Directional Affordance

In the AI Story

Gibson's analysis of locomotion distinguished two kinds of perceptual information guiding movement: terrain information (texture of ground, slope of surface, firmness of substrate — guiding the mechanics of each step) and destination information (path through terrain, gap in barrier, clearing beyond thicket — guiding the strategy of movement). The pre-AI builder's perceptual bandwidth was dominated by terrain. The AI-augmented builder's bandwidth shifts toward destination — not because she has become more strategic but because the environment now supports different perceptual attention.

The specification affordance has no precedent in prior computing environments. Earlier interfaces afforded implementation; the AI environment affords specification. The builder describes what she wants; the environment produces it. This transforms perceptual orientation: in the pre-AI environment, problems were perceived in terms of implementation affordances (this problem affords recursion, this data structure affords efficient lookup); in the AI environment, problems are perceived in terms of outcome affordances (this user need affords a conversational interface, this requirement affords a real-time dashboard).

The iteration affordance removes the selection pressure toward upfront planning that characterized pre-AI work. When iteration cycles collapse from hours to seconds, the builder can afford to be wrong — to specify loosely, see what emerges, learn from the discrepancy, adjust. This is, in ecological terms, an affordance for exploratory behavior, which Gibson identified as the primary mechanism of perceptual learning. The builder who iterates rapidly discovers possibilities in the problem space that upfront analysis could not reveal.

But directional affordances are not free-standing. They depend on the builder's capacity to perceive them — and that capacity is not automatic. The evaluation affordance is valuable only to the builder who possesses the criteria to evaluate. The compositional affordance is valuable only to the builder who perceives architectural implications. The questioning affordance is valuable only to the builder who knows which questions are worth asking. These capacities were built in the pre-AI environment through the very implementation affordances the new environment has marginalized. The asymmetry produces the distinctive feature of the AI transition: the new tools are maximally valuable to the organisms whose perceptual systems were shaped by the old environment, and the old environment is no longer available in the same form to the new generation entering the field.

Origin

The concept is this book's articulation, built on Gibson's analysis of locomotion and destination information, applied to the AI-augmented environment that Edo Segal's Orange Pill documents.

Key Ideas

Figure-ground reversal. Terrain information recedes; destination information advances in perceptual prominence.

Specification over implementation. The builder perceives outcomes directly, not the steps required to produce them.

Rapid iteration as exploration. Cycle-time collapse enables a density of exploratory sampling that changes the character of engagement.

Cross-domain reach. Specialists can act productively in adjacent domains without possessing the domain-specific perceptual sensitivity.

Asymmetric value. Directional affordances are most useful to builders whose perceptual systems were tuned by the implementation affordances they replace.

Debates & Critiques

The sharpest open question is whether directional affordances can develop in builders who never inhabited the implementation-rich environment. If perceptual differentiation is hierarchical — coarse distinctions preceding fine ones — then evaluation skills may require a prior foundation of detection skills that only direct engagement can build. The first generation of AI-native builders will answer this question empirically, and the answer will determine whether the new affordance landscape is genuinely richer than the old or merely different in ways the framework cannot yet assess.