In this chapter, Ellis works through seven different types of top-down causation:
1. Deterministic top-down causation
2. Non-adaptive feedback control
3. Adaptive selection of outcomes
4. Adaptive selection of goals
5. Adaptive selection of selection criteria
6. Complex adaptive systems
7. Intelligent top-down causation
These seven types show a progression of very simple deterministic control to a far more dynamic and interactive type.
· The deterministic type is the easiest to understand and sets the example. He discusses machines, physical systems, living systems, logical systems, mathematical models and randomness and noise. Machines are the archetypal examples of deterministic causation. Even for living systems, there is a basic deterministic level. Ellis describes it this way “…biology is based on molecular machines at the lower level, which behave in a deterministic way and affect higher levels in a bottom-up manner. This happens through physics processes at the lower levels in the context of systems structured so as to have specific functions. The outcomes depend crucially on context. Again it happens by:
o Setting boundary conditions for differential equations.
o Setting values for contextual variables.
o Passing signals via messenger molecules.
o Constraining lower level causation through structural conditions.”
· Feedback control systems are familiar to all of us. The simplest is the thermostat which we use frequently. Our world is filled with control systems from the simple thermostat to complex engineering control systems to biological systems where there are many homeostatic systems. “Feedback control systems depend essentially on information flows from system sensors to the controller.” Feedback control, Ellis says, is top-down causation because of two factors:
o Effectiveness of goals which are at a higher level than the controlled system
o The system acts as a whole.
· Adaptive systems. Ellis discusses four different types of adaptive selection processes:
o Adaptive selection of outcomes
o Adaptive selection of goals
o Adaptive selection of selection criteria
o Complex adaptive systems
In many ways, this is the heart of the book. Essentially adaptation is a feedback control system in which the feedback system changes or adapts by changing outcomes, goals, criteria or a combination of them. He notes that “adaptive processes…take place when there is
· Variation of Interacting Entities
· Selection of Preferred Entities”
At lower levels of the hierarchy, entities such as protons and neutrons have virtually no variation. They are each the interactions of three component quarks but there is no variation. Atoms in molecules have very limited quantized variations possible but no significant variation. More complex entities, like snowflakes, come in a large variety of shapes but they do not change and their differences in configuration have little impact on their properties. But at higher levels of complexity, entities can differ from each other and can change in response to external stimuli.
Ellis eloquently charts the set of adaptive processes and shows its power, for example, in generating new information. “The key process is deletion of what is not wanted, leaving what is meaningful. It is also for this reason that it can innovate. The process generates new information that was not there before, or rather, finds information that was hidden in noise.”
In the section on adaptive selection of selection criteria, Ellis delves deeper into learning theory and into the mind itself.
“Between them, ethics, aesthetics, and meaning form the topmost level of the hierarchy of adaptive selection criteria…They are the highest level abstract principles that are causally effective in the real physical world, crucially guiding what happens in choosing goals at all levels.”
Ellis collects all these different levels of selection into the category of “Complex Adaptive Processes”. He explains that “Adaptive processes take place when many entities interact, for example, the cells in a body or the individuals in a population, and variation takes place in the properties of these entities, followed by selection of preferred entities that are better suited to their environment or context.”
Finally, Ellis comes to intelligent top-down causation. This occurs when symbolic representation guides what happens. “A symbolic system is a set of structured patterns, realized in time or space, that is arbitrarily chosen by an individual or group to represent objects, states, and relationships.”
This chapter is one of the longest and difficult to read. It is really the heart of the book whose purpose is to show that top-down causality exists and is ubiquitous. Part of it is catalog style which makes for choppy reading but it does demonstrate the incredible range of top-down causal processes. In this chapter, and so far in this book, Ellis hasn’t attempted to show how these systems came into being. Rather he is describing what systems are like and showing that these causal factors play a key role in virtually all the systems we build and observe.