22.2 Microscopic mentality

At the ‘microscopic’ level, every idea is an object within a larger structure of multiple generative levels. If we focus on a specific level or degree Q, say, then we must recognize that there are prior generative dispositions P (prior), and also subsequent dispositions R (results). The level P dispositions produce the level Q dispositions conditionally on what is already existing at Q. Similarly, the level Q dispositions produce the level R dispositions conditionally on what is already existing at R. There will be other higher dispositions O, and other derivative dispositions S that may sometimes be relevant. All these dispositions are in fact some kind of love that determines what its effects are. Furthermore, they have some form or internal structure that contributes to the nature of the idea.

I envisage that a mind consists of a small set of mental objects existing within the mental body. Each object is a perception or an idea. Its meaning or content is given in part by its internal form and in part by its relation to other ideas that are in the mind or in memory. The internal content consists of the particular dispositional-substances that make up the idea and of the structure of those substances within the idea. The ability of an idea or perception to represent something else (in the mind or in the world) depends on that particular structure. The ability of an idea or perception to enter into dynamical processes with other ideas depends on the nature of its dispositional-substances. One example of a ‘dynamical process’ is the process of logical thinking, but there are many other processes which also occur, from imagination to planning to perceiving to accomplishing tasks in the world. The extent to which ideas enter into such dynamical processes of the mind will vary from idea to idea and from time to time. We may attribute some kind of ‘activation level’ to each idea in this way, measuring how much each idea is disposed to have strong effects on others.^22.2We say that strongly activated ideas are in the forefront of our consciousness, or in our short term memory, or in the ‘global workspace’ of Baars (1988).

Each level of ideas Q exists in its own space, with its own metric and topology. This is not a physical space but a space of its own kind for each level. Spaces, in general, are sets of possibilities of interactions: this is a definition equally valid for physical and mental spaces. I envisage that the precise details of mental spaces depend on the particular previous operation of prior levels O, P, in ways to be investigated. I only note that this means that, just as with physical space according to general relativity, the metric and topology are not necessarily constant but might well (and often do) change with time according to to the specific processes which have taken place. As a result of these considerations, the metric of mental spaces is expected to make them associative spaces, in which similarity of content implies that the objects are more likely to interact and hence that they are closer together in mental space. It remains to be determined whether similarity should be measured in terms of agreement of all aspects of an idea’s meaning or in terms of the strong agreement of just a few aspects of meaning. The later option would enable ideas with just a few similar components of meaning to be more likely to interact with each other: that seems to be what is observed in minds.

There is an enormous amount of detail to be discovered in the operation of minds, and consequently there is already a very extensive psychological literature discussing the nature and modeling of human mental processes. In this section, I can only hint at a few of the necessary ingredients that allow us to think correctly about minds within our theistic and generative framework and hence get these models started on a realistic basis. Up to now, the basis of mental activity has been obscure. Some researchers prefer to model in terms of beliefs, desires and actions. Others prefer explanations in terms of the processing of symbolic information. Still others try to explain mentality in terms of networks of connections, whether or not the meaning resides entirely in terms of the relational context. In the theory of this book, each idea does have its own symbolic content, but ideas can also be arranged in networks according to how they cause other ideas’ activation levels to vary with time. Let me begin by discussing the simple ingredients for the processes of reflective awareness, mental associations, production rules, and memory.

As laid out above, all ideas Q are produced as derivative dispositions from some prior (upstream) level P, and in turn produce lower (downstream) levels R as results. When a new idea Q₁ first exists, for example, it may well provide the circumstance for some previously-dormant prior degree P₁ to operate. This is the minimal scenario for us to say that the higher level P₁ ‘sees’ the new idea Q₁. Chapter 25 will discuss perception more generally and give the reasons why this process is called seeing or perception and why multiple levels need to be involved for seeing. This particular perception at prior level P₁ can also be called the reflective awareness of processes at level Q.

If the activated higher-level idea P₁ operates, it will produce new idea Q₂ in the same level as Q₁. If we were not aware of multiple levels, it would appear that Q₁ produces Q₂ directly, by a kind of direct ‘associative reflex’, but life is not that simple. In general there is a sequence , where the first arrow is a conditional activation, and the second arrow is a derivative generation. The P₁ idea is in fact necessary for the associative reflex to act in a regular manner, and it can well be called a rule for generating Q₂ whenever Q₁ appears. This pattern can be taken as the basis for much more complex and flexible production rules, detailed planning, and general modular functions in the mind. Each prior degree can be regarded as a set of rules for processing patterns of activated ideas in lower-level degrees. The ability to have functions existing in the mental architecture has long been assumed by those who model in terms of symbol processing but is difficult to achieve in a general way within purely connectionist networks.^22.3The rules here, it should be noted, are sensitive not only to the overall structure of the input ideas but also to their internal dispositional content and hence to all aspects of their meaning. This means that the new rules are not purely syntactic but are sensitive to the semantic content of the initial ideas that start the operation of the rule.

Ideas enter into long-term memory if they were sufficiently activated at some earlier time. This memory must consist of declarative as well as procedural memories, which in our context means that ideas Q as well as production rules P can be remembered. Memory is generally considered to result from the production of permanent effects or traces in some downstream (lower) substratum S, such that later processes involving that substratum may lead to the activation of new ideas Q’ or P’ that are sufficiently similar to the initial mental contents. In our multilevel generative framework, much mental content eventually produces effects at lower levels R and S, first in the mind, then in the brain, and hence finally as effects in the physical world. That is their overall purpose. Memory traces will therefore exist if the effects in those lower levels are permanently stored. It remains a topic of research, though, to find out which levels in particular are used for this storage process. Once the traces are formed, wherever they might be, then retrieval will occur in the future when the traces are the precondition for the operation of some prior degree P. If the operation of parts of that degree is conditional on the traces in memory, this leads to the re-formation of the remembered Q’, etc.

This is a barely-adequate sketch of the process: memory retrieval is most commonly driven by particular associations which retrieve memories with some semantic features in common. If this is to happen, then something about the memories, some ‘weakly activated’ idea q, say, must continue to exist at the level Q in order that its position in the associative space at that level lead it to be retrieved on the basis of possible associations.

There are many more mental processes, including symbols themselves, language, grammar, etc., that should soon be addressed in the framework of this book. Languages, for example, seem to have their own self-contained generative modules for their semantic, syntactic and phonetic generation of spoken content. There is also the ability to make names that can be associated with almost any other idea in the mind with only limited reference to that idea’s actual content. There are whole research programs waiting here.