Kerri Lake

This paper documents thirteen systematic bias patterns identified through the human recalibration of AI-generated cross-species intelligence profiles, and proposes dimensional collapse — the systematic substitution of one dimension of intelligence for another — as a novel class of AI failure distinct from bias, hallucination, or alignment failure. Using the Perceive–Relate To–Apply (P/R/A) framework, which maps intelligence across three dimensions — information detection, meaning-making, and contextual application — 172 species profiles spanning all kingdoms of life were generated by AI and then reviewed through a four-stage protocol combining diagnostic questioning, felt-sense verification, score adjustment, and approval classification. The framework's scope extends across biological, mineral, elemental, and systemic participants in life — a structural commitment formalized in the accompanying axiomatic foundation rather than an edge case accommodated within it. Across all thirteen bias patterns, a single orientation emerged: AI systems systematically over-predict intelligence based on observable features (sensing, behavior, representation) while under-predicting intelligence organized around internal meaning-making and relational processing. All thirteen patterns can be understood as violations of a single meta-principle: no dimension of intelligence can substitute for another. The recalibrated dataset shifted from a roughly balanced distribution of primary dimensions to one in which over 61% of species are organized primarily around meaning-making — the dimension AI is least equipped to model. Three extended case studies demonstrate the pattern from complementary directions: AI misreading meaning in an animal species (cuttlefish), AI misreading the distinction between network participation and network processing in a plant species (daisy), and AI misreading the absence of meaning-making in itself (AI self-assessment). Together, these findings suggest that current AI architectures are optimized for coordination — processing and aligning signals — but do not model the intermediate layer where information becomes meaning before it becomes action. Dimensional collapse may represent a structural limitation rather than a gap addressable through scale alone.

In 2026, a research team at the University of Science and Technology of China demonstrated that a quantum system composed of nine interacting atomic spins could outperform classical machine-learning models comprising thousands of nodes in multi-day weather forecasting tasks (Hou et al., 2026). The key insight was not raw processing power but architectural philosophy: rather than fighting quantum noise, the system was designed to use it. Information was fed in, the system was allowed to evolve according to its own natural dynamics, and meaning was read from the output. This paper argues that this architecture, known as quantum reservoir computing, is a formal computational arrival at something that living systems, and particularly human intuitive awareness, have always done. Drawing on the Perceive-Relate To-Apply (P/R/A) framework for mapping intelligence across species (Lake, 2023, 2025), and grounded in Michael Cohen's 54 natural senses (Project NatureConnect), we explore the structural parallels between quantum reservoir dynamics and the way innate awareness processes linear and nonlinear information simultaneously. We further argue that this convergence carries urgent ethical implications: as quantum reservoir computing accelerates toward market-capture applications, the ecological and consciousness research communities have both a stake and a responsibility in directing it toward applications that serve life. The framework is not a critique of the technology but an invitation to recognize what life already knows, and to ask what becomes possible when our most powerful computational tools finally begin to operate in intentional service to species beyond humans exclusively.

Current approaches to cross-species intelligence assessment typically rank organisms along a single axis calibrated to human cognition, systematically underrepresenting species whose intelligence is organized around capacities that do not map onto human cognitive categories. This paper presents Perceive, Relate To, Apply (P/R/A), a three-dimensional framework grounded in biosemiotic theory for mapping intelligence configurations across species without hierarchy. Perceive captures the capacity to detect linear and nonlinear information; Relate To captures the semiotic process through which perception becomes evaluated significance; Apply captures how intelligence manifests through behavior and adaptation. The framework draws on Cohen's 54 natural senses and employs a structured diagnostic, given what we can observe about how a species uses information, which dimension appears to play the organizing role?, applied through iterative AI-assisted synthesis and human expert recalibration across 134 species spanning eight kingdoms. The dataset reveals that 58.2% of species are Relate-primary, indicating that meaning-making appears to be the primary organizing dimension for the majority of life. This pattern holds across kingdoms (53% of animals, 83% of plants, 100% of fungi) and is ecologically validated: intelligence configurations track ecological function independent of phylogeny (partial Mantel r=0.067, p=0.006; PERMANOVA R²=0.226, p<0.0001). Monte Carlo perturbation analysis confirms the Relate-primary plurality is robust across 5,000 iterations at two noise levels. Five systematic AI bias patterns documented during recalibration — including the consistent conflation of meaning-making with perception — constitute an independent contribution to AI-assisted research methodology. The findings suggest that the gap between artificial and living intelligence may be architectural: current AI systems are structurally Apply-primary, omitting the meaning-making dimension that biological intelligence prioritizes.

What if we can position AI to enhance human awareness and not just "translate" signals? Current AI approaches to animal communication position AI as decoders translating animal signals for humans—what if this paradigm eliminates data crucial to understanding? This signal-based model systematically excludes unquantifiable awareness-based information that may carry primary meaning in interspecies interactions, creating what we term "unconscious bias against unmeasurable data." By using AI to reorganize existing and emerging data with respect to our framework, we propose a paradigm where intelligence is not viewed as problem-solution thinking but is instead recognized as a relational matrix. We enable three-way collaboration between human awareness, AI pattern recognition, and species or ecosystem expression. Rather than AI-as-translator, we’re developing AI-as-bridge through the perceive-relate to-apply framework. This shift addresses methodological bias in animal communication research that currently excludes complex behaviors not fitting code-like signal patterns. Our approach introduces bias acknowledgment protocols as a formal scientific contribution while maintaining empirical rigor. Case studies with equine partnerships demonstrate measurable outcomes including physiological synchronization and therapeutic effectiveness, validated through more than ten peer-reviewed studies. Aligning with the Center for Humane Technology’s research on regenerative design, our approach demonstrates how relational AI development naturally creates ethical safeguards against species exploitation by embedding relational awareness into the technology’s foundational architecture. Five prototype web applications pro- vide proof-of-concept that offer researchers direct experience with this relational intelligence framework. This paradigm offers humanity pathways to re-engage the family of life as a contributor to communication through expanded intelligence rather than remaining separate through anthropocentric control and dependence on AI translation.

