ChatGPT and trustworthiness of artificial intelligence in medical education

Abstract

Dear Editor, The case study on adopting ChatGPT to advance competency-based medical education in attitude, ethics, and communication (AETCOM) knowledge and skills is of great interest.[1] Educational research leveraging artificial intelligence (AI) technology can help in tackling systemic challenges in medical education. However, the following observations merit further deliberation. Methodological Considerations First, given the rapid evolution of large language models (LLMs), specific details of the GPT usage in the study are important for robustness of the methodology. This includes mentioning details such as OpenAI (2023) ChatGPT-3.5 May 24 version without plugins and the use of regenerated responses to enhance reproducibility of the results. Furthermore, benchmarking against other prominent LLMs like Google Bard (renamed as Gemini) or Anthropic Claude would have significantly enhanced the overall scientific impact.[2] Second, knowledge and skills in the AETCOM framework are intricate to the doctor–patient relationships (for example, Case 2: conflict between autonomy and beneficence). These competencies are grounded on fostering trust developed through rigorous pedagogical approaches established over the years. Implementing technological solutions that lack stakeholder validation at design and development stages could be potentially imperious. It is important to keep in mind that the current study primarily is not designed to project the nuances of adult learning, narrative medicine as well as training of virtuous medical professionals.[3] Third, most LLMs rely largely on Western datasets with embedded cultural biases and a risk of perpetuating such biases within the local contexts.[4] While GPT-3, the underlying model for ChatGPT, was trained on a diverse range of internet text, including sources from various countries and cultures, it is essential to note that biases can still exist in the data used for training. Multifaceted Approach to AI Trustworthiness The current eagerness to integrate AI technologies like LLMs into medical education before addressing trustworthiness considerations merits caution. Medical education aims not just to transfer information but nurture wisdom, insight, empathy, and ethics—uniquely human capabilities that AI is yet to develop sufficiently.[5,6] Hence, ChatGPT integration in AETCOM sounds premature in the absence of framework-based safeguards. Trustworthiness of AI goes beyond ethical principles to embeds in itself, a multidimensional perspective spanning technical dimensions of reliability, accuracy, reproducibility, and risk mitigation as well as social aspects of fairness, inclusivity and cultural sensitivity.[7] Adherence to relevant laws and regulations governing data, and algorithmic systems safeguards legal dimension. Finally, the intertwining ethical components ties it all together. Trustworthiness of AI for higher education therefore demands an evolving, holistic approach constructed on the fundamental pillars of trust (lawfulness, ethics, and robustness).[7] Conclusion Rather than taking a rapid techno-solutionist approach to systemic issues like faculty shortages, we as educationalists must act as gatekeepers to ensure integration of trustworthy AI systems. For instance, interprofessional collaboration through design, development, and contextualized real-world evaluations, students codesigning educational interfaces could build trust over time. While AI promises to transform learning, current tools like ChatGPT have limitations of factual inaccuracies and bias which is well acknowledged in the study. However, the proposal of ChatGPT as a replacement for medical teachers overlooks compromises in teaching nuanced judgment skills vital for patient-centered care and nurturing of emotionally intelligent clinicians. AI integration with time-honored educational values requires accountable adoption with consideration of ethical consequences. Therefore, a strategic approach to using AI to augment human capabilities would be more beneficial than replacing medical educators altogether.[3–5] Financial support and sponsorship Faculty research grant from the National University of Science and Technology, Oman to develop an artificial intelligence technology-based tool to train medical students in diagnostic reasoning. Conflicts of interest There are no conflicts of interest.

Ähnliche Arbeiten