Sampath Mandava (University of Colorado) has posted Bridging Human Values and Machine Goals: Advances in AI Alignment and Safe Autonomy on SSRN. Here is the abstract:
The expansion of artificial intelligence (AI) and autonomous systems is influencing the alignment of machine objectives and autonomy with the safety and ethical value of human trust. This study examines the recent literature on value alignment within AI and autonomous safe systems through a multidisciplinary lens. This work investigates the emerging works and models on the incorporation of moral reasoning, equity, and accountability into machine learning. There is a well-developed ethical AI governance model, including the moral and legal boundaries of AI-driven decision-making (Kamaldeen, 2023), the human-aware AI systems that understand and react to human emotions and intentions (Mechergui & Sreedharan, 2024), and the socio-technical frameworks that human-centered design and the socio-technical frameworks for responsible innovation (Shneiderman, 2022). They illustrate the new societal framework of AI and human cooperation on collaborative intelligence-the co-evolution of ethical AI and AI systems. This study emphasises the integration of human value frameworks and machine objectives, achieving cross-disciplinary work in ethics, law, data governance, and technology (Nay, 2022). An AI autonomous system will retain coherence if it is morally aligned with the algorithms, and morally autonomous AI ebb and flow with human-centred moral autonomy. Stakeholder This paper focuses on a multidisciplinary audience which includes AI policymakers, researchers, ethicists, legal scholars, and technology developers concerned with the development and deployment of human-centred and safe autonomous systems. It focuses on the development of governance structures which ensure accountability and transparency in the development of AI (Kamaldeen, 2023), the deployment of ethical AI in the promotion of equity and inclusiveness (Ayinla et al., 2024), and the integration of human-centred values in the design and learning processes of intelligent systems (Han et al., 2022). Moreover, the paper targets all academics and practitioners in the responsible AI value chain in infrastructure core areas, including healthcare, finance, and mobility (Holzinger et al., 2024). Grounded in the integration of ethics, law, and socio-technical systems, the research seeks to help decision-makers and human welfare innovators as the basis of trust, and ensure autonomous systems function safely in line with societal expectations and norms.