AI Ethics and Social Norms: Exploring ChatGPT's Capabilities From What to How

Bias and transparency are major concerns: The study identified "significant obstacles related to transparency and bias in unsupervised data collection methods" as key ethical issues with ChatGPT, suggesting the AI system may perpetuate biases from its training data without clear visibility into how this happens.

Six critical ethics dimensions need evaluation: The research identified six important aspects of AI ethics that should be assessed when deploying AI tools like ChatGPT in real-world settings: bias, trustworthiness, security, toxicology (harmful outputs), social norms, and ethical data practices—providing a framework for responsible AI evaluation.

Expert and public perspectives both matter: The study combined insights from 111 survey participants and 38 expert interviews, revealing that understanding AI ethics requires input from both everyday users and specialists, not just technical developers alone.

Social norms and ethics must guide AI deployment: The research emphasizes that before using AI systems in sensitive domains like healthcare and collaborative work environments, organizations must examine whether these tools actually operate according to established ethical principles and social norms that humans expect.

Current AI systems may not be ready for high-stakes applications: The identification of multiple unresolved ethical concerns suggests that deploying ChatGPT and similar tools in critical areas like healthcare requires addressing these foundational issues first to ensure safe human integration.