What Can Analogy Tell Us About Artificial Consciousness?
Holyoak and Monti tackle the question of AI consciousness by developing a causal framework for evidential analogy. Their central idea is that behavioral similarity to humans carries far less weight than shared causal mechanisms plausibly generating subjective experience. The paper argues that animals inherit strong analogical support from evolutionary and neurobiological homology, whereas contemporary AI—despite impressive performance—lacks established causal correspondences with human consciousness. This makes the work a careful attempt to redirect a debate often mired in hype toward principled, mechanism-first reasoning.
The paper offers a timely, conceptually clear argument that the analogy from human to animal consciousness is far stronger than the analogy from human to AI. The soft INUS formalism is presented with appropriate epistemic modesty, and the critique of purely behavioral criteria is well-supported by neuroscientific evidence. However, because the causal basis of consciousness remains deeply contested, the framework’s ability to decisively adjudicate competing claims is limited; it organizes uncertainty rather than resolving it.
The distinction between pedagogical and evidential analogy is sharply drawn and helps expose why superficial behavioral parallels do not constitute evidence for inner experience. The review of biological consciousness—emphasizing that locked-in patients remain conscious without motor output and that complex cognition can occur unconsciousness—effectively severs the link between intelligence and phenomenology. The animal calibration case is handled with nuance, acknowledging that homology weakens with phylogenetic distance but can be partially offset by convergent functional organization.
Although the formalism in Box 1 is elegant, its variables—$w_c$, $s_c(T)$, and $P(U_T)$—are impossible to calibrate with current knowledge, making the equations illustrative rather than operational. The authors acknowledge this, stating that "The purpose of the formalism is not to assign precise numerical probabilities." Nevertheless, the framework tacitly assigns high causal weight to biological substrate features that AI lacks, a move that functionalists may reject as question-begging unless one already accepts that substrate matters. Additionally, while $P(X_T)$ is left open, the paper offers almost no guidance on what would count as positive evidence for a genuinely non-biological route, leaving the alternative-mechanism clause as a largely empty placeholder.
The comparisons between humans, animals, and AI are generally fair and empirically grounded. The authors correctly note that computer implementations of global workspace theory captured neural and behavioral findings but "explicitly did not conclude that their computer simulations of access consciousness had acquired inner experience." Their engagement with recent LLM interpretability is similarly measured: they grant that identifying verbalizable vectors is interesting but insist that "identifying vectors in an LLM that approximate words does not constitute evidence that the LLM has any sort of inner experience." These rebuttals are consistent with their causal framework and avoid the strawmanning common in debates about AI sentience.
Because this is a theoretical paper, reproducibility rests entirely on conceptual clarity and the completeness of the stated framework. Box 1 supplies explicit equations and Box 2 enumerates a concrete research agenda, which aids disciplined application by others. Still, the authors admit that "The framework is intended primarily as a way of organizing evidence rather than assigning precise numerical probabilities," and without operational rules for determining causal weights or similarity scores, independent researchers are likely to apply the formalism inconsistently. No code, data, or empirical protocols are provided or expected.
Who or what is conscious? Because subjective experience is directly accessible only in the first person, judgments about consciousness in other entities depend partly on analogy. Historically, such inferences have focused on nonhuman animals, but advances in artificial intelligence have raised the possibility of conscious AI. Here we develop a causal framework for evaluating such evidential analogies. The key distinction is between similarities in factors plausibly involved in generating consciousness and similarities in downstream behavioural or cognitive effects. Our framework weights source-target similarity by causal relevance while allowing for unknown causes, disabling differences and alternative routes to consciousness. Applied to biological systems, it explains why analogical support generally weakens with increasing causal distance from humans. Applied to contemporary AI, it suggests that behavioural similarity provides only limited evidence for consciousness because relevant causal correspondences remain poorly established. The framework also clarifies what evidence would strengthen claims of artificial consciousness.
Pick a starting point or write your own. Challenges run in the background, so you can keep reading while the AI investigates.
No challenges yet. Disagree with the review? Ask the AI to revisit a specific claim.