Umeå University
As AI technologies advance, ensuring their transparency, scalability, and ethical governance becomes increasingly vital. Traditional AI paradigms, rooted in a rational framework [9], prioritize efficiency, problem-solving, and optimization. This paradigm emphasizes autonomy and a task-oriented approach, but often leads to AI systems that are monolithic and opaque. Moreover, this approach neglects crucial aspects such as societal impact, ethical considerations, and long-term maintainability [10]. To address these gaps, we propose integrating relational ethics, exemplified by feminist philosophies, with compositional AI principles from software engineering. This integration emphasizes community, interconnectedness, modularity, and transparency, ensuring that AI systems are both socially responsible and technologically robust, capable of meeting the complex demands of modern socio-ecological systems. That is, by embracing a structured, modular approach and a relational, community-centric ethic, AI can better meet the complex demands of modern socio-ecological systems, ensuring responsible and sustainable technological development. Current AI systems, particularly Generative AI (GenAI), are designed to prioritize immediate performance over long-term maintainability and ethical considerations. This leads to several issues, including complexity and opacity, as the interconnected nature of large AI models makes them difficult to navigate and understand, leading to challenges in verifiability and governance. Additionally, these AI approaches often reinforce existing power structures and fail to account for societal biases, as they are designed from a rational, individualistic perspective [1]. Furthermore, the societal impact of AI, including its potential to reinforce gender and racial biases, is not adequately addressed in current paradigms [2]. To address these critical limitations, a new paradigm is needed that shifts the focus from rationality to sociality [4]. Feminist epistemology challenges the traditional, individualistic concept of AI by emphasizing empathy, sensitivity, and community. It aims to include principles of accountability, responsibility, and transparency in AI design and use, focusing on how AI systems can affect and be affected by societal values and power structures [1, 6]. Other philosophical schools, such as Ubuntu, a non-Western philosophy rooted in African traditions, also emphasize interconnectedness, community, and the idea that one’s humanity is tied to the humanity of others. Such alternative paradigms promote norms of reciprocity, selflessness, and symbiosis, advocating for an AI that enhances communal relationships and is sensitive to the societal context in which it operates and help mitigate biases and promote social justice [5]. At the same time, to address the technical limitations of current AI systems, a shift is needed towards modular AI systems that draw on established software engineering principles such as modularity, abstraction, and separation of concerns [3]. This compositional AI paradigm promises greater flexibility, transparency, and integration with human expertise. Decomposing complex AI systems into smaller, manageable components allows for independent development, testing, and optimization, enhancing flexibility and transparency. Simplifying interfaces and protecting data ensures that AI systems are robust, reliable, and maintainable [10]. Combining AI with human expertise and analog models adds robustness and contextual relevance, especially in scenarios with significant uncertainty or ethical considerations. Simplifying interfaces and protecting data ensures that AI systems are robust, reliable, and maintainable [10]. Combining AI with human expertise and analog models adds robustness and contextual relevance, especially in scenarios with significant uncertainty or ethical considerations. By integrating relational ethics with compositional AI principles, we can develop AI systems that are both socially responsible and technologically robust. This integration involves embedding ethical considerations, transparency, and accountability into the core design of AI systems, ensuring they align with human values and societal needs [7]. Furthermore, cross-disciplinary collaboration in the development and deployment of AI systems is essential to address to the complex interactions and dependencies characteristic of modern socio-technicalecological systems [8]. Despite its potential, developing a modular, hybrid, and human-centric AI paradigm presents several significant challenges. Managing interactions between distributed nodes and ensuring data privacy and security require advanced techniques and robust infrastructure. Developing transparent and verifiable AI operations, including explainable AI techniques, is essential for building trust and accountability. To effectively integrate relational ethics with a compositional approach to AI development and use, the following characteristics and properties are essential: By integrating these characteristics and properties, we can develop AI systems that are both socially responsible and technologically robust. However, several challenges must be addressed to realize this vision. Ensuring transparency and accountability requires overcoming the complexity of AI systems, which can obscure decision-making processes. Embedding ethical considerations into AI design involves tackling ingrained biases and ensuring fairness across diverse societal contexts. Achieving scalability and maintainability without sacrificing performance demands innovative engineering solutions. Furthermore, fostering cross-disciplinary collaboration requires bridging gaps between different fields and ensuring effective communication. Despite these challenges, this integrated approach fosters a more inclusive and sustainable technological future, addressing the multifaceted challenges of modern socio-ecological systems and ensuring the responsible and ethical use of AI. Adopting an ethical relational approach to AI, combined with a structured, engineering-focused compositional paradigm, will lead to the development of AI systems that are not only powerful and efficient but also aligned with human values and societal needs. This integrated approach paves the way for a future where AI serves as a force for positive change and inclusivity.
References