What supertasks are

    My reading framework is presented in the functional-computational-representational fashion that I spoke of in Chapter 2. To accomplish this, I need to describe the tasks which make up reading and the knowledge which they make use of. However, when dealing with a behavior as complex as reading, the task level breakdown described in Section 2.5 is not always efficient. That earlier approach divided the study of a cognitive behavior into describing the set of tasks which is sufficient to explain the behavior, as well as describing the knowledge which is necessary to support the task set. Each task represents a function which needs to be accomplished in order to explain the high-level behavior one is interested in; as such, an individual task has a somewhat simple goal. I could attempt to explain reading at this level. However, reading contains a small number of sub-behaviors, smaller in scope than the entire reading process but covering a larger area than an individual task. Many of these sub-behaviors have been discussed when overviewing the related research on the topic of reading; as Chapter 3 showed, a number of researchers chose to focus on these sub-behaviors. I want to target this higher level of behavior description--if I elected to remain solely at the task level, the number of individual tasks needed would be overwhelming.

  I call this higher level of description a supertask. A supertask is a collection of tasks which act together to produce a piece of the overall reading behavior at the sub-behavior level I just described. In order to develop my theory of reading and the creature dimension of it, I required a description of six supertasks, as well as understanding how the six are related to one another. Briefly, the supertasks are (refer to Figure 23 for the functional model of reading provided by the supertasks):

  

Figure 23: Functional model of reading. The hexagons represent supertasks, the ovals are structures which are produced by the process, and the lines represent pathways of communication.

 
• The control supertask is responsible for organizing the activity of the remaining supertasks to ensure that they are acting towards a common goal, as well as handling various meta-level aspects of the comprehension process. This supertask also produces the representation capturing the cognitive process which the reader undertook in order to comprehend the story.

   

• sentence processing is the supertask responsible for interpreting English sentences and sentence fragments, including the understanding of punctuation. It is given a sentence (or phrase) by control and returns a concept designed to represent the meaning of the English words.

   

• The story structure comprehension supertask handles the tasks related to the fact that stories are cultural artifacts and contain certain expected elements, such as characters and settings. This supertask produces the representation which captures the structural information of a story.

   

• The scenario comprehension supertask is responsible for comprehending events which occur; in the case of reading, these events occur within a story. This is the supertask which produces the representation of the story corresponding to the events which occurred within it.

   

• reasoning is the supertask which understands individual concepts. This allows the individual concepts to be integrated into the comprehension being built for an entire story. Other supertasks in the system may request that reasoning perform understanding on some concept.

   

• Finally, the memory supertask is responsible for the retrieval and storage of concepts from long-term memory. Additionally, this supertask handles the possibility of spontaneous retrieval from memory. As all memory requests are given to this supertask, the other supertasks may directly request action from it.

This particular set of supertasks was selected due to their support of the overall reading process to a level which enabled me to explore creative understanding. The six supertasks are not intended to be the definitive description of all reasoning needed for reading to occur successfully. They were developed because they fill specific functional roles in my overall theory. For example, the sentence processing supertask is required to take English phrases and sentences and produce a representation which the rest of the model could use. In other theories of reading, the functionality of this supertask may be more tightly integrated into the overall reading behavior; for my purposes, it was sufficient to act as a ``front-end'' to the rest of the system. Similarly, memory and reasoning exist in order to give the model general-purpose reasoning capabilities, to a level which is sufficient to support the   creative understanding process. The story structure comprehension and scenario comprehension supertasks exist in order to produce the output of the reading process. Finally, the control supertask was needed to organize the activity of the rest of the system. It was also discovered to be useful to place the reflective portions of the model in this supertask, based on what it has access to during reading.

A given instantiated model of my creative reading could choose to implement the functionality of these supertasks in different ways. For example, a single monolithic process could be designed which would encompass the entirety of the six supertasks. Or, one could use an architecture which more closely mirrors the supertask division of labor; this is the approach I take with the ISAAC model. But, any implementation of reading at the level that I am dealing with (the high-level, conceptual aspects of the process) will need to address the functionality I am describing in the supertask breakdown if it is to handle the narratives I use. My particular implementation is done to a level which is sufficient to support the core of my creative understanding theory.

Each supertask is made up of a number of individual tasks which work together to achieve the goal of the supertask. In many respects, supertasks work independently of one another; an exception is the pipelining which is required for the   sentence processing supertask as it must act as an interface between the textual world and the reasoner's internal world. But, beyond that, the supertasks can treated as individual elements within the larger reading framework.

However, it is functionally beneficial if the supertasks have a method of communicating among themselves, thereby allowing them to be loosely interconnected. In this fashion, any supertask can immediately take advantage of information which is being given to the reasoner, either through the direct process of reading or through the discovery of some other supertask--for example, the   story structure comprehension and   scenario comprehension supertasks will later be seen to often support each other. Thus, there is strong functional evidence that the supertasks should communicate, for the sake of efficient comprehension. In addition to this communication, supertasks will also need access to specific background knowledge which will permit them to function. The knowledge requirements of each supertask will differ, but each is knowledge-driven.