Temporal Causality Analysis of Sentiment Change in a Cancer Survivor Network

Area of Science:

• Computational Social Science
• Health Informatics
• Artificial Intelligence

Background:

• Online health communities provide vital information and social support for patients.
• The Cancer Survivor Network (CSN) is the largest online community for cancer patients, survivors, and caregivers.
• Previous research indicates benefits of online community participation, but lacks causal explanations for observed positive outcomes.

Purpose of the Study:

• To introduce a novel framework for examining temporal causality in sentiment dynamics within online health communities.
• To identify factors contributing to the social support benefits derived by participants in the CSN.
• To analyze the causal influence of replies on the sentiment of thread originators.

Main Methods:

• Developed a framework using Probabilistic Computation Tree Logic and probabilistic Kripke structures to model sentiment changes over time.
• Employed a machine learning-based sentiment classifier to categorize posts as positive or negative.
• Analyzed the probabilistic Kripke structure to determine prima facie causes and significance of sentiment shifts in thread originators.

Main Results:

• The sentiment expressed in replies was found to causally influence the sentiment of the thread originator.
• The findings remained robust despite variations in the sentiment classifier's threshold and specific model used.
• The framework was extended to account for uncertainty arising from imperfect sentiment classification.

Conclusions:

• Online community interactions, specifically replies, can positively and causally impact the sentiment of originating members.
• The developed temporal causality analysis framework offers insights for optimizing online community design and moderation for enhanced social support.
• The methodology has broad applicability for analyzing dynamic systems with state variables influenced by inputs.