Trustworthy Artificial Intelligence for High-Stakes Decision-Making: Explainability, Human Oversight, and System Reliability
International Journal of Engineering & Extended Technologies Research · 15.08.2023 · 5 (4)
DOI: 10.15662/ijeetr.2023.0504011 ↗Kurz gefasst
Fachartikel in der Zeitschrift International Journal of Engineering & Extended Technologies Research, online erschienen am 15.08.2023. Verfasst von Ganesh Sai Kopparthi.
Abstract (Originalsprache)
The world of Artificial Intelligence (AI) has several areas where AI is making a strong presence in decision-making processes, such as health, finance, criminal justice, and even public administration. The upside of the benefits of AI is that it may be able to streamline processes, ensure consistency, and make more accurate analytical decisions, but there were concerns of transparency, accountability, fairness and reliability, especially when it comes to decisions with impact. This study dwells on three areas of trustworthy AI: explanation, human oversight and the reliability of systems, that are closely related and necessary pre-conditions for responsible decision making. It talks about the possibilities of Explanable AI to empower decision-makers to understand how algorithms work, and what may be going wrong if they don't, and to audit the rationale for automated recommendations. The need for human participation in the activity of AI, ethics and intervention aspects of decisions taken by AI in the scope of professional judgment would be the focus of another aspect of this research. Moreover, it's a test and evaluation of the reliability of the system measured by accuracy, robustness, consistency and performance under various conditions. The research is conceptual, leveraging on existing academic research, guidelines and principles for governance of AI and relevant regulatory frameworks. According to it, transparency of processes, job descriptions for humans and regular audits are essential features of successful business usage of AI, as is appropriate risk-management mechanisms. It is clear that organisations need to create evaluation criteria that are contextually appropriate, know what they can and cannot do in their systems and establish processes of assessing and addressing any negative impacts. It finds that explainability, supervision by humans, and reliability of the AI system need to be built into the design of the AI system from its early stages and implemented throughout the service life. These dimensions can help to build stakeholder trust, improve transparency in decision-making processes and promote responsible use of AI, especially in sensitive environments.
Zitieren
APA
Kopparthi, G. S. (2023). Trustworthy Artificial Intelligence for High-Stakes Decision-Making: Explainability, Human Oversight, and System Reliability. International Journal of Engineering & Extended Technologies Research, 5(4). https://doi.org/10.15662/ijeetr.2023.0504011
Vancouver
Kopparthi GS. Trustworthy Artificial Intelligence for High-Stakes Decision-Making: Explainability, Human Oversight, and System Reliability. International Journal of Engineering & Extended Technologies Research. 2023;5(4). doi:10.15662/ijeetr.2023.0504011
BibTeX
@article{Kopparthi2023,
author = {Kopparthi, Ganesh Sai},
title = {Trustworthy Artificial Intelligence for High-Stakes Decision-Making: Explainability, Human Oversight, and System Reliability},
journal = {International Journal of Engineering & Extended Technologies Research},
year = {2023},
volume = {5},
number = {4},
doi = {10.15662/ijeetr.2023.0504011}
}
EUVIMED