Behavior-Driven Adaptive Multimodal User Experience in Virtual Reality: The AMUX-VR Framework for Cultural Heritage Applications
Tez Türü: Doktora
Tezin Yürütüldüğü Kurum: Ankara Yıldırım Beyazıt Üniversitesi, Mühendislik Ve Doğa Bilimleri Fakültesi, Bilgisayar Mühendisliği, Türkiye
Tez Danışmanı: Muhammed Abdullah Bülbül
Tezin Onay Tarihi: 2026
Tezin Dili: İngilizce
Özet:
Immersive virtual reality applications commonly rely on predefined interaction and presentation structures, while existing adaptive approaches often focus on specific functions rather than continuous user behavior. This dissertation presents AMUX-VR (Adaptive Multimodal User Experience for Virtual Reality), a behavior-responsive framework developed for virtual cultural heritage environments. AMUX-VR integrates observable user behavior, operational state classification, adaptive environmental responses, and continued interaction within a closed-loop architecture.
The framework was implemented in a virtual madrasa containing six tasks involving behavioral progression, reading, object-directed attention, process observation, contextual narrative interaction, and multisensory storytelling. Head angular velocity constituted the principal continuous behavioral input. The resulting Earth_Slow and Fire_Fast states were treated as movement-based operational system classifications rather than psychological or emotional states. Task, targeting, interaction, proximity, and navigation events additionally supported context-dependent system responses.
AMUX-VR was evaluated using a within-subject experimental design with 34 participants who completed both the Adaptive and Static conditions. The evaluation combined time-series data, event and adaptation logs, task outcomes, navigation measures, and post-condition questionnaires. The analytical approach distinguished five levels of evidence: behavioral input and inference, delivery of the system response, subsequent behavior, task-specific consequences, and subjective experience.
System-validation results demonstrated internally consistent operation across 2,925 strict state transitions, with agreement among the trigger, resulting state, state label, and event order. Session-order analyses showed that changes in total task time and stuck count were more strongly associated with repeated exposure and a second-session advantage than with experimental condition. Among the 20 participants with valid paired temporal coverage, the reduction in angular velocity three seconds after high-motion onset was greater under Adaptive exposure (W = 34.0, p = .008, rank-biserial correlation = −.676). This finding represents a short-window kinematic difference and should not be interpreted as evidence of psychological calming or general user-experience superiority.
Task-specific findings did not demonstrate uniform superiority of the Adaptive condition. T1 actor-viewing behavior was influenced by repeated exposure, while no corrected condition differences were detected for T2 reading or T3 astrolabe-directed attention measures. T4 observation outcomes were associated with condition-dependent presentation timing and session order. T5 results indicated that the additional time under Adaptive exposure occurred in the interaction pathway preceding the fixed reading interval. T6 measures showed a directional pattern favoring Adaptive exposure, but did not provide conclusive evidence of behavioral superiority after correction for multiple comparisons. The Adaptive navigation policy reduced guidance exposure by delaying the route line, but did not produce superior navigation performance.
No corrected condition differences were found among the general subjective-experience measures. However, perceived atmospheric change was significantly greater under Adaptive exposure (Holm-adjusted p = .037). Perceived environmental responsiveness did not remain statistically significant after correction for multiple comparisons.
In conclusion, AMUX-VR was supported at four levels: consistent execution of behavior-responsive rules, greater short-term movement reduction at the focused three-second endpoint, implementation of a delayed-guidance policy, and production of a perceptible atmospheric difference. These findings did not translate into uniform superiority across all tasks, navigation performance, or general user experience. The dissertation contributes both an implemented behavior-responsive multimodal VR architecture and a five-layer evaluation approach that distinguishes correct system operation from evidence of human benefit.
Keywords: adaptive virtual reality; behavior-responsive interaction; cultural heritage; multimodal interaction; task–response alignment; user experience