One of the approaches that has gained attention in recent decades for embedding social values in technology is the Value-Sensitive Design (VSD) approach. This idea is based on the premise that technology is not "value-neutral," but rather "value-laden." Although this approach has generally been considered in discussions of designing and embedding values in technological systems, due to the unique characteristics of artificial intelligence, its generalization to AI-based systems and its implementation and execution have encountered various challenges.

This paper, employing an analytical-inferential method within an interdisciplinary framework, focuses on the methodology of embedding ethical and social values in artificial intelligence technology and its challenges. The research findings indicate that in the pursuit of embedding ethical-social values in AI, two of its unique features—self-learning and self-decision-making—create challenges at different levels of value-sensitive design. These challenges include ambiguity in defining and prioritizing values, implicit marginalization of some groups, lack of transparency, opacity, and unexplainability of systems, the phenomenon of value disembody, difficulty in translating ethical concepts into technical requirements, and the complexity of assessing the real-world impact of values in practice. To address these challenges, suggestions such as adopting a hybrid structure in Value Sensitive Design, developing regulatory and evaluation frameworks, enhancing AI literacy, and implementing mechanisms for monitoring intents are proposed. This research’s aim is to contribute to a deeper understanding of the obstacles and challenges in embedding values and ethics in AI by clarifying and conceptually expanding the Value-Sensitive Design approach in AI. Furthermore, by emphasizing the importance of interdisciplinary synergy, it seeks to outline practical strategies for the development of responsible AI aligned with social values.