Context-Aware Adaptive Visualizations for Critical Decision Making

Lopez-Cardona, Angela; Masias Bruns, Mireia; Attygalle, Nuwan T.; Idesis, Sebastian; Salvatori, Matteo; Raftopoulos, Konstantinos; Oikonomou, Konstantinos; Duraisamy, Saravanakumar; Emami, Parvin; Latreche, Nacera; Sahraoui, Alaa Eddine Anis; Vakalellis, Michalis; Vanderdonckt, Jean; Arapakis, Ioannis; Leiva, Luis A. · 2025 · Crossref

DOI: 10.3233/faia251433

archive: archived pipeline: cataloged verified

Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)

Summary

This paper addresses the limitation of static Information Visualization (InfoVis) dashboards, which fail to adapt to users’ real-time cognitive states during critical decision-making tasks. The authors propose SYMBIOTIK, an intelligent, context-aware adaptive visualization system that leverages neurophysiological signals to estimate mental workload (MWL) and dynamically adjust dashboard interfaces using reinforcement learning (RL). The research is motivated by the need for user-centric systems that account for individual cognitive differences and real-time mental states to improve task performance and engagement in high-stakes environments, such as law enforcement investigations. The study employed a user study with 120 participants to generate training data for three specific visualization types: distribution charts, timelines, and network graphs, modeled after a Crime Investigation Dashboard. Participants performed question-answering tasks while their electroencephalography (EEG) and eye-tracking (ET) signals were recorded. The system architecture processes raw EEG data in 2-second epochs, extracting power spectral density features from specific frequency bands (delta, theta, alpha, beta) to infer MWL levels. Separate RL agents were trained for each visualization layout using tabular Q-learning. These agents learn to select optimal adaptation strategies—ranging from no change to partial or full modifications of visual attributes like color, size, and shape—based on the user’s current cognitive state and the complexity of the task. The system utilizes a Kafka-based message broker to ensure scalable, real-time communication between sensing modules, the RL agent, and the adaptation engine. The results demonstrate that the SYMBIOTIK framework successfully improves task performance and user engagement compared to static dashboards. By continuously tracking cognitive load, the system provides granular, workload-sensitive adaptations that align with the user’s perceptual mechanisms. The study validates the methodology for neuroadaptive user interfaces, showing that RL agents can effectively learn policies to optimize visual presentations based on implicit physiological feedback. The authors note that behavioral indicators, such as accuracy and response time, strongly correlate with visual and question complexities, suggesting that multimodal approaches could further refine MWL proxies. The significance of this work lies in its contribution to the field of neuroadaptive interfaces, offering a scalable, real-time architecture for closed-loop implicit monitoring. It establishes a validated methodology for using EEG-driven RL to personalize data exploration experiences. The authors conclude that future work should focus on developing robust, personalized MWL metrics and integrating eye-tracking to model visual attention more precisely. The ultimate goal is to create layout-agnostic agents capable of generalizing across diverse settings, facilitating a symbiotic cooperation between humans and machines in critical decision-making scenarios.

Provenance

The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed.

Summary generated by qwen3.6-27b-nvidia on 2026-08-10; verification: verified.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

• neuro workload indices