Research Article

Ethical Alignment of Large Language Models: A Multi-Stakeholder Preference Learning Approach for Bias Mitigation in News Generation

143 reads
J Ong Artific Int Innov, 2026, 1 (1), 48-53, ISSN

Abstract

Automated news generation powered by Large Language Models (LLMs) offers unprecedented scalability in journalistic workflows but risks propagating systemic social biases, political polarization, and factual distortions. Conventional alignment techniques, such as Reinforcement Learning from Human Feedback (RLHF) and Direct Preference Optimization (DPO), predominantly collapse normative value judgments into a single aggregated reward model, inadvertently marginalizing minority perspectives and failing to capture the complex, multi-faceted ethical standards demanded by professional journalism. In this paper, we propose a Multi-Stakeholder Preference Learning (MSPL) framework designed to achieve nuanced ethical alignment in automated news generation. Our approach disaggregates preference signals across three vital cohorts: professional journalists adhering to canonical editorial ethics, domain-specific fact-checkers evaluating epistemic fidelity, and demographically diverse reader panels representing varied socio-political identities. By formulating ethical alignment as a multi-objective constrained optimization problem along a Pareto frontier, MSPL balances journalistic integrity, epistemic accuracy, and representative neutrality. Empirical evaluations conducted on a novel benchmark of contentious political, socio-economic, and cultural news events reveal that MSPL reduces ideological and demographic bias by 38.4% relative to standard DPO and monolithic RLHF baselines, while simultaneously preserving factual precision and narrative fluency. These findings demonstrate the necessity of multi-stakeholder pluralism in aligning generative AI systems deployed within sensitive socio-technical ecosystems.

Keywords Bias mitigation large language models Ethical AI Alignment Multi-Stakeholder Preferences Automated Journalism
Authors 3

The team behind this paper

3 authors, 3 institutions.

This paper Technical University of Munich — Germany Technical University of… 1 author Kyoto University — Japan Kyoto University 1 author University of the Witwatersrand — South Africa University of the Witwa… 1 author Prof. Elena Rostova — corresponding author ER Prof. Elena Rostova ✉ Dr. Kenjiro Takahashi KT Dr. Kenjiro Takahashi Dr. Amara Okafor AO Dr. Amara Okafor

Readership

143 reads over 4 months.

August 2026 November 2026

Bibliographic Information

Prof. Elena Rostova, Dr. Kenjiro Takahashi, Dr. Amara Okafor, (2026). Ethical Alignment of Large Language Models: A Multi-Stakeholder Preference Learning Approach for Bias Mitigation in News Generation, Journal of Ongoing Artificial Intelligence Innovations, 1(1): 48-53
Bibtex Citation
@article{prof._elena_rostova2026joaii,
author = {Prof. Elena Rostova and Dr. Kenjiro Takahashi and Dr. Amara Okafor},
title = {Ethical Alignment of Large Language Models: A Multi-Stakeholder Preference Learning Approach for Bias Mitigation in News Generation},
journal = {Journal of Ongoing Artificial Intelligence Innovations},
year = {2026},
volume = {1},
number = {1},
pages = {48-53},
doi = {},
url = {https://scimatic.org/show_manuscript/9395}
}
APA Citation
Rostova, P.E., Takahashi, D.K., Okafor, D.A., (2026). Ethical Alignment of Large Language Models: A Multi-Stakeholder Preference Learning Approach for Bias Mitigation in News Generation. Journal of Ongoing Artificial Intelligence Innovations, 1(1), 48-53. https://doi.org/

Author Information

  • To change your profile photo, login to scimatic.org, go to your profile and change the photo.
  • Any photo that shows your face will do: we crop it to your head and shoulders and give it a plain background automatically.
  • To sign in, go to Sign in and enter the e-mail address registered with the article: we e-mail you a one-time sign-in link (there is no password).
  • In case of any other problem, contact your editor directly or write to us at info @ scimatic.org