JMIR Publications

ABSTRACT

Background:

Online social support, interaction among individuals that one helps another within a difficult situation through online platforms such as online forums or social media. Despite the growing usage and importance of online social support, prior studies have mainly focused on only either the characteristics of support-seekers or identifying supporting types, and there are a few rooms to improve the prior research in terms of two perspectives. First, intensity of support, which indicates the strength of the willingness represented in the given support message, has been little explored. In addition, the models proposed in previous research have relied on hand-crafted features or simple text embeddings of the whole support message generated, which are difficult to support explainability for the decision by the prediction model.

Objective:

This study proposes a deep learning model that determines the intensity of online social support, particularly for informational and emotional support provided in online communities. While prior studies mainly concentrated on the identification task for support types, this research quantifies the strength of support, which may better explain how strongly supporters indicate their informational and emotional support.

Methods:

The proposed model categorizes informational and emotional support into three levels—strong, moderate, and weak. It collaborates with a large language model (LLM) to extract key sentences or phrases expressing supportive intent and to compute sentiment scores from both the original post and the reply. All the computed features are fed into the neural network layers for the final prediction. The proposed model was evaluated on a manually annotated dataset consisting of post–comment pairs.

Results:

The model achieves superior accuracy compared to baselines (0.713 for informational and 0.700 for emotional support). Throughout the analysis of case examples, we demonstrate that the extracted textual evidence provides interpretability: explicit empathy or detailed advice corresponds to strong support, while vague or indirect comments correspond to weak support.

Conclusions:

Cooperating with Large Language Models (LLMs), quantifying the intensity of online social support is highly achievable. In addition, LLMs can provide the great explainability for the decision made by the model.