This chapter delves more specifically into the essence of emergence. Ellis points out that emergence requires three aspects: modular, hierarchical, and structure. He explains that “hierarchical structures have different kinds of order, phenomenological behavior, and descriptive languages that characterize each level of the hierarchy.” As a simple example, consider the hierarchy of elementary physics, nuclear physics, atomic physics, and chemistry. Each layer of the hierarchy has a different kind of order and behavior. Elementary physics combines quarks into protons, neutrons, etc. Nuclear physics combines protons and neutrons into nuclei while atomic physics looks at nuclei coupled with electrons. Chemistry emphasizes the combination of atoms into molecule structures. And so on. A different descriptive language is required at each level. The concepts at one level would not be appropriate for a different level. Describing water molecules in terms of quarks would be useless.
Ellis provides a diagram of how one might describe the hierarchical levels in both inanimate and biological entities:
Ellis notes that it is unclear what the topmost and bottommost levels of the hierarchy are. To find bottom, one might need to go to quantum gravity or something and that may not be the lowest either. The top appears to be a metaphysical level and who knows if there is more to the story. But the concepts are nevertheless effective without knowing the limits of the levels.
At each level, the dominant entities form modular structures that comprise the entities of the next higher level in the hierarchy. The elementary physics level is comprised of quarks that combine into a modular form of protons and neutrons which comprise the entities of the nuclear physics level, etc. Working at the nuclear physics level, one can ignore the structure of quarks and study the nucleons as a module. Ellis describes this as “information-hiding” in the sense that a higher level can hide the detailed information in the next lower level. The high levels of macroscopic materials need little or no specific information of the individual atoms and molecules. That information is conveniently hidden.
Abstraction occurs whenever a module can be treated as a single unit and referred to by some appropriate label. In the simplest case above, the label “proton” refers to the module of three specific quarks. Encapsulation is whenever the internal workings of an abstract module are completely hidden. No characteristic of the module depends on the details of the components.
Bottom-up causation is well-known and is generally considered unique and sufficient in a reductionist perspective. The atoms and molecules behave according to well-known laws of nature and that action comprises the characteristic of the cell and so on up the hierarchy. Strong reductionism would claim that all behavior could be explained by such bottom-up causation. Such explanations are limited only by the complexity of expressing the equations.
Ellis claims “the phenomenon of emergent order is when higher levels display new properties not evident at the lower levels…higher level structures are created out of lower level entities and then exist as entities in their own right. They are described by suitable higher level variables.”
Variables that describe traits are “structural if they are basically of a static nature—they give the higher level its identity; and they are dynamic if they are essential to its behavior—they are time dependent in crucial ways.
Bottom-up causation “is the ability of lower levels of reality to have a causal effect on the higher levels which emerge from them, sometimes uniquely determining what happens at the higher levels.
Top-down causation “is the ability of higher levels of reality to have a causal power over lower levels.
“Emergence of complexity takes place where quite different laws of behavior hold at the higher levels than at the lower levels.” In my field of condensed matter physics, I understand this as phenomena like semiconductivity or superconductivity of macroscopic solids represent fundamentally different laws of behavior than those at the level of individual atoms. Yet the higher level is entirely determined by the behavior of the lower level. The key is that there is also top-down causation in the sense that that higher level behavior is context-dependent. External constraints limit the action of the elements of the lower level.
Multiple representation occurs when many different microstates of one level all correspond to the same higher level state. For example, many different combinations of individual molecular positions and velocities correspond to the same volume and pressure of a gas. The number of different lower level representations determines the entropy of that state.
“…emergence is when phenomena arise from and depends on more basic phenomena yet are simultaneously autonomous from that base…A phenomenon is emergent if it cannot be reduced to, explained, or predicted from its constituent parts.”
Ellis argues that top-down causation can occur if and only if there are equivalence classes of lower-level states. Consider a state in some level that can result from many different lower-level states, as in the case of multiple representation. This state “must lead to the same top level outcome, independent of which lower level states instantiates the high level state.” He calls this the Principle of Equivalence Classes. In other words, changing a high level variable will lead to a change in the lower-level equivalence class for that variable but this change cannot depend on which member of the equivalence class exists at the time of the change. Only in this way can there be consistency and reproducibility of the effect of changing a high level variable. Only then can top-down causation work and in that case, top-down causation does happen.
Ellis discusses four ways in which top-down causation can be demonstrated.
1. Altering context. Changing a high-level variable reliably induces change in the system.
2. Identifying equivalence classes. One can show directly that equivalence classes exist.
3. Identifying dynamics. One can show specific mechanisms that effect top-down control or specific feedback control systems.
4. Computer Modelling. One can model hierarchies and demonstrate the top-down causation.
Finally, he points out that the lower levels set constraints on the properties that higher levels can realize. For example, matter and energy conservation laws at the low levels put limits on the higher level possibilities.
This chapter is more theoretical than chapter two. He introduces and defines many core concepts of emergence. There seems to be a lot of overlap with chapter one, largely due to the independence of each chapter. Here he expands in detail the concepts of emergence and complexity mentioned in chapter one. It’s not easy reading but I’ve almost deluded myself into thinking that I might actually be starting to understand it.