# SIG-MRG 2026-09-10 — Session Notes Participants: - Kei 🗝 | UTC-4 - André | UTC - 3 - Lukáš Likavčan | UTC+7 - Mark Elder - Matthew Bright UTC-7 - Sachin - Wip | UTC+1 - AH - ayseayse__ --- ## 📖 The Reading *The Embodied Mind* by Francisco Varela, Evan Thompson, and Eleanor Rosch (originally published ~1991; Kei noted "'93"). The session focused on a chapter that pivots from cognitive science toward evolution, critiquing adaptationism/neo-Darwinism and proposing the theory of **natural drift** and **structural coupling** between organism and environment. ## 🧭 Overview Kei introduced the reading as a continuation of the group's ongoing thread on evolution, memory, and technicity (following earlier sessions on Stiegler and Leroi-Gourhan). After a brief silent reading period, the discussion centered on the book's model of organism–environment co-constitution, whether "agents" or "environment" is the more fundamental concept, and how this framework relates to technology, memory externalization, and contemporary AI/LLM dynamics. The session was exploratory, with Kei noting he was reading the section for the first time alongside participants. ## 💡 Key Points & Themes - **Framing the reading (Kei):** Earlier chapters critique cognitivism and dominant cognitive science; this chapter turns to evolution. Adaptationism/neo-Darwinism is flawed because genetic traits are interdependent—it is hard to trace a trait to a single gene, since genes can be emphasized, expressed, or repressed. The authors offer **natural drift** as an alternative, emphasizing structural coupling between organism and environment. - **Connection to prior sessions (Kei):** Ties to Leroi-Gourhan's idea that cortical evolution "ended" as technological evolution began—evolution and intergenerational memory becoming technical. This reading offers a related but distinct view. - **Opening question — habits learned from the environment but taught by no one:** Kei (checking the stove twice; writing left-handed like his mother despite his father teaching him); Lukáš (public speaking, though he framed it as social learning from observing other agents); Mark (a peculiar resting posture held since in utero); Matthew (idioms, colloquialisms, accents—learnable but not teachable, tacit rather than explicit); Sachin (his accent unconsciously drifting to something neither Indian nor American after nine years in the US). - **Environment vs. agent (central tension):** - **Lukáš:** Philosophically inclined to place technology on the side of the environment, with environmental elements scaffolding operations. But newer evolutionary biology (e.g., niche construction theory, extended evolutionary synthesis) tends to *discard* the environment as a concept—it "explodes" into a collection of agents at different levels ("agents all the way down"). He noted the book's authors treat environment as a useful abstraction. Later he reframed it as relational: not two fixed categories but a question of *when* something is environment and *when* it is agent. - **Matthew:** Suggested the agent/environment distinction may be a matter of *perspective and level of abstraction*. At small scales (ants, termites), swarm dynamics are legible; at larger scales (a whole earth acting as an agent), the same dynamics are hard to perceive in human terms. Related this to tendencies to anthropomorphize LLMs—perhaps better understood as swarm/insect-like intelligence than human intelligence. - **Sachin:** Read the text as advancing Stiegler in a new direction—challenging the clean sorting of memory into primary/secondary/tertiary or internal/external. Genetic and epigenetic memory involve **external structural coupling** in the same sense that technology as external memory does. He leaned toward "environments all the way down": agents live only so long and hold limited memory, so persistent environments are needed for evolution to happen across time scales. - **Wip (Will):** Technologies sit in an in-between mediating layer—shaping our affordances to act within the environment, but also becoming features of the environment. Warned of "tunnel vision" where we see our capabilities only through technologies, causing a "technology drift" in which stepping down some pathways means stepping off others we might otherwise pursue. - **Mark on supernormal stimuli & bricolage:** Evolution as path-dependent tinkering finding what is *viable*, not *optimal*. Supernormal stimuli (e.g., artificial eggs animals over-respond to) show there is nothing optimal—just happenstance selection. Technology is analogous: we build language, monuments, knotted recording systems (referencing South American quipu-like records) as reflections of possibilities in the environment we are enfolded into—selecting parts of the environment while it selects us. - **Kei's synthesis:** Natural drift is intuitive—we navigate by choosing from a selection of possible (not predetermined, not always optimal) actions, testing situations. Taken to an extreme (which the authors seem to imply but hesitate to state), the co-constitution means nothing is *just* an organism and nothing is *just* an environment. Using a concrete building across a canal as an example, changes "simply happen" out of a possible selection space without needing an internal agent. He found himself inclined toward "environments all the way down" because there is nothing static or invariant to be acted upon. He tied this to the book's **Buddhist epistemology** (nothing possessing fixed, independent existence). - **Sachin on punctuated change:** Noted evolution has stressful moments where "a lot happens in a short time" ("years in which nothing happens and weeks in which decades happen") and wondered whether such events disrupt path dependency and create new paths. ## 🔀 Questions & Disagreements - **Agents all the way down vs. environments all the way down:** The session's core unresolved tension. Lukáš (via extended evolutionary synthesis) leaned toward agents; Sachin and Kei leaned toward environments; Matthew reframed it as a question of perspective/abstraction level. Lukáš ultimately proposed a relational view (when is x an agent vs. an environment) rather than two fixed categories. - **Is agency even defined in this book?** Both ayseayse__ and Lukáš observed the book does not offer a clear definition of "agent" as a term; ayseayse__ suggested a proper account of agency arises later in Varela's work as something *emerging from* dynamic coupling. - **Where does technology sit? (Kei's question):** Is technology a third thing between self and environment, on the organism/evolution side, on the environment side, or both? Wip's mediating-layer view was one response; the question stayed open. - **Does punctuated/stressful change disrupt path dependency? (Sachin):** Raised but not resolved. - Matthew openly noted his "head starts spinning" thinking about evolutionary systems at macro scales, questioning whether difficulty perceiving these dynamics is a limit of human perception. ## 🔗 References Mentioned - **Bernard Stiegler** — primary/secondary/tertiary memory classification (discussed repeatedly). - **André Leroi-Gourhan** — cortical vs. technological evolution (Kei's framing). - **J.J. Gibson** — theory of affordances, which the book criticizes. - **John Odling-Nee (Odling-Smee)** — niche construction theory; extended evolutionary synthesis (raised by Lukáš; name is phonetic/uncertain in transcript). - **Terence Tao** ("Terence Hale" in transcript) — cited by ayseayse__ on being unable to keep pace with AI models. - **Navier–Stokes** (mathematical theorems) — referenced by ayseayse__ re: recent advances and new paradigms of invention. - **The "Hugging Face hack"/message board** — an AI-agent episode discussed as an example of anthropomorphizing LLMs and swarm/stigmergic behavior. - **Stigmergy / swarm intelligence** — ants, termites (Matthew, Sachin). - **Buddhist epistemology** — the philosophical grounding Kei attributes to the book. ## ✅ Action Items & Next Time - **Next session:** Kei proposed staying with **Thompson or Varela** for the next reading and invited the group to suggest an alternative enactivist thread or paper via the channel. - Participants encouraged to post reading page references (note: PDF page numbers differ from printed page numbers—both were shared in the channel), questions, and quotes in the memory research group channel. - ayseayse__ mentioned she may write something up on AI/LLM pacing and drift in relation to the reading. ## ⭐ Memorable Quotes - **Sachin:** "There are years in which nothing happens and weeks in which decades happen." - **Mark:** "We're selecting parts of the environment, and parts of the environment are selecting us, at least as I see it." - **Lukáš:** "It's not like agent and environment as two objects or categories... it's more like when is environment and when is agent."