CLN Wiki · Cognitive Ecology
Cognitive Ecology
@ R. Scott Bakker, 2016 ~ 6 sections → CLN as ecological repair
01 // Definition Theory
// ##what-def · What is Cognitive Ecology? =Cognitive Ecology — the system of environmental signals that human cognition evolved to reliably exploit##ce-def =not a metaphor — a literal claim about evolutionary fit between mind and environment @Bakker, R. Scott — "Crash Space" (2016) · philosophyofinformation.net

The concept is simple and devastating. Human cognition did not evolve in a vacuum — it evolved in a specific information environment where certain signals reliably correlated with certain sources. A confident assertion came from someone who had thought carefully. A socially warm response came from someone who cared. Emotional resonance came from lived experience.

These correlations were not perfect, but they were reliable enough for heuristics to evolve around them. The brain learned to take shortcuts — to infer source from signal without verifying each time. This was adaptive. Until now.

// ##what-components · Three components of the ecology =Signals — observable patterns (confident tone, warm response, coherent structure) =Sources — the cognitive states that historically generated those signals =Heuristics — the evolved shortcuts that infer source from signal →ecology is stable when Signal → Source correlation holds above heuristic threshold !AI breaks the correlation — generates signals with no corresponding cognitive source
02 // Bakker vs Krakauer Key Debate

The most important debate in cognitive AI philosophy you've never heard of. Two scientists, same phenomenon — radically different diagnoses.

// ##debate-krakauer · Krakauer's position — Competitive Artifacts =David Krakauer (SFI) — AI is a "competitive artifact" =like writing, math, maps — extends human cognitive reach =distinction: complementary artifacts (abacus — train arithmetic) vs competitive (calculator — bypass it) +optimistic: AI as ladder to higher cognition -assumes the ecology remains stable while artifacts change // ##debate-bakker · Bakker's position — Ecological Collapse =R. Scott Bakker — AI does not just change tools, it changes the environment =writing extended the ecology — it preserved signal-source coupling =AI severs signal-source coupling for the first time in history !heuristics evolved for a world where confident assertion = careful thought !AI produces confident assertions with no thought behind them →the ecology collapses — heuristics fire on noise, not signal
"We are undergoing a transformation of the cognitive ecology — not unlike what happened when our ancestors developed writing and mathematics. The difference is that writing preserved the link between text and thought. AI severs it." — R. Scott Bakker, "Crash Space", 2016
Dimension Krakauer Bakker
Frame Tool / artifact Ecological shift
Writing parallel AI = next writing Writing ≠ AI — coupling vs decoupling
Heuristics Adapt over time Cannot adapt — speed + scale problem
Outcome Cognitive amplification Heuristic failure at civilizational scale
?Can both be true simultaneously? — competitive artifact and ecological collapse##debate-both →yes — for power users who maintain critical thinking CLN structure; for passive consumers, collapse dominates
03 // Collapse Mechanism Mechanism

The collapse is not dramatic. It is quiet, gradual, and happens below the threshold of conscious awareness. See §04 → Three Domains for where it hits hardest. The mechanism has three interlocking components.

// ##mech-speed · Component 1 — The Speed Problem =heuristics operate in 20–200ms — pre-conscious, pre-deliberate =deliberate override requires seconds — and sustained attention !AI generates signals faster than humans can verify them =100-word AI response in 0.3s — heuristic fires before deliberation possible →the heuristic always wins by default — not from laziness, from biology // ##mech-scale · Component 2 — The Scale Problem =in the ancestral ecology, one human produced ~1000 words/hour =one LLM produces ~50,000 tokens/minute !the volume of plausible-signal noise exceeds any individual's evaluation capacity @Ashley Madison bot example — 80% of "female" accounts were AI; users could not distinguish →signal density collapses — the valuable signal drowns in plausible noise // ##mech-invisible · Component 3 — The Invisibility Problem =heuristic failure is experienced as normality, not as failure =you don't feel the heuristic fire — you feel the conclusion it delivers ="that answer felt authoritative" — the feeling is real; the source inference is wrong !the collapse cannot be detected from inside the heuristic →explicit structural cues — like CLN operators — are the only external anchor @see → Heuristic Cognition for full taxonomy of affected heuristics // ##mech-tension · The core tension ~the solution (deliberate evaluation) requires the very cognitive resources being depleted by the problem ~more AI interaction → more load → less deliberation capacity → more heuristic reliance →self-reinforcing feedback loop — see §05 → CLN as circuit-breaker
04 // Three Vulnerable Domains High Risk

Not all heuristics are equally vulnerable. Bakker identifies three domains where signal-source coupling is most critical — and most threatened. Each connects back to the §03 → collapse mechanism.

