One good way to build teaching systems is to populate them with stories told by experts, and then design the instruction so that the right story appears at the right time for the student to be informed by it.

When a human is confronted with a problem, questions will often come to mind. When a human hears a story or watches a video, questions will also come to mind. When we ask a question we hope to get an answer and, if we do, further questions will often come to mind.

A series of questions and answers is a dialog.

Question-based dialog networks are built in order to simulate human-like, problem-solving dialogs with a computer.

Question-based dialog networks are a kind of hypermedia comprised of

1. nodes containing content: stories, essays, articles, etc. and
2. links representing questions (each within a category)

Every story raises some questions and answers some other questions.

• questions-raised indicate information a text is missing
• questions-anwered indicate the main points a text contains

Creating a QBDN is simply a matter of matching all the questions raised in one story with the corresponding questions-answered in other stories. And then doing that for all stories in the network.

However, imagine a set of 100 stories (not a large number).

• Each story could raise 10-30 questions.
• Each story could answer 10-30 questions.

Therefore, matching could be between two sets of 3000 questions each: an intractable job.

A Theory of Conversation

Relative Indexing and Associational Indices

• Relative indexing is the method used to manage the connections between texts (as opposed to the "absolute indexing" methodology for indexing cases in an invariant way).
• Associational indices are the relational labels used for relative indexing.

Associational indices define the relationships individual texts can have to other texts: one text might provide the background for another, it may illustrate a consequent of a previous text, etc.

The idea of relative indexing is to capture the shifts of focus that can occur in a conversation, changing the topic and yet remaining on the same subject.

The set of relative indices is usually kept to a minimum in order to maintain generality. There is no universal agreement on the proper inventory, but the standard set usually includes relationships such as

1. "context" - what is/was the big picture
2. "background/causes" - what is the history, what led to this
3. "indicators" - what are/were the important points
4. "opportunities" - what are/were the possibilities
5. "results" - what are/were the outcomes
6. "examples/specifics" - what are/were the details
7. "warnings" - what could/did go wrong
8. "alternatives" - what are/were the choices
9. "analogies/stories" - what situations/problems are/were similar

Example Story: The Airframe Manufacturer

"This was an interesting problem. An aerospace defense contractor on the west coast decided they would not be competitive within a particular Air Force program unless they were able to manufacture and assemble their aircraft faster. We worked closely with both middle management and the trade union people to reorganize the plant floor in order to eliminate non-value-added steps and gain efficiencies ..." The following questions, under the following conversational categories, are arrayed before the user.

opportunities

How can continuous improvement be achieved?
How can skills be transferred to the client?
How can the lessons learned on this project be used in the future?

warnings

How were union concerns handled?
What resistance to change was encountered on this project?

causes

How was the project organized?
What qualified your team for this type of project?

results

What were the main outcomes of the project?
How long did the project take?

specifics

How was the need for change communicated to the project team?
How was the process automated?
What were the weaknesses of the current process?
How were planned changes communicated to the workforce?

Note: not every story has follow-up questions in every conversational category. This story has questions indexed as opportunities, warnings, causes, results, and specifics, but no questions indexed as either alternatives, analogies, or context.

Also Note: question categories can be thought of as paired. For example

• "background/causes" are opposite "results"
• "context" is opposite "examples/specifics"
• "opportunities" are opposite "warnings"
• "indicators" are opposite "alternatives"

Suppose the following question is asked (and answered with the story, "Problems with Middle Management"):

USER: What resistance to change was encountered on this project?

Followup Story: Problems with Middle Management

"Quite a bit of this project involved changes in procedures on the shop floor, which meant direct involvement with unions. Meanwhile, middle management was feeling out of the loop because they were feeling like they were being robbed of their people. We messed up, because middle management became afraid they didn't matter. So, if you're going to restructure the work on the shop floor make sure you load middle management with work that's relevant to the effort in a high-value way ..."

The story of "Problems with Middle Management" is also one that could lead to further mentoring dialog. However, the questions arising from this story are different from the questions arising from the Airframe story. The conversation has moved forward, and the following set of questions now arise before the user. Once again, not all conversational categories are available at every point in a dialog.

opportunities

How can a negative faction be stopped from sabatoging the changes?

causes

Why does middle management often resist the introduction of new technology?

context

When are people most willing to accept changes?
Why does resistance to change occur?

specifics

What is another example of middle management resisting changes?