Educational Technology & Society 2(1) 1999
ISSN 1436-4522

If only my intelligent pedagogical agent could ... !

Moderator & Summarizer: Scott Overmyer
Associate Professor, College of Information Science and Technology
Drexel University, 3141 Chestnut Street, Philadelphia, Pennsylvania, 18912 USA.

Discussion Schedule
Discussion: 28 Sep. - 7 Oct. 98
Summing up: 8 - 9 Oct. 98

Pre-discussion paper

This discussion is to talk about the requirements for an intelligent pedagogical agent, and it's role in an overall pedagogical strategy for distance or other on-online education. Recognizing that there are domain-specific requirements for pedagogical agents, perhaps it would be possible for us to arrive at a set of basic functional capabilities that are common to all or most agents. It would be interesting to brainstorm a "wish-list" of capabilities that we would all like intelligent agents to be able to do, and then partition the list into those things we can currently do, will be able to do in the foreseeable future, and those capabilities that may never be possible. We could then try to fit those capabilities into an overall pedagogical strategy for distance education. It's a large topic, and this start could provide a bit of structure for future discussions on pedagogical agents.

Post-discussion summary

This discussion started with a brief introduction from the moderator, and a paragraph offered as a brainstorming "seed":

" Imagine an electronic education system via which students engage in some kind of learning activity. They are studying the subject for which you are responsible. You have the opportunity to create a clone of yourself in software, and be there with each student all of the time that they are using the system. You may monitor all students simultaneously, being constantly aware of their activities and their goals (combined with your goals for them). You may watch, help, intervene, perform demonstrations, tell anecdotes, offer examples or illustrations, or perform what ever activities that you deem appropriate to aid the student's ability to learn using the system. If you could do this, which capabilities and characteristics of yourself would you give the software?"

First of all, there was a general consensus that an IPA should monitor a number of aspects of the student's activities and performance, and act upon this information depending on the role that the IPA was to play. Some of the suggested student and environmental parameters that an IPA should monitor are:

When the learner (V. Vulfs):

  • Does not trust the information being presented
  • Is bored or not paying attention
  • Is not reading or listening for comprehension
  • Is frustrated
  • Is anxious
  • Does not understand after making efforts ot do so
  • Finds the information too easy
  • Knows the information already
  • Does not find the information relevant
  • Does not feel confident
  • Finds the overall treatment inappropriate (too cute, too serious, etc.)
  • Has an emotional state that is counterproductive to learning (sleepy, angry, etc.)
  • No longer needs the service of the agent (Kinshuk)
  • Has missed an essential resource (S. Overmyer)

Having monitored these parameters, the IPA should be able to provide a number of features that are context dependent. The features that were suggested loosely fell into a number of distinct roles that the contributors envisioned for potential IPAs, and were most commonly related to the current work of the contributor. Contributors envisioned agents in roles of:

  • Information Source
  • Tutor
  • Mentor
  • Co-worker or partner
  • Motivator
  • Group facilitator

Within each of these roles, contributors to the discussion suggested a number of capabilities that they would like to see instantiated in an intelligent pedagogical agent.

Information Source
1. The IPA should provide fast access to directions on assignments. (A. Dufresne)
2. The IPA should provide help on initial technical issues for a learning session. (A. Dufresne)
3. The IPA should provide information based upon perceived student needs. (S. Overmyer)

1. The IPA should lend assistance when needed, and get out of the way when not (V. Vulfs, S. Overmyer)
2. The IPA should suggest follow-up or additional learning activities (G. Free)
3. The IPA should post context-relevant problems or questions (G. Free)
4. The IPA should reflect on a student's reflections (J. Benda)

1. The IPA should coach the approach to problem solving in a particular domain (C. Quinn)
2. The IPA should meta-learning with domain-specific learning (C. Quinn, Kinshuk, S. Overmyer)
3. The IPA should create leading questions, naturally motivate the learner, develop a thought logically through questioning, and emphasize learner activity - be Socrates (M. Zenanko)
4. The IPA should encourage the student to move from general to more specific thinking (J. Witherow)
5. The IPA should encourage student to reflect on their learning process. (J. Benda)

Co-worker or Partner
This role was mentioned, however, no specific functions or capabilities were suggested.

1. The IPA should stimulate learners by engaging them (emotionally and otherwise). (D. Graves, T. Fenton-Kerr)
2. The IPA should use animation to motivate learners to use the system. (A. Dufresne)
3. The IPA should remind students of deadlines, etc., to give the impression that someone "cares for their work" (A. Dufresne)
4. The IPA should regulate cognitive load so as to keep the learner interested, but not overwhelmed. (Kinshuk)

Group Facilitator
1. The IPA should enable group members to give their opinions and ideas. (J. Whatley)
2. The IPA should allows all members to have a say. (J. Whatley)
3. The IPA should plan and check group progress in order to stimulate online face-to-face meetings. (J. Whatley)

Finally, J. Moshinskie posted the results of a survey in which a number of desirable features of pedagogical tools were mentioned. The features mentioned were:
1. Templates
2. Branching
3. Feedback
4. Providing Help to the Learner
5. Ability to import graphics
6. Ability to distribute education globally
7. Reporting capabilities
8. Animation, and
9. Test shells

This discussion provided some insight for developers of intelligent pedagogical agents into the capabilities that educators and eventual users of their products might need. In addition, this discussion has resulted in significant reflection on the part of the discussant, and the addition of several functional requirements for his current IPA project environment.