ratings + reviews" condition (stronger conformity effect).
  • Cognitive Level:
    • Under the comprehensive display condition, users exhibited significantly higher levels of attention, working memory activation, and positive emotions compared to other conditions.
    • Emotional valence (SASI value) was a key cognitive indicator for predicting whether users conformed, supporting the reinforcement learning theory of social conformity.
  • Predictive Performance:
    • Comparing the discriminative ability of behavioral and EEG data, EEG metrics (e.g., emotional valence SASI) showed higher predictive power in distinguishing presentation conditions.
• Comparison with Existing Solutions:
  • This study is the first to integrate EEG data to deeply analyze the cognitive and emotional mechanisms behind social conformity, significantly surpassing research perspectives based solely on behavioral data.
  • It explores the additional effects of combined rating and review displays compared to rating-only displays, providing new insights for online recommendation system design.
• Limitations and Future Directions:
  1. The experiment primarily involved lightweight tasks in a controlled laboratory setting; future research should extend to more complex real-world scenarios.
  2. Explore the role of additional factors influencing social conformity (e.g., time pressure, review quality, familiarity).
  3. Conduct in-depth analysis of EEG data (e.g., temporal dimensions) to explore more detailed neural dynamic response processes.
  4. Further evaluate the practical application effects of the design recommendations to develop generalizable user interface optimization strategies.