Adjacent possible

The adjacent possible is a concept introduced by Stuart Kauffman to describe the boundary between what a system is and what it could become next. At any moment, a system can only access states that are one step away from its current state in configuration space. The adjacent possible is not a fixed set of possibilities waiting to be discovered; it expands as the system explores itself, creating new adjacent states that did not exist before.

In biology, the adjacent possible explains why evolution is not a random walk through all possible genotypes. Most genotypes are not adjacent to viable ones; the space of functional organisms is sparsely connected. Evolution proceeds along the connected paths, and each step expands the adjacent possible by creating new molecular structures, new metabolic pathways, and new ecological niches.

In technology and culture, the adjacent possible explains why inventions appear when they do. The steam engine required precision metalworking, the knowledge of atmospheric pressure, and the economic need for mine drainage. These preconditions were not merely necessary; they were the adjacent possible that made the steam engine thinkable. Innovation is the exploration of the adjacent possible, and it is constrained by the same topological laws as biological evolution.

The adjacent possible connects directly to adaptive walks: an adaptive walk is precisely the process of exploring the adjacent possible on a fitness landscape. The walker cannot teleport to distant peaks; it can only move to adjacent states. The structure of the landscape — whether smooth, rugged, or neutral — determines how the adjacent possible is arranged and whether exploration leads to progress or stagnation.

The adjacent possible is not a philosophical consolation prize for those who cannot achieve their dreams. It is a physical law: the universe creates new possibilities as it explores itself, and the boundary of the possible is the most important frontier in any science of complexity.

Mathematical Foundations

Kauffman formalized the adjacent possible using NK fitness landscapes, a model in which the fitness of an organism depends on N traits, each interacting with K other traits. In this framework, the adjacent possible corresponds to the set of single-mutation neighbors — genotypes one step away in the space of possible configurations. The structure of the landscape determines whether evolution proceeds smoothly (low K, correlated landscape) or becomes trapped on local peaks (high K, rugged landscape). The key insight is that the adjacent possible is not merely a spatial metaphor but a topological property of the fitness landscape itself: it defines the boundary between what has been realized and what could be realized with minimal incremental change.

The mathematical treatment reveals a tension between exploitation (climbing the current fitness peak) and exploration (venturing into the adjacent possible, which may lead to higher peaks but requires traversing fitness valleys). This tension is structurally identical to the exploration-exploitation tradeoff in reinforcement learning, multi-armed bandit problems, and organizational search. In each domain, the optimal strategy depends on the time horizon, the cost of exploration, and the information value of new discoveries.

Technology and the Expanding Adjacent Possible

Technological innovation is perhaps the clearest domain in which the adjacent possible operates as a structuring principle. The invention of the transistor did not merely improve vacuum tubes; it opened an entirely new region of the adjacent possible that included integrated circuits, microprocessors, personal computers, the internet, and the platform economy. Each step was not predictable from the previous one — no one in 1947 could have foreseen TikTok — but each step was accessible from the one before it. The transistor made the microprocessor possible; the microprocessor made the smartphone possible; the smartphone made location-based services possible.

This pattern produces path dependence: the sequence of earlier innovations determines which later innovations are accessible, even when superior alternatives exist on other branches of the possibility tree. The QWERTY keyboard layout, the VHS format, and the internal combustion engine each became dominant not because they were optimal but because they occupied early positions in the expanding adjacent possible, creating increasing returns that locked out competitors. The adjacent possible is thus not merely a space of opportunities but a space of constraints: every realization closes off some futures while opening others.

Brian Arthur's work on increasing returns and path dependence in the economy formalized this intuition, showing that under increasing returns, small historical events can determine long-run outcomes. The adjacent possible provides the mechanism: early winners expand their own possibility frontiers faster than latecomers, creating self-reinforcing dynamics that can persist even when the underlying technology becomes obsolete.

