How Does Inattention Influence the Robustness and Efficiency of Adaptive Procedures in the Context of Psychoacoustic Assessments via Smartphone?

• Published in: Trends in hearing Vol. 28; p. 23312165241288051
• Main Authors: Xu, Chen, Hülsmeier, David, Buhl, Mareike, Kollmeier, Birger
• Format: Journal Article
• Language: English
• Published: Los Angeles, CA SAGE Publications 01.01.2024; Sage Publications Ltd
• Subjects: Acoustic Stimulation; Attention - physiology; Auditory Perception - physiology; Auditory Threshold; Computer Simulation; Efficiency; Humans; ISAAR 2023; Monte Carlo Method; Psychoacoustics; Psychometrics; Reproducibility of Results; Smartphone; model-free adaptive procedure; Monte-Carlo simulations; mobile listening test; inattention model
• ISSN: 2331-2165; 2331-2165
• DOI: 10.1177/23312165241288051

Inattention plays a critical role in the accuracy of threshold measurements, e.g., when using mobile devices. To describe the influence of distraction, long- and short-term inattention models based on either a stationary or a non-stationary psychometric function were developed and used to generate three simulated listeners: fully-, moderately-, and non-concentrated listeners. Six established adaptive procedures were assessed via Monte-Carlo simulations in combination with the inattention models and compared with a newly proposed method: the graded response bracketing procedure (GRaBr). Robustness was examined by bias and root mean square error between the “true” and estimated thresholds while efficiency was evaluated using rates of convergence and a normalized efficiency index. The findings show that inattention has a detrimental impact on adaptive procedure performance—especially for the short-term inattentive listener—and that several model-based procedures relying on a consistent response behavior of the listener are prone to errors owing to inattention. The model-free procedure GRaBr, on the other hand, is considerably robust and efficient in spite of the (assumed) inattention. As a result, adaptive techniques with desired properties (i.e., high robustness and efficiency) as revealed in our simulations—such as GRaBr—appear to be advantageous for mobile devices or in laboratory tests with untrained subjects.