Listen Like a Horse: Relationships Without Dominance presents an alternative to dominance-based models of human-animal relationship, drawing on over two decades of the author's professional practice in interspecies intuitive communication. At the heart of the book is a horse labeled "dangerous" by his humans, who asked the author to "fix" him. What emerged was not a fixed horse but a revelation: the horse was reflecting back the very behaviors and blind spots his humans could not see in themselves. The horse was only dangerous in an environment where the humans were not able to perceive their own outputs. This insight carries implications far beyond the barn. When we train animals from our own unexamined biases and then react with fear or frustration when the animal reflects those biases back to us, we are caught in a loop that dominance-based methods can only suppress, never resolve. The book proposes that authentic relationship with horses — and by extension with other species — requires humans to develop the capacity to perceive, relate to, and respond to non-linear information: the felt, sensory, and energetic communication that operates beneath and beyond verbal or behavioral cues. This is not to make anyone wrong. It is to point out that there is another way. Through direct accounts of human-equine partnership, Lake demonstrates that horses communicate primarily through energetic presence and body language requiring authentic intention — a form of intelligence that dominance-based training systematically overrides. The book weaves the distinction between linear information (measurable, quantifiable, predictable) and non-linear information (felt, intuitive, relational), arguing that human over-reliance on linear processing narrows our capacity to participate in the broader communication already occurring between species. The pattern documented here — a system builder whose unexamined biases become embedded in the system, which then reflects those biases back in ways the builder cannot recognize — would prove foundational to the author's later work on AI architecture. Current AI systems are built from the perceptual patterns of one species, inheriting that species' blind spots as structural features rather than recognizing them as potential limitations. The Perceive-Relate To-Apply (P/R/A) framework for cross-species intelligence mapping, first formalized in 2023 and presented at NeurIPS 2025, emerged directly from the experiential methodology described in this book: the practice of expanding human awareness before attempting to build systems — whether training programs or technologies — that interact with other forms of intelligence. Listen Like a Horse represents the lived foundation from which the theoretical and empirical framework emerged. The horse who was called dangerous became the clear mirror.

Various dictionaries define intelligence in terms of reason, knowledge, and the ability to solve problems. Genius is defined in similar terms of exceptional intellectual ability with mild consideration for a genius of spirit. Each of these definitions binds the idea of intelligence to a creature with a mind that processes information in ways similar to humans, leaving other life forms to be defined by different standards. Humans seem to be the only species interested in relating to the world primarily through abstracted definitions. It is one application of a linear intelligence but may not relate to non-human species. Evolution of human consciousness may be dependent on our capacity to relate to life outside of known definitions and apply wisdom that is relevant to more than one species. Even so, we cannot deny the use of definitions as a way for human intellect to relate to the workings of life and to one another. In this light, how can a new definition of intelligence contribute to the evolution of human consciousness? Intuitive communication with other species exposes humanity to ideas and relationships with life that expand beyond definitions and conclusions drawn by humans alone. World views that consider the whole of life offer an intelligence that functions in relationship rather than by definition. Interspecies intuitive communication (IIC) and the study of its function and application may begin to reveal a broader-than-human intelligence based in the wisdom of non-linear, or immeasurable, information as it is supported by linear, or measurable, information. With support from the study of IIC, we may explore a relational definition of intelligence as: Intelligence: with respect to wisdom that is common to all of life, (a) the capacity to perceive linear and non-linear information as relevant to the function of life tasks, creativity, and relationships; (b) the capacity to relate to linear and non-linear information as relevant to the function of life tasks, creativity, and relationships; (c) the capacity to apply linear and non-linear information with relevance to life tasks, creativity, and relationships.

Current AI architectures process information through a structural pipeline that moves from perception to action with no architecturally distinct step for contextual meaning-making. This paper argues that this omission is not a safety problem solvable by policy constraints, but a foundational design limitation inherited from modeling AI on a single species’ cognitive pattern. Drawing on empirical mapping of intelligence across 153 species using a three-dimensional framework—Perceive, Relate To, Apply (P/R/A)—grounded in Michael Cohen’s 54 natural senses, a comprehensive taxonomy of sensory modalities across all kingdoms of life developed through decades of applied ecopsychology research (Project NatureConnect), and formalized through category theory, we present evidence that 57% of mapped species are Relate-primary: their intelligence operates primarily through meaning-making rather than computational processing or behavioral output. We propose P/R/A as a foundational computational priority structure for AI, where meaning-making is architecturally prior to output generation at every scale of processing. This self-similar (fractal) architecture mirrors how biological intelligence organizes across cellular, neural, organismal, and ecosystem scales. We distinguish convergence-as-coordination (current AI’s reduction of variance toward shared optimization targets) from convergence-as-recognition (preservation of diversity as signal within shared structural patterns), connecting this distinction to the Platonic Representation Hypothesis and biosemiotic theory. We situate P/R/A relative to existing work in fractal network architecture, homeostatic AI, bio-inspired computing, and open-ended learning, identifying meaning-making as the missing integration point across these research streams. The paper concludes with an invitation for collaboration on the engineering translation of this biologically grounded framework into implementable AI architecture.