| _version_ | 1866902057819045888 |
|---|---|
| author | Adeliyi, Olutomiwa Adebaminle |
| author_facet | Adeliyi, Olutomiwa Adebaminle |
| contents | <p>This literature review synthesises current academic research on the prevalence, causes, and mitigation of bias and hallucination in business focused applications of large language models (LLMs). It explores the forms and sector-specific manifestations of both implicit and explicit biases, discussing their potential to influence business outcomes and decision-making. The review also examines existing approaches to hallucination detection, highlighting how natural language processing (NLP) techniques are used to improve the factual accuracy of LLM outputs in business contexts. In evaluating current artificial intelligence governance frameworks, the study observes variation in their effectiveness across sectors and highlights the need for more targeted and practical approaches. A suggested set of ethical guidelines addresses transparency, accountability, and trust, which are important principles for the responsible use of LLMs in business settings. The review also considers the role of explainable NLP methods in addressing issues of bias and hallucination. By improving the interpretability of model decisions, these techniques support better alignment between model outputs, user expectations, and regulatory requirements. The findings suggest that while LLMs offer promise for business applications, their ethical, transparent, and context-aware use is essential to support fairness and accuracy. This review outlines current challenges and provides direction for future research aimed at improving the responsible use of artificial intelligence in different business environments.</p> |
| format | Recurso digital |
| id | zenodo_https___doi_org_10_5281_zenodo_16607702 |
| institution | Zenodo |
| language | |
| publishDate | 2025 |
| publisher | Zenodo |
| record_format | zenodo |
| spellingShingle | Bias and Hallucination in Business-Focused LLM Applications: Risks and Mitigation Adeliyi, Olutomiwa Adebaminle <p>This literature review synthesises current academic research on the prevalence, causes, and mitigation of bias and hallucination in business focused applications of large language models (LLMs). It explores the forms and sector-specific manifestations of both implicit and explicit biases, discussing their potential to influence business outcomes and decision-making. The review also examines existing approaches to hallucination detection, highlighting how natural language processing (NLP) techniques are used to improve the factual accuracy of LLM outputs in business contexts. In evaluating current artificial intelligence governance frameworks, the study observes variation in their effectiveness across sectors and highlights the need for more targeted and practical approaches. A suggested set of ethical guidelines addresses transparency, accountability, and trust, which are important principles for the responsible use of LLMs in business settings. The review also considers the role of explainable NLP methods in addressing issues of bias and hallucination. By improving the interpretability of model decisions, these techniques support better alignment between model outputs, user expectations, and regulatory requirements. The findings suggest that while LLMs offer promise for business applications, their ethical, transparent, and context-aware use is essential to support fairness and accuracy. This review outlines current challenges and provides direction for future research aimed at improving the responsible use of artificial intelligence in different business environments.</p> |
| title | Bias and Hallucination in Business-Focused LLM Applications: Risks and Mitigation |
| url | https://doi.org/10.5281/zenodo.16607702 |