Picture a roomful of AI researchers all claiming to study 'consciousness' in machines. One means the ability to use the word 'I' correctly. Another means sensory integration. A third means quantum coherence in microtubules. They are not disagreeing about answers; they are not even asking the same question. That is the diagnostic premise of a mini-review published this summer in the journal Cognitive Systems Research, and once you see the problem the way its authors frame it, the whole literature on machine consciousness looks less like a scientific debate and more like a multilingual argument conducted without a translator.
The paper is authored by Werner Vogd and Jonathan Harth, and it appears in a special issue titled 'Machines in Thoughts: Philosophy, History, and Epistemology of Robotics and Artificial Intelligence,' edited by scholars at the University of Bologna and the University of Urbino Carlo Bo. The special issue is itself a signal of how seriously the academic philosophy of AI has started treating what was once considered a fringe question: whether the conceptual tools researchers grab off the shelf when they study machine minds are actually fit for purpose. Vogd and Harth's contribution is a systematic audit of that toolbox. They explore the possibility that machines can develop consciousness and map seven distinct theoretical approaches, including 'consciousness' as language ability, sensory perception, social interaction, and as an emergent phenomenon in complex systems. Read the list slowly: each item is a coherent research program, each has serious defenders, and each implies a completely different empirical test for whether any given AI system passes.
The seven approaches are where the paper's real intellectual payload sits. They run from 'consciousness' as the ability to participate in the language game of I-sayers, through sensory perception and acting accordingly, self-consciousness as a result of the structural coupling of social organisms, consciousness as re-entry, consciousness as the ability to lie, consciousness as a bodily phenomenon, and finally consciousness as a quantum phenomenon. That seventh item is not a typo. The Penrose-Hameroff Orchestrated Objective Reduction hypothesis, which roots subjective experience in quantum processes in neuronal microtubules, is a live enough position in the literature to earn its own chapter. The authors examine each perspective in terms of its feasibility in artificial systems, for example through neural networks, embodied cognition, or self-referential processes. Some translate into AI architectures with surprising ease; others resist any computational implementation almost by definition, which is itself an important finding.
The punch lands in the concluding section, whose title asks a pointed question: 'consciousness without quotation marks?' The authors' broader argument is that the word has been doing far too much work for far too long. They identify a central paradox: on the one hand, there is no reason not to implement consciousness processes in technical systems, but on the other hand, the gap between objective modelability and subjective introspective quality remains unbridgeable. That is the hard problem restated with engineering precision, and it is not going away. The field is essentially stuck choosing between a version of consciousness that can be operationalized but may not be what anyone actually cares about, and a version that clearly matters but may be permanently beyond the reach of any external test. Compounding the confusion, AI researchers, philosophers, and natural scientists each bring structurally different conceptual frameworks to the table, and the policy decisions and ethical frameworks surrounding AI consciousness are likely to be shaped primarily by whichever discipline dominates the room.
Skeptics will note, reasonably, that a mini-review mapping frameworks is not the same as resolving them. Eduardo Garrido-Merchan has argued in print that machine consciousness research risks becoming pseudoscience precisely because it borrows the prestige of 'consciousness' without agreeing on what the word means. As of 2026, the scientific consensus is that no current AI system has been confirmed conscious, and the field has shifted toward probabilistic frameworks that assess consciousness across multiple competing theories rather than a single yes-or-no test. That shift is real progress, but it also risks letting every theory stay alive indefinitely on partial evidence. What Vogd and Harth add to that landscape is not a verdict but something arguably more useful in the short term: a map of the conceptual terrain that makes clear exactly which framework any given claim is operating inside, and therefore which evidence could actually settle it. The linkage between the philosophical and methodological development of AI is crucial for understanding how theories about mind, perception, and cognition get operationalized in these fields, and this translation from theory to practice has both shaped technological advances and raised new philosophical questions about machine autonomy. Without that translation being made explicit, researchers are not just talking past each other; they are building systems premised on incompatible definitions of the very thing they hope to study.
Until the field agrees on which definition of 'consciousness' is even on trial, every result it produces is a verdict in a case no one has formally filed.