Definition
Open-Endedness
Open-Endedness is the capacity of a system—typically an evolutionary or computational one—to continually produce novel and increasingly complex structures, behaviours, or functions without being bounded by a predefined state space or an ultimate goal.
Theoretical Foundations
The concept is central to the transition from simple replication to complex life. According to John von Neumann, open-ended evolvability requires a fundamental separation between the constructor and the instructions (the “tape”).
Observation
Encoding instructions for self-construction in a form that is itself replicated is the key to open-endedness. This allows evolution to select for arbitrary design changes that can be inherited by subsequent generations.
Requirements for Open-Endedness
For a system to exhibit open-ended evolution, several conditions are generally required:
- Instruction-Based Replication: The ability to store and copy a “Standard Description” or program (e.g., DNA).
- Excess Capacity: The replication machinery must be able to carry and propagate additional information beyond the minimum required for reproduction.
- Symbiogenesis: The ability for existing dynamically stable entities to merge and form higher-order cooperative wholes, opening new combinatorial design spaces.
- Niche Construction: The process by which organisms modify their environment, creating new challenges and opportunities for further evolution.
Distinction from Optimisation
Unlike traditional machine learning or search algorithms that seek to minimise a specific cost function, open-ended systems have no fixed “objective”. They are characterised by the ongoing creation of new dynamic stability and the emergence of “purposive” code that was not explicitly designed or anticipated.