Lauren Ross Mechanistic Explanation and What the Consensus Ambition Needs From Philosophy of Science at MoC7
Lauren Ross is Professor of Logic and Philosophy of Science at UC Irvine. She will keynote Models of Consciousness 7 (MoC7) at the HC Ørsted Institute, University of Copenhagen, October 12-16, 2026. Her research examines causation, explanation, and scientific practice in the life sciences, with a particular focus on whether proposed explanations in biology and neuroscience are actually doing the explanatory work their proponents claim.
Her presence at MoC7 addresses a specific problem the consensus paper will face: a methodological consensus is only as good as the theories it is built around, and those theories need to have set themselves tractable explanatory targets to have made progress that measurement standards can consolidate.
What philosophy of science contributes here
Ross’s research programme argues that neuroscience uses a narrower range of causal concepts than the phenomena it studies actually require. Mechanistic explanations, which identify component mechanisms and their interactions, are the dominant mode of explanation in neuroscience. Ross argues that biological systems also require pathway explanations, cascade explanations, and process-level explanations that cannot always be reduced to mechanism specification without loss.
This critique of mechanistic reductionism has a direct application to consciousness science. IIT explains consciousness in terms of the integrated information of a system’s causal structure. GNW explains it in terms of the broadcast dynamics of a workspace architecture. Both are mechanistic in the sense that they identify a property of a specific structural or functional mechanism and claim that property is what consciousness consists in or correlates with. Ross’s framework asks whether that explanatory strategy is sufficient for the phenomenon.
The philosophical diagnostic tool she brings is the concept of explanatory targets. A theory sets a tractable explanatory target if what it is trying to explain is specific enough to be measured, if the available methods are capable of measuring it, and if a positive result would constitute evidence for the claim the theory is making rather than merely evidence for a weaker associated claim. A theory with a vague explanatory target, or a target that available methods cannot measure, can always generate publications but cannot make genuine progress.
Why this matters for the Cogitate result
Liad Mudrik’s CS26 presentation on adversarial collaboration and consciousness theory testing, reviewed on this site, describes the methodological context of the Cogitate Consortium’s adversarial test of IIT and GNW. That test produced mixed results: frontal activity was found unnecessary for conscious perception (against GNW), and early-onset activity did not predict consciousness as IIT predicted. Neither theory was decisively confirmed or disconfirmed.
Ross’s philosophy of science framework provides a diagnosis of why adversarial tests between major consciousness theories tend to produce mixed results. The explanatory targets of IIT and GNW are each specified at a level of abstraction that allows partial match with many possible experimental outcomes. IIT says consciousness consists in integrated information above a threshold. This statement is formally precise but explanatorily vague: a system could satisfy the formal condition through many different neural mechanisms, and a negative experimental result could always be attributed to measuring the wrong kind of integration rather than to the theory being wrong.
A theory with a tractable explanatory target would instead specify not only what property consciousness correlates with, but what specific mechanism produces that property, what the precise measurement conditions are, and what experimental result would constitute genuine falsification rather than methodological revision. Mudrik’s adversarial collaboration protocol is a genuine attempt to force falsification. Ross’s analysis explains why the attempt produced ambiguous results: the theories’ explanatory targets were not specific enough to make clean falsification possible.
The relationship with Megan Peters
Ross and Megan Peters collaborate at UC Irvine on exactly the intersection of philosophical analysis and empirical measurement that MoC7’s consensus paper will need to navigate. Peters’s MoC7 profile on this site documents her finding that standard consciousness measurement frameworks were designed for and validated against biological subjects, and that applying them to non-biological systems, including AI systems, requires redesigning the tests before those tests can produce valid results.
The two researchers address adjacent problems from different directions. Peters asks whether the tests are valid for the population shift from biological to artificial subjects. Ross asks whether the theories the tests are built around have set tractable explanatory targets. Together, their contributions identify a two-level problem that the consensus paper needs to address: the theories may have vague targets (Ross), and even if the targets were well-specified, the tests may not be valid across the population shift to AI (Peters).
A consensus paper that specifies measurement standards without addressing both levels will be methodologically fragile. Standards that are technically precise but built around theories with vague explanatory targets will not produce results that the field can agree on, because any result can be interpreted as partial confirmation or methodological artifact. Standards that address validity but assume the original theories’ targets are well-defined will not catch the underlying problem that Ross identifies.
What the consensus paper needs from philosophy of science
The MoC7 speaker programme overview identifies the conference’s goal as producing a consensus paper with methodological standards that researchers from five distinct theoretical traditions can accept. Ross’s contribution to that process is a systematic evaluation of whether the proposed standards are methodologically sound rather than merely technically elaborate.
The specific questions her framework poses are: Does each proposed consciousness measure have a specific explanatory target? Can available methods measure that target rather than a proxy for it? Does a positive measurement result constitute genuine evidence for the theory’s central claim, or could the same result arise from a system that satisfies the measurement criterion without satisfying the claim? If the answer to any of these questions is negative for a given measure, that measure should not be included in a consensus paper that aims to guide AI consciousness evaluation.
These are demanding standards. They may not be satisfiable by all the measures currently under consideration. What they produce, if applied systematically, is a smaller set of measures with better-defined evidential standing, rather than a larger set of measures whose relationship to consciousness remains unclear.
What this means for AI consciousness evaluation
For AI consciousness research specifically, Ross’s diagnostic framework has a practical implication. The current approach to AI consciousness evaluation applies indicators derived from biological consciousness theories and counts the number that an AI system satisfies. If the underlying theories have vague explanatory targets, then satisfying more indicators does not straightforwardly mean having more evidence for consciousness: it may only mean satisfying more formal criteria whose connection to consciousness is unclear.
Ross’s framework does not establish that current AI consciousness evaluation is methodologically useless. It establishes that the field needs to do the philosophical work of checking whether the proposed indicators have tractable explanatory targets before the counts they produce can support strong conclusions about AI consciousness.
The 2026 consciousness research landscape documents a field with many active research programmes and significant theoretical disagreement. Ross’s philosophy of science contribution at MoC7 is less visible than a new empirical result, but it may be more consequential for the long-term quality of the consensus paper: measurement standards built around well-specified explanatory targets are standards the field can build on. Registration for MoC7 closes August 31, 2026.