Causal Control
# Continue PAPER_TPL
OTHER
Clarifies the distinction between causal interventions (difference-making/control) and mere correlates in neuroscience, using neural correlates of consciousness as a key example.
"One example of miscommunication is when mechanism language is used to refer to neural correlates of interest, such as the neural correlates of consciousness, and it is not clear whether mechanism is intended in a causal sense or in a non-causal, ‘mere’ correlation sense (which is suggested by the term ‘correlate’)112–115. This usage has engendered confusion about whether a mere correlate or cause of consciousness is being referred to, which is supported by slippage between a causal and a non-causal use of mechanism116,117."
Miscommunication, p. 87
The authors highlight that claims about mechanisms of consciousness require causal, interventionist evidence rather than correlational signatures, directly motivating causal-control tests in both brain and AI systems .
"Causal relationships provide explanations and understanding of ‘how’ a system — such as the brain, a neural ensemble or a cellular circuit — works, functions and behaves... causal models identify factors that control, explain and ‘make a difference’ to them1,27."
Why do causation and mechanism matter for neuroscience?, p. 82
By grounding explanation in factors that control outcomes, the paper frames mechanistic understanding as difference-making, reinforcing intervention-based assays (e.g., stimulation, lesions) as the gold standard for consciousness-relevant causal control .
Figures
Fig. 1 (p. 86)
: Figure 1 contrasts narrow, reductive mechanisms with broader causal systems, clarifying evidentiary standards for claiming mechanisms (vs correlates) in consciousness studies and AI evaluations .
Tables
Table 2 (p. 85)
: Table 2 | Examples of grant and funding information that mentions the ‘mechanism’ concept and ‘mechanistic research’
Limitations: This review is conceptual and does not present new empirical interventions or quantitative tests; it provides guidance on causal standards but no direct measurements or model ablations linking specific consciousness signatures to control.