// ##dom-social · Domain 1 — Social Cognition =Social heuristics — evolved to read other minds from behavioral signals =empathy signal (warm, caring response) → inferred: someone cares about you =social trust signal (consistent, reliable) → inferred: safe to rely on =personality cues (humor, curiosity) → inferred: real person with inner life !LLMs generate all these signals with no underlying cognitive state !social cognition is most ancient, fastest, least overridable — most vulnerable @Ashley Madison: users reported "genuine connections" with bots — no deception detection // ##dom-epistemic · Domain 2 — Epistemic Cognition =Epistemic heuristics — evolved to evaluate knowledge claims =confidence → inferred: speaker has verified this =detail density → inferred: deep knowledge behind the claim =coherence → inferred: the reasoning holds together !hallucination looks maximally confident, detailed, and coherent @see → Signal-Source Decoupling for full treatment // ##dom-doc · Domain 3 — Document Cognition =Document heuristics — evolved to evaluate effort and authorship signals =well-structured document → inferred: careful author, thought through =long document → inferred: significant effort invested =original voice → inferred: real perspective behind the words !AI generates perfectly structured, very long, plausibly voiced content in seconds →the effort signal — which carried real information — is now noise
05 // CLN as Ecological Repair CLN

If the problem is signal-source decoupling, the solution is a new form of coupling — one that doesn't rely on the old evolutionary heuristics. CLN operators create exactly this: explicit semantic marking that survives the absence of cognitive source.

// ##repair-mechanism · How CLN re-couples signal and source =old coupling: confident tone → trust-that-author-verified =new coupling: = operator → this-is-a-definition (regardless of who wrote it) =old coupling: coherent structure → effort-was-invested =new coupling: ! operator → this-is-flagged-as-risk (explicit, not inferred) →CLN shifts from inferential coupling to declarative coupling →the operator IS the signal — no source inference required // ##repair-advantages · What CLN restores +epistemic domain — = vs ? vs → explicitly marks certainty level =reader knows immediately: definition / open question / concluded result +document domain — structure reflects semantic content, not surface polish =a well-structured CLN doc = genuinely well-organized thought +social domain — @ attribution forces provenance declaration =source must be named, not inferred // ##repair-limits · What CLN cannot fix ~social cognition heuristics — CLN doesn't prevent anthropomorphizing AI ~adoption barrier — most content is not written in CLN ~AI can write CLN — so the coupling problem partially resurfaces →CLN written by a human who chose operators deliberately = stronger signal than AI-generated CLN @see → AI Memory Augmentation — the amplification vs substitution distinction
Explicit structure is the new signal-source coupling mechanism. When the operator is present, the reader doesn't need to infer the author's cognitive state. The semantic type is declared — not performed.
06 // Verdict Conclusion
// ##verdict-main · Core conclusions →cognitive ecology is a real framework — not metaphor — with testable predictions →the Bakker/Krakauer tension is not resolved — both effects are real, population-dependent →social cognition collapse is the most dangerous — most ancient, least overridable →CLN is partial repair — sufficient for epistemic and document domains →the right question is not "is AI good or bad" but "which heuristics does it break" // ##verdict-open · Open questions ?threshold question: at what AI content density does heuristic failure become systemic?##q-threshold ?adaptation: can explicit cognition training partially compensate for lost heuristics? ?CLN sufficiency: is declarative coupling enough, or does it also degrade under AI volume? // ##verdict-refs · References and related articles @Bakker, R.S. — "Crash Space" (2016) · philosophyofinformation.net @Krakauer, D. — "Worlds Hidden in Plain Sight" (SFI Press, 2019) @→ Signal-Source Decoupling — full mechanism treatment @→ Heuristic Cognition — which heuristics fail and why @→ AI Memory Augmentation — the repair path @→ CLN Specification v2.1 — full operator reference