Reference-Assisted Decoding of Sparse Tactile Codes Through Active Sequential Exploration: Time Costs and Similarity-Driven Errors in a Low-Resolution Vibrotactile Interface

Low-resolution tactile interfaces often require users to identify spatial codes through active sequential exploration, but the interaction costs and error structures of reference-assisted sparse-code decoding remain undercharacterized. We studied how 15 novice users decoded Braille-like alphabetic patterns on a 3 × 2 vibrotactile interface using blindfolded tactile exploration followed by visual reference-chart consultation. Participants completed a baseline single-location discrimination task and a three-block character-decoding task without explicit letter-pattern training or trial-wise correctness feedback. Single-location performance was near ceiling, whereas character-decoding accuracy increased across blocks. Exploration accounted for the larger share of task time, but incorrect trials showed a disproportionate prolongation of the composite post-exploration reference-assisted stage. Errors were structured by pattern similarity: most occurred between patterns separated by a Hamming distance of 1, and higher dot counts were associated with lower accuracy and longer response times. These findings inform low-resolution vibrotactile interface design by characterizing interaction costs and pattern confusability.