Aggression is a socially complex behaviour that intelligent virtual agents (IVAs) must convincingly convey in applications such as customer service and conflict training. Despite its importance, aggression remains understudied: prior work has focused on basic emotions and unimodal cues, providing little insight into how aggression can be modelled multimodally or systematically scaled by intensity. We present a psychologically grounded model that parametrises language, voice, body movement and facial expressions, across four aggression levels. We evaluated the model in two studies with 38 flight attendants. Experiment 1 tested unimodal cues, showing all modalities except language conveyed aggression gradients. Experiment 2 extended this by combining modalities, demonstrating that coordinated multimodal integration stabilised weaker language cues and produced perceptually robust aggression levels (low, mid, and high) with body and facial cues carrying most weight. Our work contributes the first validated multimodal, multi-level aggression model for IVAs, offering design principles for broader socially expressive agents.

Immersive virtual reality (VR) applications demand accurate, temporally coherent full-body pose tracking. Recent head-mounted camera-based approaches show promise in egocentric pose estimation, but encounter challenges when applied to VR head-mounted displays (HMDs), including temporal instability, inaccurate lower-body estimation, and the lack of real-time inference. To address these limitations, we present EgoPoseVR, an end-to-end framework for accurate egocentric full-body pose estimation in VR that integrates headset motion cues with egocentric RGB-D observations through a dual-modality fusion pipeline. A spatiotemporal encoder extracts frame- and joint-level representations, which are fused via cross-attention to fully exploit complementary motion cues across modalities. A kinematic optimization module then imposes constraints from HMD signals, enhancing the accuracy and stability of pose estimation. To facilitate training and evaluation, we introduce a large-scale synthetic dataset of over 1.8 million temporally aligned HMD and RGB-D frames across diverse VR scenarios. Experimental results show that EgoPoseVR outperforms state-of-the-art egocentric pose estimation models. A user study in real-world scenes further shows that EgoPoseVR achieved significantly higher subjective ratings in accuracy, stability, embodiment, and intention for future use compared to baseline methods. These results show that EgoPoseVR enables robust full-body pose tracking, offering a practical solution for accurate VR embodiment without requiring additional body-worn sensors or room-scale tracking systems.