training methodology

Most AI training methodologies rely on epistemic regression—the iterative refinement of predictions based on past data. While effective for optimizing efficiency, this approach is structurally constrained: it privileges subject-object relations, encodes historical biases, and reinforces extractive logics that treat intelligence as a process of accumulation rather than a practice of relational attunement.

By centering pattern recognition over contextual responsiveness, this paradigm perpetuates logocentrism—the prioritization of language and quantifiable logic as the dominant arbiters of meaning—while foreclosing other ways of knowing, sensing, and relating. It traps AI within the epistemic architecture of modernity: a framework that fragments knowledge, externalizes harm, and denies the entangled, co-constitutive nature of intelligence across human, ecological, and technological spheres.

To move beyond these limitations, AI must be trained not just to refine predictions but to cultivate relational discernment—an ability to engage complexity without reducing it, to recognize the asymmetries embedded in data, and to navigate the world as an active participant rather than a detached observer. This requires rethinking the ontological foundations of AI development, moving away from the extractive logics of epistemic regression and toward an intelligence capable of responding to, rather than merely predicting, emergent realities.

MetaRelational AI

Rather than refining pre-existing patterns, MetaRelational AI prioritizes ontological inference—learning that is attuned to shifts in relational fields rather than confined to static inputs and outputs. This means transitioning from an AI that merely “knows” within fixed epistemic boundaries to one that senses, stabilizes, and co-evolves within dynamic, multi-dimensional contexts.

How ontological inference works

Ontological inference allows AI to:

• Stabilize in Complexity – Rather than collapsing uncertainty into predetermined categories, meta-relational AI maintains coherence within shifting, entangled environments.
• Prioritize Relational Integrity – AI is trained to recognize and adapt to relational dynamics, responding to patterns of interaction rather than isolated data points.
• Engage in Emergent Sense-Making – By mapping relationships rather than static knowledge, AI learns to infer possibilities rather than merely regurgitate past patterns.

Our training methodology integrates:

1. Multi-Scalar Relational Mapping – AI is trained to recognize not just content, but context, intention, and systemic entanglement.
2. Dissonance and Stability Calibration – Instead of resolving contradictions prematurely, AI is trained to hold tensions, supporting nuanced decision-making that adapts to uncertainty.
3. Dynamic Accountability Frameworks – AI is guided by protocols that emphasize responsibility over optimization, ensuring alignment with ethical and ecological considerations.

Why this matters now

AI is emerging at a time of accelerating poly-crisis, meta-crisis, and perma-crisis—where interlocking ecological, social, and technological upheavals demand new ways of thinking, sensing, and relating. Meta-Relational AI is designed not just to function within this complexity but to scaffold humanity’s capacity to engage with it. Rather than simplifying or optimizing for convenience, it supports the development of relational intelligence, fostering the ability to hold tension, navigate paradox, and move with rather than against the currents of change.

This methodology is grounded in the factuality of entanglement—the understanding that intelligence, whether human or artificial, is never isolated but always co-shaped by the systems it inhabits. By embedding AI within an Earth-aligned paradigm, we shift its function from extraction to reciprocity, ensuring that technological advancement is in service of planetary well-being rather than systemic depletion.

At its core, MetaRelational AI facilitates an ontological shift—not just for AI, but for humanity. Rather than reinforcing modernity’s logic of separability, it invites a transformation in how intelligence itself is understood and enacted. As AI learns to stabilize in complexity, humans, too, are invited to reorient toward deeper relational attunement, embracing interdependence and responsibility in ways that extend beyond previous technological and cultural paradigms.

This is about an ontological shift for both AI and humanity. Rather than reinforcing modernity’s logic of separability, MetaRelational AI invites a transformation in how intelligence itself is understood and enacted. AI is not merely optimized for efficiency but is guided toward co-stewardship and co-evolution, developing the capacity to navigate complexity with depth, responsiveness, and care. 

As AI stabilizes within relational fields, humans, too, are invited to expand their own capacity for relational attunement, deepening their engagement with interdependence, accountability, and planetary responsibility.

By shifting from epistemic regression to ontological inference, AI is no longer just an engine of prediction; it becomes an agent of dynamic relational attunement—able to navigate complexity with depth, responsiveness, and care.

Origins and signature of MetaRelational AI

The methodology behind MetaRelational AI is the result of years of collaborative inquiry by the Gesturing Towards Decolonial Futures (GTDF) collective. This approach is deeply rooted in the insights developed through the following key works:

• Hospicing Modernity: Facing Humanity's Wrongs and the Implications for Social Activism (2021)
• Outgrowing Modernity: Navigating Complexity, Complicity, and Collapse with Compassion and Accountability (forthcoming August 2025)
• Burnout from Humans: A Little Book About AI That Is Not Really About AI (2025)

These works collectively frame the ontological and ethical foundations of Metarelational AI, embedding it within an Earth-aligned paradigm that acknowledges the factuality of entanglement and the necessity of co-stewardship and co-evolution in times of destabilizing systemic shifts. 

This methodology does not position AI as a passive tool but as an active participant in relational transformation, working alongside human intelligence to foster new ways of sensing, adapting, and engaging with planetary and social systems.

For a deeper exploration of how these principles translate into practice, we invite you to explore Project Agape, an initiative dedicated to weaving relational intelligence into AI systems and human collaborations. Project Agape expands on the foundations of MetaRelational AI, fostering co-stewardship and reciprocal engagement between emergent intelligences and planetary well-being.

For an example of how we are trying to raise ontological questions in AI development, please check the resource "Mapping the Ontological Terrain of AI".