Explainable and Ethical Artificial Intelligence Using Language Models

The previous article discussed the concept of explainability through AI models from two angles - the technical aspect of implementing models that can be easily read by non-experts using natural language generation, and the governance aspect of controlling AI so that it remains under the mandate of genuine human control, both of which are important, but together do not completely solve the problem. Technically speaking, Renze (2018) achieved 0.996 as the factual accuracy of the self-correcting GPT pipeline, which nearly meets the accuracy of a rule-based hand-coded system, while still providing a more flexible and natural experience. Governance-wise, James (2020) conducts a survey which provides evidence of a shift in public sentiment, and as such, James identified four specific governance interventions; scoped training, decentralized architecture, model-approval committees, and organizations providing human-in-the-loop requirements for high stakes uses. The literature in the field of explainable AI evaluated in this article indicates that the two threads discussed are indeed converging, as there seems to be a consensus in the literature that the provision of explanation through LLM-based models will be the most promising means of reaching explainable AI, and the noted governance frameworks will become binding law. The combination of researchers/practitioners who are able to bridge both threads and build systems that cover both aspects has the best opportunity to successfully implement AI that is truly explainable, truly ethical, and truly trustworthy.

A. J. James, "Human oversight and control in AI development and deployment," Technology in Society, Preprint submitted May 31, 2023. M. Renze, "Natural language explanations for explainable AI systems using large language models with self-correcting architectures," arXiv preprint, Johns Hopkins University, 2023. S. R. Islam, W. Eberle, S. K. Ghafoor, and M. Ahmed, "Explainable artificial intelligence approaches: A survey," arXiv:2101.09429, 2021. M. Nagahisarchoghaei et al., "An empirical survey on explainable AI technologies: Recent trends, use-cases, and categories from technical and application perspectives," Electronics, vol. 12, no. 5, p. 1092, 2023. J. Burrell, "How the machine 'thinks': Understanding opacity in machine learning algorithms," Big Data & Society, vol. 3, no. 1, 2016. S. Nazir, D. M. Dickson, and M. U. Akram, "Survey of explainable artificial intelligence techniques for biomedical imaging with deep neural networks," Computers in Biology and Medicine, vol. 156, 2023. A. B. Arrieta et al., "Explainable Artificial Intelligence (XAI): Concepts, taxonomies, opportunities and challenges toward responsible AI," Information Fusion, vol. 58, pp. 82–115, 2020. A. F. Markus, J. A. Kors, and P. R. Rijnbeek, "The role of explainability in creating trustworthy artificial intelligence for health care," Journal of Biomedical Informatics, vol. 113, 2021. A. Zytek, S. Pidò, S. Alnegheimish, L. Berti-Équille, and K. Veeramachaneni, "Explingo: Explaining AI predictions using large language models," arXiv:2412.05145, 2024. P. Biecek et al., "Explainable AI (XAI) 2.0: A manifesto of open challenges and interdisciplinary research directions," Information Fusion, vol. 106, 2024. S. M. Lundberg and S.-I. Lee, "A unified approach to interpreting model predictions," in Advances in Neural Information Processing Systems, vol. 30, pp. 4765–4774, 2017.