TRANSFORMATIONS IN THE GOVERNANCE AND MANAGEMENT OF PUBLIC ORGANIZATIONS TO IMPLEMENT ARTIFICIAL INTELLIGENCE SYSTEMS THAT THEY CONSIDER ETHICAL PRINCIPLES
Governança de IA, Regulação de IA, IA responsável, Ética na IA, IA nas organizações públicas, Fuzzy QCA.
The race for regulation and AI global governance has advanced in the field of discussions which fostered many proposals for legislation and the defense of ethical principles’ definition by researchers, public and private managers, government officials, jurists, legislators and by society itself. While observing this challenging movement for all stakeholders, public organizations that produce AI systems are faced with the need to structure themselves to offer digital services based on AI systems that consider ethical principles. This research aimed to investigate whether public organizations have incorporated the guidelines presented by the academy, by legislation and by international standards, to their governance and management models in the production of AI systems that consider ethical principles. The investigation began with bibliographic research on AI regulation and AI governance, which supported the design of an AI regulation and governance framework that encompasses public and private institutions in a country, the AIR – Artificial Intelligence Regulation framework. Eleven propositions were elaborated on the processes and practices recommended by the academy, by the legislation under discussion and by international standards for the implementation of AI governance. A search was carried out in public organizations that had AI systems in operation. As a result, a sample was composed of twenty-eight public organizations, distributed in seventeen countries. With an exploratory and descriptive purpose, through a qualitative-quantitative analysis approach, the Quantitative Comparative Analysis (QCA) method was carried out, in crisp-set and fuzzy modes, both based on questionnaire responses. The QCA results joined to the interviews and public documents’ content analysis. The results supported a few findings in the studied sample: a) how organizations that were most advanced in the implementation of AI governance combined processes and practices distributed at different levels of the organization, aiming at the development of reliable AI systems; b) how they used drivers to help AI governance implementation; and c) how they faced the obstacles in such journey. A framework was built regarding how the joint participation between business units and the IT unit can be adjusted to respond to situations involving AI systems outsourcing, and to expand the benefits of government recommendations (soft laws). As theoretical contributions, the research highlights the need for collaboration between stakeholders in complex intra and interorganizational relationships supported by the Stakeholder Theory; as well as efforts to ensure accountability as an ethical principle which is part of AI governance, supported by the Agency Theory. In addition to the AIR framework, a second framework was elaborated on how processes and practices are distributed in a public organization, with a partnership between the business area and the IT area, in an AI governance model. As a contribution to managers and researchers, in addition to knowledge of the transformations identified in organizations, the research presented an integrated model of governance and management processes and practices, which extend from the strategic level to the operational level of a public organization for its AI governance implementation. By this way, it is expected that public organizations can add itself into its country's AI governance, and thus joins to the global AI governance movement.