Active Sampling for Efficient Subjective Evaluation of Tactons at Scale

• Published in: 2021 IEEE World Haptics Conference (WHC) pp. 1 - 6
• Main Authors: Demers, Marc, Fortin, Pascal E., Weill--Duflos, Antoine, Yoo, Yongjae, Cooperstock, Jeremy R.
• Format: Conference Proceeding
• Language: English
• Published: IEEE 06-07-2021
• Subjects: Authoring systems; Conferences; Explosions; Haptic interfaces; Information theory; Sociology; Statistics
• DOI: 10.1109/WHC49131.2021.9517257

Traditional tacton evaluation studies often rely on pre-defined haptic effects that are specifically tailored to explore a handful of design parameters. To prevent combinatorial explosion, researchers are forced to constrain their exploration to very limited subsets of the parameter space. In this work, we propose a hands-off active sampling strategy grounded in probability and information theory that automatically generates tactons to maximize the perceptual information gain at each stimulus presentation. As a proof of concept of the proposed technique, we present the results from a crowdsourced study investigating the perceived similarity between tactons with over 200 participants. Without researcher intervention in the tacton selection process, our method allowed a set of the most salient features for perception of tacton similarity to emerge naturally from the data. This approach is highly scalable and allows for a more efficient exploration of a larger haptic space than typical laboratory study designs aimed at evaluating perceptual attributes of tactons.