Immersive Experimental Teaching Interaction Design Based on Virtual Reality Environment

Immersive virtual reality allows broadening access to laboratory experiments by minimizing the need for tangible devices and repeated procedural learning, but pre-defined instructional cues give little help to learners in case of diverging from complex multi-stage experimental procedures. To solve this problem, an adaptive teaching framework for experimental procedures in acid-base titration was developed using Unity 2022.3 LTS, Meta Quest 3, behavioral event logging, Hidden Markov Model (HMM) and Viterbi decoding. The system detected five states of operation and provided feedback through visual, haptic, and textual feedback in case of illegal transitions, missed steps, or repeated invalid moves. In a controlled experiment conducted on 72 undergraduate students, a model-guided VR condition and a basic-VR condition were compared. The HMM gave an accuracy of 90.8%, Macro-F1 score of 0.908 and mean inference delay of 37.6 ± 6.2 ms. The group with the model guidance made fewer procedural mistakes (1.78 ± 1.35 vs 4.39 ± 1.90), completed more steps (94.82% vs 89.77%), gained more knowledge (17.47 ± 6.83 vs 11.39 ± 5.15) and had lower NASA-TLX scores (37.13 ± 10.01 vs 48.26 ± 10.40). This study proves that adaptive interaction with state awareness allows connecting real-time behavior recognition with improvements in procedural accuracy, cognitive efficiency and learning in laboratories.