Abstract

This chapter explores how to ensure integral accountability in AI systems, particularly in relation to fragile stakeholders—those whose inherent or structural characteristics (e.g., age, disability) make them especially susceptible to harm. The chapter introduces a dual modality of accountability: backward-looking (answerability for past actions) and forward-looking (commitment to preventing future harm), with the second being enrooted into an ethical-relational proactive organizational strategy (ongoing exchange in stakeholder relationships). Following the humanistic perspective, we argue that integral accountability concerning fragile stakeholders must incorporate all three. In order to discuss how to pursue integral accountability in AI ethics, the chapter discusses current frameworks of anticipatory ethics which enable a shift from reactive to proactive accountability (the precautionary principle, the ethical foresight analysis, the NEAAT framework). Through the discussion of illustrative cases taken from real-world ethical controversies on the use of AI, the chapter then discusses how organizations should provide a specific anticipatory ethics focus on at least three categories of fragile stakeholders: minors, seniors, people with special needs. Minors require that the organization balance the quality of their digital experience with their safety, moving beyond legal compliance to include ethical design, empowerment, and participation. Seniors require that the organization identifies and overcomes ethical challenges such as algorithmic ageism, exploitation, and over-dependence on AI care systems; the neglecting of such ethical challenges can turn beneficial AI from for seniors into potential harm and vice versa. While AI for people with special needs poses to the organization similar challenges to other fragile stakeholders, specific ethical dilemmas can arise in contexts of specific needs tied to this fragility. Overall, the chapter contributes to current knowledge on AI accountability and anticipatory ethics frameworks, by introducing and discussing the concept of the “preferential option for the fragile,” which builds upon the CST notion of the “preferential option for the poor.” This principle asserts that organizations should prioritize fragile stakeholders in their AI anticipatory ethics frameworks, so to provide integral human accountability to these groups. By applying the principle of the preferential option for the fragile, technologies are designed to meet the highest standards of safety and inclusivity. As a result, if an AI system is safe and accessible for fragile stakeholders, it is inherently more robust, ethical, and safe for all users. In this way, safeguarding the most vulnerable sets a universal benchmark for responsible and human-centered innovation.