Modeling social interaction dynamics using temporal graph networks

The project represents a academic research artifact associated with a 2024 arXiv paper. Despite the recent paper date, the repository shows zero stars and minimal activity (0 velocity), suggesting it has not yet gained traction within the Human-Robot Interaction (HRI) or Graph Neural Network (GNN) communities. The core innovation—applying Temporal Graph Networks to social dynamics—is a sound application of existing GNN techniques to a niche domain, but it lacks a technical moat or a proprietary dataset that would prevent others from replicating the work. Frontier labs (OpenAI, Google) are unlikely to compete directly as this is a specific robotics/HRI application, but it faces high displacement risk from other academic labs (e.g., Stanford's Social-STGCNN or MIT's CSAIL) and commercial robotics startups developing proprietary social navigation stacks. The defensibility is currently limited to the specific 'adapted' architecture described in the paper, which functions more as a reference implementation than a deployable tool.