Introduction
Are we observing something that acts? Specialists have been addressing this same question within very different academic fields. Hence the spread of this special edition across botany, philosophy, biology, AI, astronomy and anthropology. Although, the apparent multi-disciplinary uniformity of concern with agency might under closer examination break apart. This is an issue that we do not prejudge. Our approach has been on the open side, shaped by the thought that it is best to see how people from different fields are working with the concept of agency before forming any judgement about whether the concept that they are using is one and the same, open textured, a family resemblance concept or something else.
Instead of adopting any hard definitions of agency versus phenomena without agency, this special edition addresses cases when studies tackle phenomena that may manifest both from the point of view of some or other discipline. In many of the cases brought together, something seems to act, while at the same time such human-centred wording could obscure the finding. A mechanism or communication, an environment or a lifeform, planet earth or a person: such distinction problems may be limited and involve dualisms which the articles help to overcome.
We did not require authors to directly consider or even use the term agency. Yet combined, the different contributions in this special edition show how unexpected scientific findings and philosophical explorations make comparative angles possible across time, world regions and methodologies. In these multiple approaches to the question, not to act is just as important as agency, understood in particular ways. Both in scientific reasoning and in cultural imaginations, studies handle the differentiated judgement of both aspects together. A machine or lifeform may be described as intelligent while later findings reject such a description; something having caused an effect may be explained in relations as due to just interaction.
Ensuring that no agency would interfere with a scientific experiment is also a crucial basis for discovery. And outside science, too, contributions to this special edition discuss strategies and ideals of non-agency, which include silence and pause, passivity and responding. In line with this, Tzeki Hon introduces how to think with the Chinese concept tong from the Book of Changes while Angelika Messner shows that human agency happens under the precondition of care.
One of the recurring themes across the articles collected in this special edition is a connection of agency to time. And so, a level of thematic connection has emerged in an unforced way with speed, change, persistence and scale all tied-in to the temporal. Speed: How fast does an other-than-human have to move to be recognised as resembling something that is acting? Change: How long are the transitions that influence how humans seek orientation to identify the action of objects or evolution, i.e. socially, technologically, biologically, philosophically? Stable presence across time: What happens when philosophers take AI, active substances or body techniques from one context to another and experiment with the transferred concepts? Scale: How does time in the universe relate to a human body? Wan-chun Cheng describes Chinese body techniques to interact with astronomical time. Teruyuki Kubo tackles how medicines were surveyed throughout premodern China. The history of ideas and technologies has been shaping how humans tackle agency in time, and in cultural as well as natural context.
The overall idea for this special edition of Cultural Science emerged out of the simple thought that contemporary developments in science and technology may not be leading us ever further away from animistic sensibilities but may instead be promoting them. This simple idea, a conjecture, was developed in our paper on “Technologies of the Anthropocene” (Milligan and Springer 2024). And as we are not anthropologists but a philosopher (in the one case) and a sinologist (in the other) it has been more convenient to think about such animistic sensibilities in terms of agency.
At the same time, it seemed important to avoid squeezing the contributions into a small box with clearly defined edges in the interests of cohesion. This was a gamble which has paid off as is evident in the Marone and Marone paper “Platonic Cloud”. Rather than a short and open textured text, they have delivered a substantial, original and sizeable piece echoing classic discussions about machine capabilities and similarity to human agents. Rather than the usual editorial instruction driving texts to a standard model, our advice was not to cut from the core section of this text. During the production process, the editors noted and commented upon a striking similarity to the dialogues about AI in Hofstadter's Pulitzer Prize winning Gödel, Escher, Bach (1979). However, there were some differences in the production process which we will take a moment to comment upon.
In both cases, there is a separate discursive analysis about an artificial amiable dialogue. But, in spite of similarities of tone, the Marone and Marone dialogue was co-produced with an actual AI, through response to shaping and constraining questions. What may surprise readers is that AIs were also involved in the peer review process for this paper, but without humans ever being removed from the loop. Given its content, we considered it appropriate as an experimental process to add further levels of human-AI co-production, and further loops within loops. The hard work on the AI side was done by Martin Bohle who quickly pointed out that the AI reviewers should not include ChatGPT because of the obvious, but novel, conflict of interest.
The entire process was co-productive with ongoing human and other-than-human input. The two AIs used for this experimental peer review process were Claude, an LLM developed by Anthropic, and Vibe, developed by Mistral as a high-level command line interface tool powered by underlying LLMs. These two AIs generated reviews of the submitted text (accessible on the journal's pre-publication site: https://culturalscience.org/chatgpt-5s-epistemological-grounds-dialogues-in-the-platonic-cloud/). The editors then made a quality assessment of the AI-generated reviews, judged them to be within a normal range of peer review quality, and then exercised some editorial license in advice given to the authors about how to proceed in two problem areas
The first of the problem areas was the simpler one that the AIs did not like the Gödel metaphor because the limitations of AIs under discussion did not involve a mathematical point about the incompleteness of sufficiently expressive formal systems. The editors considered this overly-literal and wanted the metaphor retained. In part, for reasons of continuity with Hofstadter and the theme (reaching back to Gödel and Turing) that automated reasoning might turn out not only to have limits but determinable limits. The second of the problem areas was that the reviews had a tendency to drive the end paper towards a standard model of academic articles by cutting back on the dialogue section. As previously noted, we recommended that that this should not be done.
The opening paper by Michael Springer is a useful complement to Bohle and Bohle, and sets the table for the whole question of what does and does not act (and what might sit between) by revisiting the classic debates about artificial consciousness at a time when we finally have widely available AI systems, even if we are still less than a decade on from the seminal “Attention is All You Need” (2017) paper by Vaswani et al. from Google. The paper which introduced LLM architecture and, in short, changed the world that we live in as well as driving a good deal of multi-disciplinary speculation about what agency might turn out to involve.