This paper presents a novel trajectory prediction model for autonomous vehicles (AVs) called GAVA, which integrates principles of human cognition and observational behavior to enhance the accuracy and adaptability of trajectory prediction in mixed-autonomy traffic environments. The model introduces an "adaptive visual sector" mechanism that mimics how human drivers dynamically allocate attention based on factors like speed, spatial orientation, and proximity. It also develops a "dynamic traffic graph" using Convolutional Neural Networks (CNN) and Graph Attention Networks (GAT) to capture spatio-temporal dependencies among agents. Benchmark tests on the NGSIM, HighD, and MoCAD datasets show that GAVA outperforms state-of-the-art baselines by at least 15.2%, 19.4%, and 12.0%, respectively. The model's contributions include a sophisticated pooling mechanism that replicates human attention allocation, a novel dynamic traffic graph for agent interaction modeling, and a multi-head attention mechanism for spatiotemporal dependency modeling. The model also incorporates a vision-aware module that adapts the visual sector based on speed, and a priority-aware module that generates multi-modal trajectory predictions. The results demonstrate that GAVA achieves high accuracy and applicability in various traffic scenarios, including highways and dense urban areas. The study highlights the potential of leveraging human cognition principles to improve trajectory prediction algorithms for AVs.This paper presents a novel trajectory prediction model for autonomous vehicles (AVs) called GAVA, which integrates principles of human cognition and observational behavior to enhance the accuracy and adaptability of trajectory prediction in mixed-autonomy traffic environments. The model introduces an "adaptive visual sector" mechanism that mimics how human drivers dynamically allocate attention based on factors like speed, spatial orientation, and proximity. It also develops a "dynamic traffic graph" using Convolutional Neural Networks (CNN) and Graph Attention Networks (GAT) to capture spatio-temporal dependencies among agents. Benchmark tests on the NGSIM, HighD, and MoCAD datasets show that GAVA outperforms state-of-the-art baselines by at least 15.2%, 19.4%, and 12.0%, respectively. The model's contributions include a sophisticated pooling mechanism that replicates human attention allocation, a novel dynamic traffic graph for agent interaction modeling, and a multi-head attention mechanism for spatiotemporal dependency modeling. The model also incorporates a vision-aware module that adapts the visual sector based on speed, and a priority-aware module that generates multi-modal trajectory predictions. The results demonstrate that GAVA achieves high accuracy and applicability in various traffic scenarios, including highways and dense urban areas. The study highlights the potential of leveraging human cognition principles to improve trajectory prediction algorithms for AVs.