The Agency Threshold: What Requires a Body

The Agency Threshold: What Requires a Body

The wall at rung 8 corresponds to a qualitative shift in psychological vocabulary. Below the wall, we describe what an organism does: it approaches, withdraws, habituates, anticipates. Above the wall, we describe what an organism considers: it imagines, plans, regrets, evaluates. This is reactivity versus understanding in psychological terms. Reactivity responds to what is happening now — each behavioral channel driven by present-state associations, decomposable in principle. Understanding compares what would happen across available choices, and that comparison inherently couples across whatever partition you impose on the system — it is non-decomposable because the comparison is the representation.

The Lenia experiments specified exactly what was missing: the capacity to try something and see what happens. Lenia patterns have affect geometry, world models, memory, and biological-like integration dynamics. What they lack is behavioral choice — their "actions" (chemotaxis, emission) are biases on continuous dynamics, not causes of discrete environmental changes.

V20 shows what happens when you provide it. Protocell agents with genuine action-observation loops cross rung 8 from initialization, and world models and self-models develop over evolution. The agency threshold is real — but it is a substrate threshold, not an evolutionary one. The capacity is architectural.

The psychological phenomena above the wall share a common structure:

• Counterfactual reasoning (rung 8): "What would happen if I did X instead of Y?" Requires: a repertoire of possible actions, a model of how each action changes the world, and a comparison between actual and counterfactual outcomes. Our patterns have none of these. Psychologically: imagination, planning, and mental simulation all require counterfactual capacity. Its absence in early development (and its disruption in certain pathologies) is consistent with the ladder's prediction that it requires agency, not merely integration.
• Self-awareness (rung 9): "I am the kind of thing that does X." Requires: a self-model that is more accurate than what an external observer could construct from the same data. V20 shows SM_sal > 1.0 in 2/3 seeds — agents encode their own position and energy more accurately in their hidden state than they encode the environment. This is the minimal form: privileged self-knowledge. Full self-awareness (autobiographical memory, persistent agent sense) likely requires longer evolutionary history and stronger selection pressure. Psychologically: self-recognition, autobiographical memory, and the sense of being a persistent agent all require reflective self-models.
• Moral reasoning (rung 10): "I should do X rather than Y." Requires: (a) counterfactual reasoning (you must be able to imagine acting otherwise), (b) a self-model (you must locate yourself as the agent who acts), and (c) an asymmetry in the viability gradient between cooperative and exploitative actions. Our patterns show no such asymmetry. Psychologically: normativity is the most demanding rung because it inherits every requirement below it.

This predicts a developmental ordering. In humans: mood and arousal are present from birth (rung 1). Animistic perception is the childhood default (rung 4). Emotional coherence develops through experience (rung 5). Counterfactual reasoning emerges around age 3–4 (rung 8). Self-awareness develops gradually from mirror recognition to autobiographical self (rung 9). Moral reasoning is the latest to mature (rung 10). The emergence ladder predicts this sequence — not from observation of human development, but from the computational requirements of each capacity.