Sad or just jealous? Using Experience Sampling to Understand and Detect Negative Affective Experiences on Instagram

Title of the Paper

Sad or just jealous? Using Experience Sampling to Understand and Detect Negative Affective Experiences on Instagram

Paper Information

• Topic Area: Emotional experiences and detection on social media, particularly negative emotions on Instagram
• Keywords: Emotion detection, social media, smartphones, experience sampling method, Instagram

Research Background and Issues

• Problems and Challenges:

  1. The use of social networking services (SNS) can trigger a variety of emotional experiences, with both positive and negative emotions coexisting.
  2. Negative emotions, such as jealousy and appearance comparison, can have profound effects on users' emotions and mental health.
  3. Although the importance of emotions in SNS experiences is widely recognized, systematic methods for recording and detecting user emotions remain underdeveloped.

• Significance of the Research:

  • Understanding the emotions triggered by SNS use and their impacts can reveal how to design social platforms that support emotional management.
  • Automated emotion detection can aid in developing intervention systems that effectively adjust negative emotional experiences.

• Motivation and Related Work:

  • The Experience Sampling Method (ESM) is a widely used technique for recording emotions in everyday environments.
  • Existing ESM methods suffer from recall bias and operational burdens, especially in contexts involving negative emotions.
  • Combining smartphone sensor data can help address the limitations of passive sampling, but such studies are mostly confined to laboratory settings and have not been validated in real SNS usage scenarios.

Solution

• Methods and Solutions:

  1. First Study: An improved ESM method was used to develop a mobile application that captures emotional states in real-time after Instagram use, recording specific emotional experiences such as appearance comparison and jealousy.
  2. Second Study: Smartphone sensor data (e.g., touch and motion data) was captured, and machine learning models were used to detect specific categories of emotions.

• Innovations:

  • A context-triggered ESM method was proposed, which records emotional responses immediately after Instagram use, thereby reducing recall bias.
  • A complete system was developed that uses smartphone sensor data to detect triggers of negative emotions in real-time, marking the first application of this approach in real-world usage environments.

• Implementation Steps and Key Techniques:

  • Step 1: Use questionnaires and scales to capture self-reported data on negative emotions such as appearance comparison and jealousy.
  • Step 2: Collect sensor data and perform feature extraction, including statistical features such as minimum, maximum, and mean values of motion and touch data.
  • Step 3: Use machine learning classifiers (e.g., SVM) to predict categories of negative emotions based on sensor data and evaluate model performance.

Research Outcomes

• Specific Findings:

  1. First Study: Found that appearance comparison and jealousy were the most frequently experienced negative emotions during Instagram use, accounting for 16.6% and 17% of all emotions, respectively.
  2. Second Study: Successfully achieved automatic detection of appearance comparison (accuracy up to 95.78%) and jealousy (accuracy up to 93.95%) using sensor data.

• Comparison with Existing Solutions:

  • The study demonstrated the feasibility of combining smartphone sensors for emotion detection, achieving high classification accuracy even in everyday scenarios.
  • Compared to traditional ESM methods, the new approach significantly reduced recall bias and improved the temporal precision of emotional descriptions.

• Experimental or Evaluation Results:

  • High-accuracy detection of appearance comparison and jealousy highlighted the superiority of sensor-based motion data features.
  • The robust performance of the classifier with a small amount of training data (especially using SVM as the classifier) provides support for future deployment.

• Limitations and Future Directions:

  1. The sampled data mainly came from young populations (especially college students), which may limit generalizability.
  2. Dependence on Instagram may make it difficult to extend the findings to other SNS platforms or media styles.
  3. Future research is recommended to enhance SNS feature integration, making functionalities similar to Instagram more comprehensive.
  4. Further exploration of other emotion detection technologies (e.g., physiological sensors) could improve the accuracy of emotion classification.

Conclusion

This study innovatively combines ESM with smartphone sensor data to detect negative emotions during SNS use. The results validate its efficiency and practicality, laying the foundation for the development of emotion-aware SNS services.