Intelligent systems, such as the animal brain or a swarm of fish, are able to change their structures to accommodate new demands. When it comes to the animal brain, structural changes occur across long timescales, through natural selection and evolution and across many generations – these changes in structure trigger the creation of new functions.

Changes in structure can also happen on shorter timescales, for example in swarm-like systems. Here, specific environmental demands such as a predator threatening the collective can cause a brief change in the swarm’s active interconnections (see principle 1), thereby modifying the whole structure of the swarm. This helps the swarm respond adaptively to the environmental demands in action.

Marlin hunting sardines in nature and in a computer model. External influence causes the swarm’s structure to change ©SCIoI/Alicia Burns and Palina Bartashevich

Structural changes have also been implemented in synthetic systems that are considered intelligent. Evolutionary algorithms, for example, allow for adaptation of structure in synthetic systems.

Another aspect related to this principle is the cognitive neuroscience concept of “neural recycling” (Dehaene and Cohen, 2007), which occurs when an intelligent system “reuses” some of the specific functions that have evolved for other reasons in order to perform a new task that wasn’t specifically part of the structure’s evolutionary plans. An example of this is the act of reading. Reading is a result of cultural evolution and has developed thanks to the previous evolutionary development of structures that are involved in specific information processing (such as parts of the visual-cortex involved in object recognition). Here the cogntive structure changes as a function of cultural evolution. Circumstantial support comes from system building in robotics, where the adaptation of structure, unsurprisingly, leads to systems with different capabilities and suited to different tasks (ecological niches) (Eppner et al., 2018).

What would a non-intelligent system do instead?

For better understanding, it often makes sense to compare the intelligent systems described above with a possible non-intelligent counterpart. A calculator, for example, may be able to conduct very complex calculations that could theoretically even contain elements of yet-to-be-discovered mathematical principles, but its structure does not evolve. So, if the calculator is not originally programmed to process the available information to discover new mathematical rules, it will not do so in the future either.

A more in-depth look

The principle of active interconnections takes the components as given and leverage active interconnections for good generalization. In contrast, the adaptation of structure re-distributes the capabilities among the system’s components to improve generalization. We hypothesize that active interconnections enable generalization for structurally similar environmental variations and on ontogenetic time scales. In contrast, adaptation of structure enables generalization for substantial deviations of environmental structure. Also, the time scales for those changes can be longer, sometimes happening ontogenetically, sometimes over generations (cultural evolution) but sometimes also genetically. As an analogy from linear algebra: Whereas active interconnections combine existing basis vectors in useful ways, adaptation of structure corresponds to changing the basis vectors themselves.

The re-distribution of capabilities across components (re-factoring) changes the regularities that can be encoded and exploited within the system. Intelligent behavior is facilitated if this change represents a move toward the regularities relevant in the current ecological niche.

Relation to SCIoI projects

Project 27: Speed-accuracy tradeoffs in collective estimation
Project 33: Shepherding behaviour in predator-prey interactions
Project 41: Self-organised criticality in animal collectives
Project 42: Weighing personal and social information in cooperative problem solving
Project 57: Internal and External Visual Information Sampling Using Eye-Movements and Overt Attention in Dynamic Scenes