A 2023 paper in BioSystems by Masumori and Ikegami proposes that spiking neural networks can produce informational closure by avoiding stimuli, suggesting a possible computational mechanism through which a network maintains a bounded, self-referential informational state rather than being driven entirely by external input.
What this finding does and does not show must be stated carefully: the paper's abstract is not available in the record used here, meaning every substantive claim beyond the title is inferred. The methodology, results, sample sizes, network architectures, and quantitative measures of informational closure are all unknown from the information provided. The finding cannot be independently evaluated on those terms.
The study appears, from its title, to involve computational modeling of spiking neural networks, which are models that mimic the discrete, timed firing of biological neurons. Whether the work includes empirical neural data, theoretical proofs, or simulation alone is not determinable. The peer-reviewed venue, BioSystems, adds baseline credibility, but it does not substitute for access to the methods and results.
The relevance to consciousness research lies in the concept of informational closure, a property some theorists associate with minimal autonomy and rudimentary awareness. If stimulus avoidance can produce such closure in a simple spiking network, that would be relevant to questions about the minimal conditions for basal cognition and autonomous agency. Those questions remain open, and this paper's actual contribution to them cannot be assessed without the full text.
Source: https://doi.org/10.1016/j.biosystems.2023.104972