HUMAN-IN-THE-LOOP PREDICTIVE ANALYTICS: A COMPREHENSIVE SYSTEMATIC REVIEW OF EXPERT-IN-THE-LOOP FEEDBACK MECHANISMS FOR ENHANCING MODEL ACCURACY, ENSURING ALGORITHMIC FAIRNESS, AND STRENGTHENING ADVERSARIAL ROBUSTNESS IN MODERN AI SYSTEMS
International Journal of Grid Computing & Applications · 28.03.2026 · 17 (1) · 123-136
DOI: 10.5121/ijgca.2026.17112 ↗Kurz gefasst
Systematische Übersicht in der Zeitschrift International Journal of Grid Computing & Applications, online erschienen am 28.03.2026. Verfasst von Sungho Kim und 9 weiteren Autor:innen.
Abstract (Originalsprache)
Despite remarkable advances in automated machine learning, purely algorithmic predictive systems remain susceptible to three persistent failure modes: prediction inaccuracy arising from distributional mismatch, latent bias encoded from non-representative training corpora, and brittleness under adversarial or out-of-distribution inputs. Human-in-the-Loop (HITL) machine learning proposes to remediate these failure modes by systematically integrating structured expert feedback into model training and evaluation cycles. This paper presents a systematic review of 74 peer-reviewed studies (2018–2025) examining HITL mechanisms across five application domains: healthcare, financial risk, NLP/sentiment analysis, image recognition, and predictive maintenance. We find that HITL-guided systems achieve mean accuracy improvements of 12.6 percentage points over fully automated baselines, reduce demographic parity violations by up to 82.3%, and maintain above-50% accuracy under adversarial perturbations where unaided baselines collapse below 14%. We further provide a formal taxonomy of feedback modalities, analyse cost-accuracy trade-offs in active learning protocols, and identify scalability and annotator-bias propagation as the principal open challenges. Our findings substantiate that HITL integration is not merely an engineering convenience but a necessary condition for deploying ethically sound and operationally robust predictive analytics in high-stakes settings.
Zitieren
APA
Kim, S., Rahman, M. M., Karim, M. A., Chowdhury, A., Emon, A. I., Rahman, M., Niha, I. A., Hellol, H. I., Khondaker, R., & Ridwan, M. A. (2026). HUMAN-IN-THE-LOOP PREDICTIVE ANALYTICS: A COMPREHENSIVE SYSTEMATIC REVIEW OF EXPERT-IN-THE-LOOP FEEDBACK MECHANISMS FOR ENHANCING MODEL ACCURACY, ENSURING ALGORITHMIC FAIRNESS, AND STRENGTHENING ADVERSARIAL ROBUSTNESS IN MODERN AI SYSTEMS. International Journal of Grid Computing & Applications, 17(1), 123-136. https://doi.org/10.5121/ijgca.2026.17112
Vancouver
Kim S, Rahman MM, Karim MA, Chowdhury A, Emon AI, Rahman M, et al. HUMAN-IN-THE-LOOP PREDICTIVE ANALYTICS: A COMPREHENSIVE SYSTEMATIC REVIEW OF EXPERT-IN-THE-LOOP FEEDBACK MECHANISMS FOR ENHANCING MODEL ACCURACY, ENSURING ALGORITHMIC FAIRNESS, AND STRENGTHENING ADVERSARIAL ROBUSTNESS IN MODERN AI SYSTEMS. International Journal of Grid Computing & Applications. 2026;17(1):123-136. doi:10.5121/ijgca.2026.17112
BibTeX
@article{Kim2026,
author = {Kim, Sungho and Rahman, Mohammad Mahmudur and Karim, Mohammad Abdul and Chowdhury, Ashaduzzaman and Emon, Animul Islam and Rahman, Marjia and Niha, Ibtisum Ahmed and Hellol, Hamim Islam and Khondaker, Riffat and Ridwan, Md Al},
title = {HUMAN-IN-THE-LOOP PREDICTIVE ANALYTICS: A COMPREHENSIVE SYSTEMATIC REVIEW OF EXPERT-IN-THE-LOOP FEEDBACK MECHANISMS FOR ENHANCING MODEL ACCURACY, ENSURING ALGORITHMIC FAIRNESS, AND STRENGTHENING ADVERSARIAL ROBUSTNESS IN MODERN AI SYSTEMS},
journal = {International Journal of Grid Computing & Applications},
year = {2026},
volume = {17},
number = {1},
pages = {123--136},
doi = {10.5121/ijgca.2026.17112}
}
EUVIMED