Abstract

Rapid advancements in artificial intelligence (AI) have heightened the need for ethical design principles, bringing Responsible AI (RAI) to the forefront across academia, politics, and industry. Despite the plethora of guidelines, RAI faces challenges due to fragmentation and the lack of a cohesive explanatory theory guiding research and practice. Existing AI literature frequently fixates on RAI attributes within usage contexts, laboring under the misapprehension that RAI can be achieved solely through specific system attributes, responsible algorithms, or minimization of harm. This narrow focus neglects the mechanisms that interlace design decisions with the realization of RAI, thereby undervaluing their profound significance. Similarly, Information Systems literature predominantly emphasizes the operation and usage of these systems, often bypassing the opportunity to weave ethical principles into AI design from its inception. In response, this study adopts a grounded theory approach to theorize RAI design from the perspective of AI designers. The Authenticity, Control, Transparency (ACT) Theory of Responsible AI Design emerged as a result. This theory posits that authenticity, control, and transparency are pivotal mechanisms in RAI design. These mechanisms ensure that ethical design decisions across three domains—architecture, algorithms, and affordances—translate into responsible AI. The ACT theory offers a parsimonious yet practical foundation for guiding research and practice, harmonizing ethical considerations with AI technological advancements

DOI

10.17705/1jais.00948

Download

Share

COinS