The advent of ChatGPT, a sophisticated chatbot using deep learning, marks a significant AI advancement. Its ability to generate human-like text promises to revolutionize various sectors. However, this progress necessitates a careful consideration of potential pitfalls. While ChatGPT offers benefits in customer service, education, and other fields, its irresponsible use raises ethical concerns regarding privacy, inclusivity, inequality, bias, safety, and even existential risks. This review aims to identify these risks and propose an ethical framework for the sustainable development and use of chatbots. The study focuses on two key research questions: What are the perils of irresponsible ChatGPT development and implementation? And, how can a multi-stakeholder RRI framework guide the sustainable development and use of chatbots?

The authors conducted a focused literature review using Scopus and Google Scholar, employing keywords related to ChatGPT, RRI, responsible development, perils, challenges, and implications. After applying inclusion criteria (English language, publication years 2011-2023, full-text articles, academic journals and conference papers), a final pool of 72 articles was selected for analysis. The review synthesized existing literature on the ethical considerations of AI and RRI principles to address the research questions.

This study employed a focused literature review, a methodology characterized by its concise examination of a specific subset of literature relevant to the research questions. This approach prioritized identifying and synthesizing key findings rather than conducting an exhaustive analysis of the entire research domain or using bibliometric analyses. The search strategy involved using keyword combinations in Scopus and Google Scholar databases, resulting in 1892 initial records. After applying inclusion and exclusion criteria (English language, publication dates, full-text availability, peer-reviewed articles from academic journals and conference proceedings), the number of records was reduced to 118. Further filtering for duplicate records resulted in 60 unique eligible records. A cited search, including a snowball technique, added another 12 records for a final pool of 72 articles. These articles were then reviewed to address the two research questions: mapping the perils of irresponsible ChatGPT implementation and proposing a multi-stakeholder RRI framework.

The review identified significant risks associated with ChatGPT and similar AI tools across individual, organizational, and societal levels. At the individual level, concerns include the spread of misinformation, targeting vulnerable individuals with harmful content, privacy violations, and the potential devaluation of human relationships due to AI's capacity to mimic human interaction. At the organizational level, risks involve labor market disruptions (job displacement), intellectual property theft, cybersecurity threats, and damage to brand reputation. Societal-level concerns encompass the negative impact on education and research (compromised academic integrity, reduced creativity), exacerbation of digital inequities, and ethical dilemmas in healthcare (system errors, privacy breaches). To mitigate these risks, the study advocates for a multi-stakeholder RRI framework based on stakeholder theory. This framework emphasizes proactive risk assessment (anticipation), inclusive stakeholder engagement (inclusiveness), reflective practice (reflexivity), and responsiveness to societal needs and concerns. The framework integrates the AI innovator, regulators, and direct and indirect stakeholders, promoting collaborative decision-making and ethical considerations throughout the AI lifecycle. Specific guidelines are suggested, including prioritizing safety and security, incentivizing responsible AI development, harmonizing AI with human values (autonomy, agency, privacy), and investing in research on the societal impacts of AI.

The findings underscore the urgent need for a responsible approach to AI development and deployment. The identified perils of irresponsible AI implementation highlight the necessity of ethical guidelines and frameworks. The multi-stakeholder RRI framework proposed in the study offers a practical and theoretically grounded approach to address these challenges. By integrating stakeholder theory and RRI principles, the framework facilitates proactive risk management, inclusive decision-making, and alignment with societal values. The discussion emphasizes the need for a collaborative effort involving AI developers, regulators, policymakers, and the public to ensure that AI technologies are developed and used ethically and sustainably.

This focused review contributes to the ongoing debate surrounding the ethical implications of AI, specifically focusing on ChatGPT. The authors highlight the significant risks of irresponsible AI development and use across individual, organizational, and societal levels. They propose a multi-stakeholder RRI framework to guide the ethical and sustainable development of AI. The framework incorporates established theories and principles, offering a practical tool for stakeholders to proactively address the challenges associated with advanced AI. Future research should empirically explore the efficacy of different stakeholder engagement approaches and investigate the effectiveness of regulatory measures in promoting responsible innovation in the AI domain.

As a focused literature review, this study's scope is limited to the available literature and the authors' interpretation. The findings are based on a specific selection of articles and may not represent the entire range of viewpoints on the topic. The proposed RRI framework is a conceptual model; its practical implementation and effectiveness require further empirical investigation. Furthermore, the rapidly evolving nature of AI technology means that new challenges and risks may emerge that are not fully addressed in this review.