Team-based Instructional Planning Julita Vassileva University of Saskatchewan

1
Team-based Instructional Planning Julita
VassilevaUniversity of Saskatchewan
2
Motivation

LOR

LOR

LOR



LOR
LOR
LOR
3
Motivation

• Heterogeneous repositories of LOs lots of
  materials available
• Metadata standards, semantic web search is
  possible
• But will the learner know how to search to find
  what is best suitable for his/her learning goals?
• Assuming this problem was solved, how to present
  the learner with a coherent learning experience
  and not with a set of fragments?

4
Instructional Planning

• An instructional planner should take the
  requirements
• goals, constraints, learning style, preferred
  teaching style of the student
• pedagogical requirements and strategies
• available LOs that meet these requirements
• and create a sequence that appears coherent,
  smooth, and making sense to the student.
• However, all the knowledge is domain-dependent! ?
  very complex planning (centralized planner will
  be a bottleneck)
• A lot of the knowledge resides with the LOs
• Prerequisite links among LOs, pedagogical
  characteristics
• Stats of learners (with their requirements) and
  their success
• Idea distributed planner!

5
Planning in a Soccer Game
6
Team Analogy





Planning a course Soccer
LO Agents Players
Students agent Coach
Pedagogical agent Rule book / Referee
Student Ball
7
LO Agents

• Self-interested
• Goal to maximize their success statistics
• Secondary goal to maximize number of students
  taught (eventually ).
• Cooperating
• Only through cooperation they can achieve max.
  number of successful teaching sessions

8
Students Agent

• Keeps information about the student (learner
  model)
• Goal, learning style, preferences, teaching style
• History
• Role
• Searching for and selecting LOs (team members)
• Watches the game and interferes when needed (when
  students constraints are violated)

Constraint-satisfaction
9
Pedagogical Agent

• Keeps information about teaching strategies in
  the domain
• Collects statistics about successof LOs with
  different learners
• Collects stats about success of sequences of LOs
• Role
• Monitors the game and ensuresthat the players
  keep to the rules
• Auctions out the ball (the attention of the
  student), accepts bids from team-members (LOs)

10
Algorithm Overview
11
Re-planning

• If student is not doing well in current plan
• Fails a test or sequence of tests or game
  constraints violated (PA interferes)
• E.g. no time left to complete the plan
• Growing evidence about student shows that some
  learner model data was incorrect (SA interferes)
• Different goal
• Different learning style / preferences
• Different background knowledge

12
Proof of Existence Simulation

• Fictitious LOs
• Different levels of granularity, for different
  types of learners, different stats of usage
• Fictitious SAs
• Different knowledge, learning types
• Very simple PAs
• Rules (try to use max stat LOs, match, keep
  time)
• Criteria for success
• Won games (accomplished learner goals and max LOs
  success stats)
• More Distant Future
• Implement in real system