By Mona-Marie Wandrey and Marta Halina, Department of History and Philosophy of Science, University of Cambridge
(See all the posts in this series here.)
We regard the proposal advanced by Will Bridewell and Alistair Isaac as a unique and promising methodological framework for advancing empirical progress in consciousness science. The explicit acknowledgement of metaphysical agnosticism is welcome, as conflicting theoretical commitments often stand in the way of reaching consensus. Although a variety of theoretically modest approaches for studying consciousness have been proposed in the literature (i.e., those compatible with many different theories of consciousness), they are also known to have shortcomings (Birch 2020, Shevlin 2021). We believe the apophatic method has the potential to avoid some of these shortcomings.
As an example of the strengths of this method, consider a recent critique of evolutionary approaches to identifying consciousness. Halina, Harrison, and Klein (2022) note how some evolutionary approaches avoid making theoretical commitments regarding the mechanisms responsible for consciousness. For example, Birch, Ginsburg, and Jablonka (2020) advance unlimited associative learning (UAL) as an epistemic marker of consciousness. UAL is a promising marker, they argue, because it requires a list of capacities that researchers generally agree are also sufficient for consciousness (e.g., global accessibility, binding, intentionality, and others). However, Halina and colleagues argue that such a strategy carries a risk. A barometer may indicate a storm reliably, but it fails to explain that storm. Similarly, UAL may correlate with consciousness while remaining mechanistically unrelated to it. What is needed are details concerning the processes underlying consciousness, which an epistemic-marker approach does not provide. The apophatic method stands in stark contrast to this by focusing explicitly on simulated or robotic instantiations of the mechanisms responsible for subjective experience, and on the iterative improvement of such models.
While we are excited to see what this new approach will bring, we are less optimistic than Bridewell and Isaac that computational modeling will be able to advance our understanding of consciousness in a deeper sense. While they emphasize throughout the paper that their commitment to apophatic consciousness science is purely methodological, they also envision that integrating more and more consciousness-relevant phenomena into computational models will, at some point, either “empirically vindicate methodological computationalism” by successfully exhausting all the possible phenomena or lead to a “crisis” in consciousness science by identifying a “circumscribed core” of consciousness-related properties that cannot be reduced to computational implementation (Bridewell & Isaac 2021, p. 8). In both cases, it seems that they regard our ability or inability to model consciousness-relevant phenomena as potentially revealing something about consciousness itself, by delineating “the boundaries of consciousness as a natural phenomenon from without” (ibid.).
We are not convinced by either of the above envisioned scenarios. Starting with the former, it is probably true that if we were, one day, able to integrate all the consciousness-relevant phenomena into a computational model that perfectly reproduces the behaviors associated with consciousness, proponents of functionalism might regard this as evidence that functional properties are, in fact, all there is to consciousness. However, opponents of functionalism are unlikely to be convinced, as the counter-arguments cited by the authors (the Chinese Room and the Zombie objection) still hold. Hence, it doesn’t seem that our ability to model consciousness-relevant phenomena computationally would be able to resolve the metaphysical agnosticism we started with (and while Bridewell and Isaac point out that this scenario would likely change our intuitions, it does not seem they think that it would vindicate functionalism itself). This entails that apophatic consciousness science would put us, ultimately, in a position where we have computational models that mimic consciousness convincingly without being able to determine whether the models are actually conscious or not.
Moving to the second scenario envisioned by the authors, that we will identify consciousness-relevant phenomena that we are not able to model computationally. Would this, in fact, lead to a “crisis” in consciousness science and advance our understanding of what consciousness is by providing negative evidence? We believe it would not, as there are many reasons why a computational model might fail to capture consciousness-relevant phenomena, which have nothing to do with consciousness per se. For example, we might fail to integrate quantum processes in the brain into our computational model merely because our computing capacities are too restricted. That does not imply that there is a conceptual connection between our modeling inability and consciousness itself, or that “negative evidence” points to an involvement of quantum processes in conscious experience. Negative results are notoriously difficult to interpret and consciousness research is no different from other sciences in this respect.
One final point: we would like to highlight that one of the features that makes consciousness science so difficult is that methodological and empirical considerations are deeply entangled with ethical concerns. Whether we regard another entity as conscious or not has major ethical implications, for example in the context of diagnosing brain-injured patients with disorders of consciousness, or when determining which species of non-human animals are conscious (Johnson 2021). Value judgments about the ethical risk of false positive and false negative results are embedded in methodological choices. Thus, we disagree that ethical concerns about conscious computational models are “completely orthogonal to methodological and empirical considerations” (Bridewell & Isaac 2021, p. 8). Instead, we think it would be best to acknowledge that the apophatic method comes with certain ethical risks and note that these need to be weighed against the risks of other approaches. Philosophy of science provides a rich resource for exploring how such risks are evaluated in other sciences (Douglas 2009).
In conclusion, we think that while the apophatic approach proposed by Bridewell and Isaac provides a promising framework shift that might advance the empirical science of consciousness by integrating and unifying the mechanisms proposed by various conflicting theories of consciousness, it seems that the metaphysical agnosticism they take as a starting point will still be there at the end. This creates ethical challenges that need to be acknowledged and managed by the empirical science of consciousness.
References
Birch, J. (2020). The Search for Invertebrate Consciousness. Noûs, 56(1), 133–153.
Birch, J., Ginsburg, S., & Jablonka, E. (2020). Unlimited associative learning and the origins of consciousness: a primer and some predictions. Biology & Philosophy, 35, 1–23.
Bridewell, W., & Isaac, A. M. C. (2021). Apophatic science: How computational modeling can explain consciousness. Neuroscience of Consciousness, 2021(1), 1–10. https://doi.org/10.1093/nc/niab010
Douglas, H. (2009). Science, Policy, and The Value-Free Ideal. University of Pittsburgh Press.
Halina, M., Harrison, D., & Klein, C. (2022). Evolutionary Transition Markers and the Origins of Consciousness. Journal of Consciousness Studies, 29(3-4), 62–77.
Johnson, L. S. M. (2021). The Ethics of Uncertainty: Entangled Ethical and Epistemic Risks in Disorders of Consciousness. Oxford University Press.
Shevlin, H. (2021). Non‐human consciousness and the specificity problem: A modest theoretical proposal. Mind & Language, 36(2), 297–314.
