Ethical dimensions of generative AI: a cross-domain analysis using machine learning structural topic modeling

• Date: 05 September 2024
• Pages: 3-34
• DOI: https://doi.org/10.1108/IJOES-04-2024-0112
• Published date: 05 September 2024
• Subject Matter: Economics,Social economics
• Author: Hassnian Ali,Ahmet Faruk Aysan

Ethical dimensions of generative AI:

a cross-domain analysis using

machine learning structural

topic modeling

Hassnian Ali

College of Islamic Studies, Hamad Bin Khalifa University, Doha, Qatar and

International Center for Research in Islamic Economics, ICRIE,

Minhaj University Lahore, Lahore, Pakistan, and

Ahmet Faruk Aysan

College of Islamic Studies, Hamad Bin Khalifa University, Doha, Qatar

Abstract

Purpose –The purpose of this study is to comprehensively examine the ethical implications surrounding

generativeartif‌icial intelligence (AI).

Design/methodology/approach –Leveraging a novel methodological approach, the study curatesa corpus of

364 documents from Scopus spanning 2022 to 2024. Using the term frequency-inversedocument frequency (TF-

IDF) and structural topic modeling (STM), it quantitatively dissectsthe thematic essence of the ethical discourse

in generative AI across diverse domains,including education, healthcare, businesses and scientif‌ic research.

Findings –The results reveal a diverse range of ethical concerns across various sectors impacted by

generative AI. In academia, the primary focus is on issues of authenticity and intellectual property,

highlightingthe challenges of AI-generated content in maintaining academicintegrity. In the healthcare sector,

the emphasis shiftsto the ethical implications of AI in medical decision-making and patientprivacy, ref‌lecting

concerns aboutthe reliability and security of AI-generated medical advice. The study also uncoverssignif‌icant

ethical discussions in educationaland f‌inancial settings, demonstrating the broad impact of generative AI on

societaland professional practices.

Research limitations/implications –This study provides a foundation for crafting targeted ethical

guidelines and regulations for generative AI, informedby a systematic analysis using STM. It highlights the

need for dynamic governance and continual monitoringof AI’s evolving ethical landscape, offering a model

for future researchand policymaking in diverse f‌ields.

Originality/value –The study introduces a unique methodological combination of TF-IDF and STM to

analyze a large academic corpus, offering new insights into the ethical implications of generative AI across

multiple domains.

Keywords Generative AI, Ethics, Structure topic modeling, Governance, Regulation

Paper type Research paper

1. Introduction

In the broader spectrum of artif‌icial intelligence (AI), generative AI stands at the forefront,

revolutionizing how machines understand, interpret and generate human-like text, images

and even codes (Sauvola et al., 2024).Due to its generalizability and accessibility, the nature

JEL classif‌ication –I23, I31, K11, O33.

International

Journal of Ethics

and Systems

3

Received23 April 2024

Revised2 June 2024

17July 2024

Accepted22 July 2024

InternationalJournal of Ethics and

Systems

Vol.41 No. 1, 2025

pp. 3-34

© Emerald Publishing Limited

2514-9369

DOI 10.1108/IJOES-04-2024-0112

The current issue and full text archive of this journal is available on Emerald Insight at:

https://www.emerald.com/insight/2514-9369.htm

of generative AI is different from other AI models and types. The rapid and mass-level

adoption of generative AI technologies has brought to light the intricate web of ethical

considerations that accompany these innovations (Ray, 2023a). As these systems become

increasingly capable of producing indistinguishable content from that created by humans,

questions of authorship, intellectual property (IP), bias and potential misuse become

paramount (Klenk, 2024). This intersection of generative AI and ethics forms a critical

domain of inquiry, necessitating a comprehensive examination of the intellectual discourse

surrounding these issues (Carnevaleet al., 2023). Ethical aspects of AI are important in this

context, as they ensure that AI development aligns with principles such as transparency,

fairness, accountabilityand privacy, which are essential forgaining public trust and avoiding

harmful consequences (Kazim and Koshiyama, 2021). Furthermore, integrating ethical

considerations at all stages of AI development, from conceptual design to deployment and

maintenance, addresses issuessuch as bias, data privacy and the impact on human rights and

societal structures (Mittelstadt, 2019). Ethical guidelines and frameworks are necessary to

balance AI’s immense potential with the need to prevent misuse and ensure equitable

benef‌its, highlighting the critical importance of substantive ethical analysis and

implementation strategies (Jobin et al., 2019). The existing studies on generative AI and

ethics are either domain-specif‌ic, such as academia (Lund et al., 2023), healthcare (Chen

et al.,2023) or discuss the ethics of using ChatGPT (Stahl and Eke, 2024). The research

lacks a robust methodologyof literature review and depth of discussion. The importanceand

objective of this study lies in its attempt to unravel the complex ethical landscape of

generative AI discussed in academic discourse, providing insights that could guide the

development, deployment and governanceof these technologies in a manner that aligns with

societal values and norms.

