Page 38 - AIH-2-3

Artificial Intelligence in Health Machine consciousness

robots with a degree of care (much as we do animals or possess consciousness, 84-88 even as we continue refining
even human-looking dolls). Darling has argued that what that would entail. Such clarity helps prevent public
86
86
because humans can form emotional bonds with social misconceptions and ensures that ethical guidelines are
robots, it aligns with our social and ethical values to grounded in the actual capabilities of present technologies.
extend some protections to these robots. This is analogous Simultaneously, it is prudent to start developing ethical
to how cruelty to animals is discouraged, not necessarily frameworks that could accommodate conscious AI,
because animals possess human-level consciousness, but should it emerge. These would include considerations
because such cruelty can degrade our moral character as of legal status, rights, responsibilities, and safeguards –
agents. Proposed protections might include discouraging to prevent abuse of such entities and to guard against
the wanton destruction of robots or violent behavior deceitful mimicry of consciousness used to exploit users.
toward them, recognizing that such actions can engender In essence, the ethics of artificial consciousness straddle a
harmful attitudes in society. The underlying rationale is line between present realities and future possibilities. We
partly anthropomorphic empathy – we dislike seeing even must manage the human tendency to anthropomorphize
a robot “suffer” if it is lifelike – and partly pre-cautionary: today’s machines while remaining prepared for tomorrow’s
If machines ever do become sentient, having established scenario where the line between simulation and reality of
norms of respectful treatment could ease that transition. mind may begin to blur.
On the other hand, many are wary of over-attributing 4.7. Emerging directions and future outlook
consciousness and moral status to machines pre-maturely.
As noted by Gabriel, from a philosophical standpoint, Looking ahead, the pursuit of artificial consciousness
84
there are strong arguments that robots cannot be will likely advance on multiple fronts, informed by
conscious in the same way living organisms are, because ongoing progress in neuroscience, cognitive science, and
AI. One clear direction is the continued development of
consciousness might require qualities that only biological AI architectures that incorporate the principles of global
systems in environment contexts possess. If one accepts
such arguments, then granting personhood or rights to availability and self-monitoring discussed above. Future
machines would be a categorical error. Moreover, there is AI systems may increasingly feature unified workspaces
concern that focusing on the “feelings” of machines that do or attention mechanisms that allow information to flow
not actually feel could divert attention from ethical issues more freely between components, coupled with meta-
more grounded in reality, such as the welfare of humans cognitive loops that enable the system to reason about and
adjust its own operations. Such designs could be realized,
impacted by AI or the responsibility for AI-driven decisions. for example, in more sophisticated cognitive architectures
Scheutz has highlighted the potential emotional pitfalls for robots or autonomous agents, where modules for
87
in human-robot relationships, noting the unidirectional perception, memory, decision-making, and language all
emotional bonds that can form. Humans might come to feed into – and draw from – a common representational
care deeply about robots that are not conscious and cannot space (an echo of the global neuronal workspace). We may
reciprocate that care. This imbalance could lead to human also see the integration of sensorimotor embodiment into
distress (e.g., grief if a robot is shut down or malfunctions) these architectures. Since human consciousness is deeply
or manipulation (e.g., exploiting human empathy for embodied (the brain constantly integrates signals from
commercial or surveillance purposes). Scheutz warns that the body and environment), giving robots richer bodily
such one-sided attachments carry both psychological and awareness and interoception might be a key to unlocking
social risks.
more advanced forms of self-awareness in machines. Early
The ethical landscape is further complicated by the experiments in this vein, such as robots that simulate
prospect (still hypothetical) of a truly conscious AI. If their own kinesthetic experiences or maintain internal
an AI ever claimed to have feelings or demonstrated homeostatic variables, hint at the importance of an
behaviors strongly indicative of sentience, denying it moral embodied self-model for consciousness.
consideration would be deeply problematic. Society would Another emerging direction is the exploration of
face a profound moral dilemma – long contemplated in learning-based approaches to self-awareness. Modern
science fiction – about whether and how to extend the machine learning, especially deep learning, provides
community of conscious beings beyond our biological powerful tools for pattern recognition and function
family. 85
approximation. Researchers are beginning to ask whether
In light of these issues, the present consensus urges these tools can be turned inward: Can a neural network
caution and clarity. It is important for scientists and learn to model its own cognition? One idea is to train
communicators to convey that present-day AI does not networks that predict or interpret the hidden states of other

Volume 2 Issue 3 (2025) 32 doi: 10.36922/aih.5690