Deliberation operates on descriptions of future reactions.
Reflective systems operate on descriptions of deliberations.
Self-Reflection operates on descriptions of reflections.
Why emphasize descriptions here? That’s because we could never learn enough low-level If-Then rules, and the only alternative is to use abstractions—as was argued in 1959 in an essay called Programs with Common Sense. [107]
John McCarthy: “If one wants a machine to discover an abstraction, it seems most likely that the machine must be able to represent this abstraction in some relatively simple way.”
We need to make our descriptions abstract because no two situations are ever the same, so as we saw in §5-2, our descriptions must not be too concrete—or they would not apply to new situations. However, as we noted in §5-3, no representation should be too abstract, or it will suppress too many details.[108]
??????????????????? We discussed how much knowledge a person could have, but perhaps it is more important to ask how we re-collect what we need so quickly when we need it?
Whenever we get a new idea, or find a new way to solve a problem, we may want to make a memory- record of it. But records are useless unless you have ways to retrieve the ones most likely to be relevant to the problems you face. I’ll argue that this needs a lot of machinery.
Citizen: If remembering is so complex, then why does it seem so effortless, simple and natural? Each idea reminds me of similar ones, which then make me think of related ideas—until I recall the ones that I need.
Why does ‘remembering’ seem so effortless? As long ago as you can remember, you could always recall things that happened to you. However, you cannot remember much of your earliest years; in particular, you cannot recall how you developed your early abilities. Presumably, you had not yet developed the skills one needs for making those kinds of memories.[109]
Because of this Amnesia of Infancy, we all grow up with simplistic views of what memories are and how they work. You might think of your memory as like a writing-pad, on which you can jot down mental notes. Or perhaps for each significant event, you store ‘it’ away in some kind of memory-box and later, when you want it back, you somehow bring ‘it’ out of that box—if you are lucky enough to find it. But, what kinds of structures do we use to represent those ‘its’ and how do we bring them back when we need them? Our recollections would be useless unless (1) they are relevant to our goals and (2) we also have ways to retrieve the ones that we need at the times when we need them.
To do this, a computer expert might suggest that we store everything in some single ‘data base’ and use some general-purpose ‘matching’ technique. However, most such systems still classify things in terms of how they are described instead of what they are likely to be useful for. The trouble with this is we do not usually know what kind of thing we are looking for, but only what we want to accomplish with it—because we’re facing some obstacle, and want to know how to deal with it.
So, instead of using some general method, I suspect that every child develops ways to link each new fragment of knowledge to goals that it might help us to achieve, as well as to other related ideas. These additional links might help to answer questions like these:
What kinds of goals might this item serve? Which kinds of problems could it help to solve? What obstacles could it help to overcome?
In which situations might it be relevant? In which contexts is this likely to help? What subgoals must first be achieved?
How has it been applied in the past? What were some similar previous cases? What other records might be relevant? See §8-Credit Assignment.
Each fragment of knowledge may also need links to some knowledge about its deficiencies—and the dangers and costs of using it:
What are its most likely side effects? Is it likely to do us more harm or more good?
How much will it cost to use it? Will it repay the effort of using it?
What are its common exceptions and bugs? In which contexts is it likely to fail us—and what might be good alternatives?
Is it part of some relevant family? [See Glossary: Ontology.]
We also link each item to information about its sources and to what other persons might know.
Was it learned from a reliable source? Some informants may simply be wrong, while others may mean to mislead us.
Is it likely to be outdated soon? That’s why this book does not depend much on current beliefs about how our brains work.
Which other people are likely to know it? Our social activities strongly depend on knowing what others may understand.
All this raises question about how we make so many connections to and from each new fragment of knowledge. I suspect that we can’t do this all at once—and indeed there is some evidence that it normally takes some hours or days (including some sessions of dream-laden sleep) to establish new long-term memories. Also, we probably add more links each time we retrieve a fragment of knowledge, because then we’re likely to ask ourselves, “How did this item help (or hinder) me at overcoming this obstacle?” Indeed, some research in recent years suggests that our so-called long-term memories are not so permanent as we used to think; it seems that they can be altered by suggestions and other experiences.
We all know that our memory systems can fail. There are things that we can’t remember at all. And sometimes we tend to recollect, not what actually happened to us, but versions that seem more plausible. At other times we fail to remember something relevant until—after several minutes or days—suddenly the answer appears—and you say to yourself, “How stupid of me; I knew that all along!” (That could happen either because an existing record took long to retrieve, or because it was never actually there, and you had to construct a new idea by using some process of reasoning.)
In any case, we should expect such ‘lapses’ because our recollections must be selective; §4-4 discussed how bad it would be to remember everything all the time: it would overwhelm us to recall all the millions of things that we know. However, none of this answers the question of how we usually retrieve the knowledge that we currently need. I suspect that this is mainly due to our already having prepared in