The novelty of the research lies in its methodicalapproach to data collection and analysis,

leveraging Scopus’s extensive database to curate a relevant corpus of literature spanning

from 2022 to 2024 on f‌inally selected 361 documents. Using term frequency-inverse

document frequency (TF-IDF) metrics (Bai et al., 2021) and structural topic modeling

(STM) (Roberts et al.,2013), the research transcends traditional bibliometric studies by

quantitatively dissecting the thematic essence of the discourse on ethics in generative AI.

This analytical rigor ensures the identif‌ication of pertinent topics, elucidating the semantic

landscape that def‌ines this evolving f‌ield and discussing general and domain-specif‌ic ethical

issues.

The f‌indings comprehensively analyze the ethical dimensions associated with generative

AI technologies across diverse domains, including art, healthcare, education and scientif‌ic

research. The research identif‌ies ten distinct topics encapsulating the ethical discourse

around generative AI by employing STM on an intellectual corpus derived from Scopus.

These range from concerns over authenticity and IP rights in AI-generated art to the

implications of AI in medical decision-making and patient privacy to the challenges and

opportunities of integrating AI in educational settings. Notably, the study highlights a

signif‌icant focus on the ethical use of generative AI in art and image creation, while also

emphasizing emerging concerns in patient care, indicating a dynamic and evolving ethical

landscape. In addition, the visualizations of word clouds, bar graphs, and scatter plots

illuminate the frequency and relevance of key terms within each topic and illustrate the

interconnectedness and distinct ethical considerations pertinent to various applications of

generative AI. This nuanced understanding of the ethical concerns, grounded in empirical

data analysis, underscores the complexityof navigating the ethical dimensions of generative

AI technologies and signals the need for tailored ethical guidelines and governance

frameworks across differentsectors.

IJOES

41,1

4

The contributions of this study are manifold. First, it delineates discrete ten topics that

capture the breadth of ethical discussions on generative AI, from its applications in art and

healthcare to its implicationsfor education and beyond. This thematic identif‌icationnot only

maps the intellectual contours of the f‌ield but also underscores the complex ethical

considerations inherent in diverse applications of generative AI technologies. Second, the

research provides a detailed exploration of these topics’distribution and structural nuances,

offering insight intothe dominant ethical dimensions and those that warrant further scholarly

attention. The systematic utilization of STM affords a structured overview of ethical issues,

serving as a navigational tool for scholars and practitioners alike in this intricate ethical

landscape. Third, and notably, the study extends beyond general ethical considerations to

dissect specif‌ic issues related to the use of generative AI across academia, healthcare and

f‌inance and businesses. Thisaspect of the study illuminates the distinct ethical challengesin

these domains, such as academic integrity and plagiarism in education, privacy and data

accuracy in healthcare, and transparency and employment in f‌inance and businesses,

providing a granular understandingof domain-specif‌ic ethical dilemmas.

The rest of the document is structured as follows: Section 2 outlines the methodology,

detailing the data collection, preparation and processing stages. Section 4 presents

descriptive outcomes along with STM f‌indings. Section 5 interprets these results, framing

them around f‌ive key agendas, while the concluding section, Section 6 concludes the study,

discussing its implicationsand avenues for future research.

2. Method

2.1 Data identif‌ication, ref‌ining and extraction

Data identif‌ication and retrieval for our study on generative AI and ethics was conducted

using Scopus, a comprehensive database widely recognized for its extensive coverage of

peer-reviewedliterature suitable for quantitative analysis (Abubakar and Aysan, 2022;Aysan

et al., 2021;Aysan and Unal, 2023;Qadri et al.,2022).The study is based on secondary data

which is in the form of existing published research documents on the Scopus database. The

decision to use Scopus as our primary metadata source was informed by its broad scope,

encompassing over 20,000 peer-reviewed active titles and its capacity to support detailed

bibliometric analyses. This choice ref‌lects a preference for depth and quality of coverage,

acknowledging the platform’s advantages in facilitating nuanced scholarly research over

alternatives like Web of Science and Google Scholar. Scopus is particularly valued for its

inclusion in journal articles, conference proceedings and book series, offering a richer data

set for comprehensive bibliometric studies. Its extensive use in prior bibliometric research

underlines its suitability and effectiveness for such analyses (Baker et al., 2020;Das et al.,

2023;García-Lilloet al.,2019;Rialpet al., 2019;Tahir et al., 2023).

On March 22, 2024, a search for “Generative AI”OR “Gen AI”OR“ChatGPT”(in titles

only) AND “Ethics”OR “Ethical”(in the Titles, abstract and keywords) within Scopus

yielded 651 documents. The set period was 2022–2024. After ref‌iningour search criteria to

include only articles and reviews and excludingother document types to ensure consistency

and relevance to our research focus, the data set was narrowed down to 364 documents.

These selected publications, comprising a mix of articles and reviews, formed the basis of

our analysis. The stepsof the research process f‌low are illustrated in Figure 1.

2.2 Preprocessing of text and metadata

Based on the data extracted from Scopus, we carefully compiled a text corpus that included

the title, abstract, keywords and publicationyear of every research article published. We also

checked the impact of publication sources on the prominence of the topics by adding source

International

Journal of Ethics

and Systems

5