Predicting perceptions for optimizing action
In preceding chapters we have discussed the role of perception, action, and feelings in the present. We also discussed how action in the present may be influenced by the past, through memory. In this chapter we shall consider the possible influence of the future. Not the actual future, of course, but action that is based on prediction of perceptions that might happen in the future. We shall try to make the case that the behaviour of some animals indicates that they make such predictions.
Let us start with a series of diagrams showing a predator chasing prey.
- We have drawn the diagrams as snapshots in successive moments of time, showing the positions and movements of the predator and the prey
- At the beginning, when the predator first sees the prey. The arrows show the directions in which they are running. We assume that the predator is guided by his perception, i.e. he runs directly toward the prey.
- For the next snapshot, both the predator and the prey have made progress in the directions they were running. At this moment the predator again looks at the prey and changes direction slightly so that he is heading directly toward the prey
- For the next snapshot, both have made further progress, and the predator needs to change direction again.
- Again, both have made further progress, and the predator needs to change direction again.
- We have now seen enough to infer that following successive sightings of the prey leads the predator in a curved line
- Following this logic we can conclude that the predator must follow such a curve if he only uses perception to guide him.
- If the predator wanted to use the shortest, and therefore the fastest path, he would have to choose the straight line between the present position and the point of interception
- Diagram 5: diagonal path for predator - prey chase
- The point of interception is in the future, and therefore requires a prediction. If the assumptions are correct, i.e. that the prey continues in a straight line, then the prediction leads to a shorter and faster chase -- an evolutionary advantage.
It is possible to verify this predictive ability for different animals with suitable experimentation.
- Relevant experimentation has already been reported in the literature
- For example, an animal can see food but cannot get there on a straight path. The animal has to figure out to go around a barrier.
- This has been done with ...
Next, let us look at perception-related prediction from an information processing perspective
- We need to be able to generate new perception sequences that have not been experienced
- The new sequence is presumably copied from and guided by past perceptions
- The simplest form presumably is to change the relationships between foreground 'blobs' and background. This kind of simple mechanism should be sufficient for the predator and prey example above.
- To generalize, we need to find a way for generating new, not previously seen perceptions
