The Emotion Machine

No matter what you are trying to do, there will be other things trying to get your attention. Some of these can be ignored, but if some are subgoals of what you are trying to do, this may require you to switch to some other Way to Think that uses different resources and bodies of knowledge. Then, once you have accomplished those subgoals, you will need to return to your previous job—but to avoid repeating what already was done, you must have retained some information about these aspects of your previous state of mind:

Your previous goals and priorities,

The representation you used for them,

The bodies of knowledge you had engaged,

The sets of resources that were active then,

The Selectors and Critics that were involved.

This means that our larger-scale model of mind needs places for keeping such sets of records. Let’s give these the name of “Cognitive Contexts.” Without them, every ‘train of thought’ would be disrupted whenever we were interrupted. In simpler brains, it might suffice to maintain no more than a single such memory, but for looking several steps ahead—or for managing larger subgoal trees—we’d need special machinery to enable us to rapidly switch among several remembered contexts.

More generally, because every hard problem keeps forcing us to switch among several different Ways to Think, a typical ‘present mental state’ must actually be part of a larger panalogy that can fluently navigate among several different points of view. In popular folk-psychology, we simply imagine all that stuff to be stored in our “short-term memories”— as though we could put such things into a box and take them out whenever we want.

However, we still do not know very much about how those “working” memories operate; there is evidence that such records are stored in various different forms and locations in different parts of our brains—but we still know little about how those various brain centers actually work. So here we shall simply assume that those short- term records are stored in some place that we’ll call the “Context Box.”[145]

If you asked Joan what she was thinking about, she might mention the subject of tidying-up, and further questions would show that she is maintaining several different representations of the kinds of changes she is planning to make. Furthermore, for her to be able to switch among these, she must be able to store and retrieve various kinds of structures like these:

Her current collection of sub-goal trees.

Some records of recent external events.

Some descriptions of recent mental acts,

Her presently active fragments of knowledge,

Simulations she uses to make her predictions.

This means that Joan’s context box for ‘tidying up’ must keep track of various aspects of that task:

Also, of course, other topics and subjects have been ‘on her mind’ for longer spans of time, so she’ll need to keep track of several of these:

Why would we need special systems like this to keep track of multiple contexts? It seems quite natural to us that after any brief interruption—for example, to answer a question that someone has asked, or to pick up a tool that you have just dropped—we can usually get back to what we were doing without needing to start all over again. It is the same when we interrupt ourselves—for example, to attend to a subgoal of a task, or briefly to think in some different way. When such a diversion is small and brief, this causes little trouble because it leaves most of our active resources unchanged.

However, a larger-scale change could cause more disruption and result in wasted time and confusion. So, as we evolved more ways to think, we also evolved machinery for more quickly returning to previous contexts. In everyday life we simply say that we’re using our ‘short-term memories’— but any good theory of how that might work must answer questions like these:

How long do recent records persist, and how do we make room for new ones? There must be more than one answer to that because various parts of the brain must work in somewhat different ways. Some memories may be permanent, while others may rapidly fade away, unless they happen to get “refreshed.” Such records also would get erased if stored in a ‘place’ that has a limited size—because then each new item would have to replace some records that are already there. Indeed, this is how modern computers work: whenever data is created or retrieved, it is first stored in a cache—a device that has been designed to be especially quickly accessible. Then, whenever such a cache gets full, its oldest records get erased—although some of them may have been copied to larger, more permanent memory boxes.

How do some memories become permanent? There is evidence that it takes hours or days for what we call short-term memories to be converted to longer-term ones. Older theories about this assumed that frequent repetitions made the original record more permanent. However, it seems more likely to me that new memories are briefly maintained in resources that act like a computer’s cache—and then, over time, more permanent versions are created in other regions of our brains. See §8-4.

In any case, some memories seem to last for the rest of one’s life. However, this could be an illusion because they might need ‘refreshment’ from time to time. Thus, when you recall a childhood memory, you often also have the sense of having remembered the same thing previously; this makes it hard for you to know whether you have retrieved an original record or merely a later copy of it. Worse yet, there now is ample evidence that those records can be changed while they’re being refreshed.[146]

How do we retrieve old memories? We all know that our memories often fail—as when you try to recall some important details but find that their records have disappeared, or that at least we cannot retrieve them right now. Now clearly, if no trace of that record remains, further search would be a futile quest. Nevertheless, we frequently manage to find some clues that we can use to reconstruct more of those memories. Here is a very old theory of this:

St. Augustine: “But what happens when the memory itself loses something, as when we forget anything and try to recall it? … Perhaps the whole of it had not slipped out of memory; but a part was retained by which the other lost part was sought for, because the memory realized that it was not working so smoothly as usual, hence, it demanded the restoration of what was missing. For example, suppose we see or think of some man we know, and, having forgotten his name, try to recall it—but some other thing presents itself, which was not previously associated with him; then this is rejected, until something comes into the mind which better conforms with our knowledge.”

—Book 10 of Confessions, 427 AD.

So if you can link a few of those fragments together, you may be able to reconstruct a good deal more—

“... by gathering together those things that the memory already contains but in an indiscriminate and confused way, and now putting them together [so that] where they formerly lay hidden,