Generating Causal Temporal Interaction Graphs for Counterfactual Validation of Temporal Link Prediction

TL;DR

This paper introduces a framework for counterfactual validation of temporal link prediction models by generating causal graphs with known ground-truth structures, enabling causality-aware benchmarking.

Contribution

It proposes a structural equation model for continuous-time events, extends it to causal graphs, and develops a distance metric for comparing causal models based on predictive error.

Findings

Predictors trained on one causal model perform worse on distant models.

The framework can evaluate causal shifts and timestamp shuffling as causal distortions.

Provides a foundation for causality-aware benchmarking of TLP models.

Abstract

Temporal link prediction (TLP) models are commonly evaluated based on predictive accuracy, yet such evaluations do not assess whether these models capture the causal mechanisms that govern temporal interactions. In this work, we propose a framework for counterfactual validation of TLP models by generating causal temporal interaction graphs (CTIGs) with known ground-truth causal structure. We first introduce a structural equation model for continuous-time event sequences that supports both excitatory and inhibitory effects, and then extend this mechanism to temporal interaction graphs. To compare causal models, we propose a distance metric based on cross-model predictive error, and empirically validate the hypothesis that predictors trained on one causal model degrade when evaluated on sufficiently distant models. Finally, we instantiate counterfactual evaluation under (i) controlled causal shifts between generating models and (ii) timestamp shuffling as a stochastic distortion with measurable causal distance. Our framework provides a foundation for causality-aware benchmarking.