Iteration Affordance

In the AI Story

Gibson distinguished sharply between passive reception and active exploration. Exploratory behavior generates samples of the ambient array from multiple vantages; each sample is informationally distinct; the invariants that persist across samples are what the perceptual system extracts. The density of sampling — how many distinct probes the organism generates per unit time — directly determines how much information the perceptual system has to work with.

In the pre-AI environment, the builder's exploratory sampling rate was rate-limited by implementation cost. Each probe — each tentative approach to a problem — consumed hours. A full day of engagement might yield three or four samples. The low sampling rate imposed selection pressure toward caution: the builder had to commit to approaches that looked likely to work, because the cost of wrong approaches was too high to risk casually.

The iteration affordance lifts the rate-limit. When cycle time collapses to seconds, the builder can sample the problem space with a density that was previously impossible. She can try approaches that look unlikely to work, because the cost of being wrong is now trivial. She can probe the edges of the problem, investigate counterintuitive framings, test hypotheses that upfront analysis would have rejected. The exploratory character of engagement returns to something approximating its function in natural habitats: a continuous probing that generates a rich stream of samples from which perceptual differentiation can occur.

But the iteration affordance presupposes — as all directional affordances do — the capacity to evaluate the samples it generates. The builder who iterates rapidly without the perceptual sensitivity to distinguish productive probes from unproductive ones is not exploring; she is churning. The exploration is productive only when the evaluator is tuned to detect the information the samples reveal, and that tuning — Eleanor Gibson's differentiation — was historically built through the slow, friction-rich engagement that the iteration affordance's cycle-time collapse has replaced.

Origin

The concept is articulated in this book's reading of Edo Segal's account of rapid AI-mediated iteration, read through Gibson's analysis of exploratory behavior as the engine of perceptual learning.

Key Ideas

Economics of exploration. Cycle-time collapse transforms the cost structure of probing the problem space.

Sampling density. Higher iteration rate produces denser sampling of the ambient affordance structure and more information for perceptual differentiation.

License to be wrong. Cheap iteration removes the selection pressure toward upfront correctness that characterized pre-AI work.

Returns to natural exploration. The character of engagement approaches the continuous probing that Gibson identified in natural habitats.

The evaluator problem. Rapid iteration is productive only when the builder can evaluate what the samples reveal; without tuned evaluation, rapid iteration produces churn rather than exploration.

Debates & Critiques

The debate centers on whether the iteration affordance's benefits require prior perceptual differentiation or can build it through the exploration itself. Optimists argue that rapid sampling, combined with AI feedback, can generate the differentiation directly — that iteration is itself a training regime. Pessimists argue, in Gibson's tradition, that differentiation requires authentic friction that AI-mediated iteration smooths away, and that the evaluative capacity exercised in rapid iteration presupposes a foundation that only slower, friction-rich engagement can build.