Consciousness

What warrants attributing experience across unfamiliar organisms or substrates? Covers proposed indicators, competing theories, and the limits of transferring evidence between humans, animals and machines.

Theories and discriminating tests

  • Comparing consciousness theories - Compares higher-order, workspace, recurrent/predictive and integrated-information approaches. Differences concern explanatory targets as well as proposed mechanisms, complicating empirical comparisons between theories. Seth and Bayne 2022.
  • Adversarial consciousness tests - Preregistered visual tests in 256 people challenge sustained posterior synchrony and offset ignition predictions, while supporting others from integrated-information and workspace theories. The tests concern specified neural implementations. Cogitate Consortium et al. 2025.

Indicators beyond familiar brains

  • Cortical organoid activity - Developing cortical organoids produce nested oscillations dependent on excitatory and inhibitory signaling. Selected electrical features resemble preterm EEG; the study measures network maturation, not subjective experience. Trujillo et al. 2019.
  • Animal consciousness profiles - Organizes evidence along perceptual richness, evaluative richness, integration within and across time, and self-consciousness. Species can differ along several dimensions; the proposed profiles are hypotheses requiring comparative tests. Birch et al. 2020.
  • AI consciousness indicators - Derives candidate indicators from neuroscientific theories and examines AI architectures against them. These are theory-dependent assessment criteria; satisfying them is not a validated test of subjective experience. Butlin et al. 2023.
  • Organoid sentience criteria - Birch proposes brainstem-like regulation of needs, arousal and sleep-wake cycles as a warning threshold for organoid oversight. This precautionary criterion is theory-dependent, not an experimentally validated test of experience. Birch 2024.

Open questions

  • Which observations discriminate competing explanations rather than satisfy several theories at once?
  • Which proposed indicators remain justified when the biological or computational implementation changes?