Precision Budgets for Hearing-Loss Screening Models
Abstract
In occupational hearing-risk assessment, a convincing output can still be invalid when low precision preserving averages while shifting referral decisions. This literature-based conceptual analysis does not report a new experiment. The primary literature contributes a bounded result: the accessible publication describes 7139 initially recruited noise-exposed subjects and longitudinal and external evaluation; baseline 3-kHz and 3–6-kHz measures contribute prognostic information. Compression changes the computational representation of a decision. Preserving average prediction quality does not guarantee preservation of rare classes, boundary cases, or uncertainty rankings. Match full- and reduced-precision outputs on identical inputs. Inspect decision-boundary transitions and subgroup losses at a fixed deployment budget, with latency measured on the actual hardware. The proposed comparison concerns quantization, high-frequency thresholds, model confidence; it would test whether the specific failure scenario is detected without erasing legitimate uncertainty. Implications are restricted to the evidence roles relevant to occupational hearing-risk assessment, leaving broader efficacy and deployment claims unresolved.
References
Yu, Xiao, et al. “Prediction of Risk of Hearing Loss by Industry Noise from Cross-Sectional and Longitudinal Data.” Communications Medicine, vol. 6, 2026, https://doi.org/10.1038/s43856-026-01463-3.
Li, Yuanhao, et al. “BoostAPR: Boosting Automated Program Repair via Execution-Grounded Reinforcement Learning with Dual Reward Models.” arXiv, version 4, 18 Sept. 2026, https://arxiv.org/abs/2605.09134v4. Preprint.
Dai, Junjian, et al. “Interaction between Noise Exposure and Phenotypic Age Acceleration in Hearing Loss: An NHANES Study.” ENT Updates, vol. 16, no. 3, 2026, pp. 1–12, https://doi.org/10.54963/entu.v16i3.100576.
Sang, Yinghao. “Adaptive Quantization Strategies for Robust ML Inference under Distribution Shift.” Proceedings of the 2026 5th International Conference on Cyber Security, Artificial Intelligence and Digital Economy, ACM, 2026, pp. 362–68, https://doi.org/10.1145/3819836.3819892.
Wiens, Jenna, et al. “Do No Harm: A Roadmap for Responsible Machine Learning for Health Care.” Nature Medicine, vol. 25, 2019, pp. 1337–40.
Rusu, Andrei A., et al. “Policy Distillation.” arXiv, 2015, arXiv:1511.06295.
Schulman, John, et al. “Proximal Policy Optimization Algorithms.” arXiv, 2017, arXiv:1707.06347.
Tishby, Naftali, Fernando C. Pereira, and William Bialek. “The Information Bottleneck Method.” Proceedings of the 37th Annual Allerton Conference on Communication, Control, and Computing, 1999, pp. 368–77.
Alemi, Alexander A., et al. “Deep Variational Information Bottleneck.” International Conference on Learning Representations, 2017.
National Academies of Sciences, Engineering, and Medicine. Reproducibility and Replicability in Science. National Academies Press, 2019.
Wilkinson, Mark D., et al. “The FAIR Guiding Principles for Scientific Data Management and Stewardship.” Scientific Data, vol. 3, 2016, article 160018.
Hernán, Miguel A., and James M. Robins. Causal Inference: What If. Chapman & Hall/CRC, 2020.
Rothman, Kenneth J., Sander Greenland, and Timothy L. Lash. Modern Epidemiology. 3rd ed., Lippincott Williams & Wilkins, 2008.
Benjamini, Yoav, and Yosef Hochberg. “Controlling the False Discovery Rate: A Practical and Powerful Approach to Multiple Testing.” Journal of the Royal Statistical Society: Series B, vol. 57, no. 1, 1995, pp. 289–300.