The paper examines the philosophical challenges involved in creating artificial consciousness within AI systems. It begins by outlining the major features of the mechanistic worldview that underpins our general conception of machines. It then proposes an account of consciousness grounded in features human observers intuitively attribute by inference to other conscious agents through empirical observation. These include features such as subjectivity (having experiences), autonomy, perspectival embodiment, self-reflection, and intersubjectivity. Building on these features and aiming to avoid the epistemological problematics of researching consciousness, the paper introduces a multidisciplinary strategy of artificially approximating consciousness, coined algorithmic phenomenology. Its guiding question is: How can we construct an algorithmic agent that elicits such a compelling sense of conscious presence that a human may spontaneously overlook its computational substrate and entertain the possibility that it is conscious? Following that, the paper surveys several promising research avenues aimed at designing artificial agents that would consistently evoke in human observers intuitions that lead them to infer features associated with consciousness: (1) autonomy and an independent, partially unknowable perspective; (2) reflectivity and introspection; and (3) intersubjectivity. Special attention is devoted to the last, which shifts the focus of research from the agent itself to the complex relational dynamics it may establish with human interlocutors—dynamics that may prove to play a role not only in attributing consciousness by inference but also in the emergence of consciousness.