?Ts?ui Pen must have said once: I am withdrawing to write a book. And another time: I am withdrawing to construct a labyrinth. Every one imagined two works; to no one did it occur that the book and the maze were one and the same thing.?

(1) In the Steven Spielberg movie entitled Artificial Intelligence, a robot boy asks an artificially intelligent system named Dr. Know (?there?s nothing he doesn?t?) for help finding the Blue Fairy of the Pinnochio fairy-tale. Imagine that an artificially intelligent entity named Shari Seldon requests help from a 21^st century prototype of Dr. Know. Seldon is designing only the natural language generation component(s) of a qualitative (not equational) physics ITS. Imagine yourself as a communicating hardware/software element of this early version of Dr. Know that responds only to questions in the fields of CS/AI/CL.

(2) The Socratically-oriented ITS will assist students to plan actions in the physical world. For example, the ITS may ask ?A man running in a straight line at a constant velocity throws a pumpkin straight up. Where does it land and why?? The student may propose, ?It falls behind the runner because, once the runner releases the pumpkin, there is no more force on it.? The ITS must then decide what to say, how to say it, and then say it.

Shari Seldon is an artificially intelligent entity with attitude that restricts the way in which Dr. Know can provide design assistance. First, assistance from Dr. Know must be consistent with the following (over-simplified, says McKeown) pipeline model (described by Dale & Reiter) of natural language generation:

text planning è sentence planning è surface realization.

Second, assistance from Dr. Know must be consistent with the principles expressed in or implied by the four readings for the class session of 04-10-02. Finally, the initial task assigned to Dr. Know is to evaluate the following design proposal by Shari Seldon:

(1) Use the algorithms of Chu-Carroll/Carberry to select: (a) a focus for modification for determining a set of beliefs to be posted by the ITS; (b) the set of candidate justifications that the ITS will present to the student in hopes that the previously mentioned beliefs will be made mutual.

(2) Interpret the qualitative physics beliefs of, respectively, the ITS and the student as what Kashihara and colleagues describe as, respectively, an explanation structure and an understanding structure. Use these structures to compute the cognitive load that each candidate justification will impose on the student and use this information to select the justification that is used to generate a set of semantic forms.

(3) For each qualitative physics topic area, construct a sentence plan generator and ranker ala SPoT. For each topic area, interpret the set of semantic forms generated via stage (2) as a text plan input to the appropriate version of SPoT. The natural qualitative physics language generated will be evaluated by a set of human physics teachers and, perhaps, Dr. Know (since ?there?s nothing he doesn?t?).