Download (direct link):
The second indexing issue is:
B: Indices can be used to find scripts (and particular episodes) in a scene.
When key or elevator is into the generalized scene CHECK-IN, $hotf.l-CHECK-IN comes quickly to mind, These indices help call in the right processing structure Thus we can imagine something like “I was dealing with some papers over a big desk and they handed me a key - oh it must have been a hotel.” Of course, it could have been a safe-deposit box in a bank too, but in a sense that’s the point. Such things come to mind given the general description of the scene in the most neutral terms (the rudiments of CHECK-IN) and a prop from the specific instance (an index to the script stored in terms of that scene). Thus, indices point to scripts from generalized scenes. And,
C: Once a script has been determined, it can be used to index its own
colorations within the generalized scene that spawned it
We have been discussing Ñ a great deal without actualy mentioning it specifically. A script colors a scene by indexing every expectation from the scene through the script. In other words we have something like
indexing and search
transfer in the case of SHOTEL-CHECK-IN leads one to expect the key Thus, just as the seven questions above can lead one to $HOTEL-CHECK-lN, so 0OTEL-CHECK-IN can lead one to each of the answers.
Thus, the initial question of how a generalized scene continually develops is that new scenes are compared against it. If there is enough of a match, the generalized scene replaces the particular scene, which becomes a script, indexed under, and coloring, the generalized scene.
So, the answer to what indices are used in generalized scenes are those mentioned in A, B, and Ñ above, which in turn depend upon the answer to a set of questions regarding the particulars of the actions described by the generalized scene.
Indexing other kinds of expectation failure
Not all expectations are straightforward. When the world does not conform to what we expected, it does not necessarily follow that what happened was unexpected. That is, we are not always attempting to predict anything and everything that might happen.
One of the most important points about an expectation is, as we have said, that it can fail. Expectations that succeed are boring. They are of use in processing, of course. They help us to make sense of the world around us. But, from the point of view of a developing memory, that is, one that makes new generalizations and learns, a successful expectation is of no help. A failed expectation, on the other hand, forces us to create indices to memories that exemplify that failure. Further encounters with those indexed memories, that is, repeated failures leading to repeated remindings, force us to alter our expectations.
The major problem is, of course, what to do when an expectation fails. Expectations brought together in one scene, when they fail, can be traced to the proper place in memory for an update of the most relevant memory structure. This allows us to modify that structure and learn from our experience.
Consider a person in a given situation. The scene in which he is operating supplies expectations relevant to that situation. (Of course, there is likely to be more than one operating scene at any one time. But we are attempting to keep things simple here.) This scene is being employed by some MOP For example in a car we might have:
SDRIVE STALK SENJOY-SCENERY
Let us assume that there are three scripts active in this scene at the same time. Now suppose a car swerves out in front of the driver. The driver reacts almost by instinct. Clearly whatever he does is part of $DRIVE by definition. But, if his later analysis of what he did makes him feel uncomfortable in some way, for example if he believes that he usually reacts correctly in crisis situations, but here did not, this analysis will modify his memory. But what does it modify?
He was in $DRlVEat the time The physical manifestation of the action he performed was encoded in $drive The act of reflecting upon what he did however, belongs elsewhere. It should modify a MOP that encodes knowledge of one’s behavior in a crisis We can call that MOP M-CRISIS-SITUA-TION. We can modify that MOP by noting that the expectation that failed in $drive was one that was of one’s own behavior in a crisis. That is, knowing the kind of expectation that failed helps us to modify the right structure, and in this case one that was not obviously in operation at the time. Thus, we might expect our driver to consider drive more carefully because expectations formed earlier about his ability to react to crises seem to be no longer applicable. In other words, we can change out world view upon reflection. Thus some MOPs can be consciously modified.
Consider another example. A hotel room is clearly a scene. When the plaster falls on you in your bed in a hotel room, you have experienced a failure of some kind. You would like to modify your memory accordingly. But what does that involve? Here again, none of the MOPs that we would like to modify were necessarily active at the moment the plaster fell. Nevertheless, we would like to possibly do the following things (classified according to what the expectations involved weie about):