Page 35 - AIH-2-3

Artificial Intelligence in Health Machine consciousness

moment-to-moment awareness we experience in daily Phenomenal consciousness refers to the subjective
life when making rapid decisions. In humans, despite a experience itself – the raw feel of sensations and thoughts,
flurry of unconscious sensory and cognitive processing, often described as “what it is like” to be in a given mental
consciousness provides a singular, integrated vantage state. It encompasses the qualitative, first-person aspects of
point (the feeling of “being aware”) that helps us adaptively mind (sometimes called qualia), such as the redness of red
navigate each moment. By mimicking this in machines – or the pang of emotion. By contrast, access consciousness
ensuring that an artificial agent has an integrating layer denotes a mental state’s availability for use by the cognitive
that monitors and directs sub-processes – their framework system. A piece of information is “access conscious” if it
aims to achieve a conscious-like functionality that is widely broadcast within the brain (or system) such that
could improve the system’s flexibility and robustness in various processes (reasoning, memory, decision-making,
unpredictable environments. Notably, these authors regard verbal report) can utilize it. In essence, access consciousness
such architecture not just as an add-on to intelligence, but concerns the functional role of conscious information –
as essential for complex adaptive behavior: A machine how it is accessible and how it guides behavior – rather
endowed with a consciousness-like integrative function than how it feels.
might better handle novel situations by flexibly combining This distinction has profound implications for artificial
information from all its subsystems. consciousness. Most neuroscience-inspired AI frameworks

The above frameworks illustrate how insights from implicitly aim at access consciousness – ensuring that
neuroscience and cognitive psychology (such as the an AI system possesses internal representations that are
importance of self-monitoring and global integration of globally available and can be used to organize behavior in
information) are being translated into AI design. Each an intelligent, context-sensitive way. For example, when
approach stresses a different facet of natural consciousness: Chatila et al. 79(p1) focus on robots “knowing what they have
From Levy’s broad pragmatism to Chatila’s self-reflective learned” and reporting that knowledge, they are dealing
knowledge, to Kinouchi’s global integration. The diversity with access consciousness: The learned information is
of these proposals also underscores that there is not accessible for future decisions and self-report. Similarly,
80
yet a consensus on a single “blueprint” for artificial Kinouchi and Mackin’s integrative layer is designed to
consciousness. Different researchers prioritize different collect distributed information and make it available to
cognitive ingredients (self-awareness, learning, integration, the whole system for coordinated adaptation – again, a
etc.), reflecting the multifaceted nature of consciousness functional, access-oriented property.
itself. This plurality suggests that the field is still in an Phenomenal consciousness, however, is a much harder
exploratory phase: Much like the blind men and the issue. It asks whether the robot or AI actually has an inner
elephant, each framework captures one aspect of the larger life: Is there something that it is like to be that robot? Does
concept. A key task for the research community moving it feel anything when it integrates information or reports
forward is to synthesize these insights and determine how on its knowledge? This is the crux of the hard problem
they might fit together. For instance, one could ask whether in the context of AI. Strong AI enthusiasts might argue
a truly conscious machine would need to incorporate all of that if we achieve a complete functional emulation of the
these elements – a shared functional understanding, self- brain’s processes (i.e., replicate access consciousness to a
awareness, and global integrative capacity – or whether high degree), then phenomenal experience might emerge
any one of them might be sufficient on its own. Addressing naturally. However, skeptics point out that no matter how
such questions requires not only engineering advances sophisticated a machine’s functional capabilities, this does
but also deeper theoretical clarity, which brings us to the not guarantee – or even necessarily imply – the presence
distinction between different notions of consciousness and of subjective experience. A machine could conceivably
83
how they apply in artificial systems. meet every external criterion for access consciousness – it
could introspect, reason about its own mental states, and
4.3. Access versus phenomenal consciousness: behave indistinguishably from a conscious being – yet still
Functional versus experiential dimensions lack any inner lights on. This skeptical view is epitomized
In discussions of both human and machine consciousness, by certain philosophical arguments (e.g., Searle’s Chinese
it is crucial to distinguish between two often-confused Room or the notion of philosophical zombies) and has
dimensions of conscious states: Phenomenal consciousness been voiced in contemporary analyses that conclude
and access consciousness. This distinction, originally robots are not – and perhaps cannot be – conscious in
82
84
articulated by Block, has proven useful in framing debates the phenomenal sense. Thus, the phenomenal versus
82
about consciousness in artificial systems. access distinction serves as a reminder that behavioral or
Volume 2 Issue 3 (2025) 29 doi: 10.36922/aih.5690