Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Peer Disambiguation in Self-Reported Surveys using Graph Attention Networks

Ajitesh Srivastava, Aryan Shetty, Eric Rice

TL;DR

The paper tackles the challenge of ambiguities in self-reported social-network data used for peer-based interventions. It introduces a Graph Attention Network (GAT) framework to solve two ambiguity tasks—pair disambiguation and link existence—by learning embeddings from node attributes and network structure, with explanations provided by GNNExplainer. Across real and simulated ambiguities in a dataset of active-duty military personnel, the approach improves network accuracy and downstream suicide-risk prediction, outperforming decision trees and MLP baselines. The work demonstrates the practical impact of resolving network ambiguities on targeted interventions, and highlights the value of interpretable GNN explanations for domain researchers and practitioners.

Abstract

Studying peer relationships is crucial in solving complex challenges underserved communities face and designing interventions. The effectiveness of such peer-based interventions relies on accurate network data regarding individual attributes and social influences. However, these datasets are often collected through self-reported surveys, introducing ambiguities in network construction. These ambiguities make it challenging to fully utilize the network data to understand the issues and to design the best interventions. We propose and solve two variations of link ambiguities in such network data -- (i) which among the two candidate links exists, and (ii) if a candidate link exists. We design a Graph Attention Network (GAT) that accounts for personal attributes and network relationships on real-world data with real and simulated ambiguities. We also demonstrate that by resolving these ambiguities, we improve network accuracy, and in turn, improve suicide risk prediction. We also uncover patterns using GNNExplainer to provide additional insights into vital features and relationships. This research demonstrates the potential of Graph Neural Networks (GNN) to advance real-world network data analysis facilitating more effective peer interventions across various fields.

Peer Disambiguation in Self-Reported Surveys using Graph Attention Networks

TL;DR

The paper tackles the challenge of ambiguities in self-reported social-network data used for peer-based interventions. It introduces a Graph Attention Network (GAT) framework to solve two ambiguity tasks—pair disambiguation and link existence—by learning embeddings from node attributes and network structure, with explanations provided by GNNExplainer. Across real and simulated ambiguities in a dataset of active-duty military personnel, the approach improves network accuracy and downstream suicide-risk prediction, outperforming decision trees and MLP baselines. The work demonstrates the practical impact of resolving network ambiguities on targeted interventions, and highlights the value of interpretable GNN explanations for domain researchers and practitioners.

Abstract

Studying peer relationships is crucial in solving complex challenges underserved communities face and designing interventions. The effectiveness of such peer-based interventions relies on accurate network data regarding individual attributes and social influences. However, these datasets are often collected through self-reported surveys, introducing ambiguities in network construction. These ambiguities make it challenging to fully utilize the network data to understand the issues and to design the best interventions. We propose and solve two variations of link ambiguities in such network data -- (i) which among the two candidate links exists, and (ii) if a candidate link exists. We design a Graph Attention Network (GAT) that accounts for personal attributes and network relationships on real-world data with real and simulated ambiguities. We also demonstrate that by resolving these ambiguities, we improve network accuracy, and in turn, improve suicide risk prediction. We also uncover patterns using GNNExplainer to provide additional insights into vital features and relationships. This research demonstrates the potential of Graph Neural Networks (GNN) to advance real-world network data analysis facilitating more effective peer interventions across various fields.

Paper Structure

This paper contains 28 sections, 4 equations, 4 figures, 6 tables.

Table of Contents

  1. Introduction
  2. Background and Related Work
  3. Peer-based Interventions
  4. Challenges in Data Collection and Ambiguity
  5. Graph Neural Networks in Social Network Analysis
  6. Contributions
  7. Data and Problem Definition
  8. Dataset Description
  9. Disambiguation Problems
  10. Method
  11. Embedding Model: Graph Attention Network (GAT)
  12. Feature Aggregation:
  13. Stacking GAT Layers
  14. Model Architecture and Training
  15. Evaluation Setup
  16. ...and 13 more sections

Figures (4)

  • Figure 1: Graph network from the dataset
  • Figure 2: Approach overview: The input graph consisting of nodes representing individual features is fed into the GAT model to generateembeddings. These embeddings are used to resolve the ambiguity between multiple candidates (pair disambiguation) and to determine the likelihood of a link between individuals (link existence).
  • Figure 3: Subgraph and features identified by GNNExplainer as decisive in identifying the existence of link given in red. (a) The nodes share common neighbors and a few important features. (b) The nodes do not share any common neighbors, but many important features
  • Figure 4: Decision Tree with class distribution