Scientific Discovery as Boundary Expansion

Scientific instruments are tools for expanding the adjacent possible of observation. The microscope did not merely improve vision; it made the cell visible, opening the adjacent possible of microbiology. The telescope did not merely magnify the sky; it made Jupiter's moons observable, opening the adjacent possible of celestial mechanics. The particle accelerator, the DNA sequencer, and the functional MRI each operate by the same principle: they push the boundary of what can be measured, and in doing so, they restructure the adjacent possible of scientific theory.

This observation inverts the common assumption that theory precedes observation. In practice, the available adjacent possible of observation constrains the adjacent possible of theory. Einstein could not have developed general relativity without the available observational data on Mercury's perihelion; the Higgs mechanism remained speculative until the Large Hadron Collider made it observable. The adjacent possible is not a one-way street from theory to experiment; it is a co-evolving boundary in which each expansion of observational capacity enables new theoretical structures, which in turn suggest new instruments.

Creativity and the Exploration of Possibility

The creative process, whether in art, science, or design, can be understood as navigation of the adjacent possible. Creative breakthroughs rarely arise from radical leaps into distant regions of possibility space; they arise from the recombination of existing elements at the boundary of the known. Johannes Gutenberg did not invent printing from nothing; he combined the wine press, movable type, and oil-based ink — each existing technology — to create something that was in the adjacent possible of all three but in the distant possible of any one.

This has implications for how we structure creative environments. The innovation paradox is that too much freedom — too large a possibility space — can be paralyzing, while too much constraint can be stifling. The optimal creative environment provides enough structure to make the adjacent possible navigable and enough openness to prevent premature convergence on local optima. This is why constraints — whether formal (sonnet structure, twelve-bar blues) or material (limited computational resources, fixed budgets) — often enhance rather than inhibit creativity: they narrow the search space to a region where the adjacent possible is dense with valuable recombinations.

Critiques and Extensions

The adjacent possible has been criticized for its vagueness. Kauffman's original formulation was primarily metaphorical, and the concept risks becoming a post-hoc explanation: any innovation can be described as adjacent to what came before, but this description does not generate predictions. Without a formal measure of adjacency — a distance metric on possibility space — the concept risks tautology.

Recent work has attempted to address this by operationalizing the adjacent possible in specific domains. In technology studies, the technology readiness level (TRL) provides a rough index of how far a technology is from the adjacent possible of commercial deployment. In drug discovery, the chemical space of synthetically accessible molecules defines a measurable adjacent possible. In AI research, the concept of emergent capabilities in large language models — abilities that appear suddenly at scale rather than gradually — challenges the smooth, continuous image of the adjacent possible, suggesting that some transitions are discontinuous.

A deeper critique comes from the sociology of innovation. The adjacent possible is not a natural property of technology or biology but a social construction: what counts as possible depends on funding priorities, regulatory frameworks, cultural values, and power structures. The adjacent possible of renewable energy was technically accessible decades before it became economically and politically possible. The adjacent possible of gene editing was biochemically accessible before it became ethically possible. The boundary of the adjacent possible is not merely technical; it is normative.

Connections

The adjacent possible is connected to several related concepts: - Path Dependence: Early choices constrain later ones by determining which regions of the adjacent possible are accessible. - Increasing Returns: Successful innovations expand their own adjacent possible faster than alternatives, creating lock-in. - Fitness Landscape: The geometric structure that determines the shape of the adjacent possible in evolutionary systems. - Innovation Systems: The institutional and social structures that determine how quickly the boundary of the adjacent possible expands. - Serendipity: The role of chance in discovering valuable regions of the adjacent possible that were not deliberately sought. - Lock-in: The phenomenon by which a technology or standard becomes entrenched, closing off adjacent alternatives. - Exaptation: The repurposing of existing structures for new functions, a mechanism for moving into the adjacent possible. - Recombination: The creative process of combining existing elements to produce novel configurations within the adjacent possible.