An Adaptive Tutor to Promote Learners Skills Acquisition during - - PowerPoint PPT Presentation

An Adaptive Tutor to Promote Learner’s Skills Acquisition during Procedural Learning

Joanna TAOUM

Workshop “eliciting Adaptive Sequences for Learning” ITS, Montréal (Canada), June 2018

CONTEXT

ECA Positive Effect

Student engagement

[Rowe et al., 2007]

Effectiveness of teaching

[Kokane et al., 2014]

Pedagogical Interaction

Communication content

[Kopp et al., 2008]

Intelligent Virtual Environment

[Aylett and Cavazza, 2001]

Actions and Procedures on Technical Systems

Pedagogical Scenarios

[Saunier et al., 2016]

1/15

STEVE [Johnson, 1998] Windmill Structure and VE Representation

taoum@enib.fr

PROPOSITION

LEARNER’S EVOLUTION

Each learner evolves differently

OBJECTIVE

Improving ECA with an adaptive tutor behavior, that is able to adapt the execution

• f a pedagogical scenario according to the learning performance of a learner

WHY?

To enhance learner’s learning performance

taoum@enib.fr

2/15

▰

Formalization of the intelligent virtual environment

INTELLIGENT TUTORING SYSTEM

Student Model Tutoring Model Domain Model Interface User- Learner

ITS four component architecture

[Nkambou et al., 2010]

taoum@enib.fr

3/15

▰

Domain expert knowledge

INTELLIGENT TUTORING SYSTEM

Student Model Tutoring Model Domain Model Interface User- Learner

ITS four component architecture

[Nkambou et al., 2010]

taoum@enib.fr

4/15

DOMAIN MODEL: MASCARET

▰

Domain model is represented by MASCARET [Chevaillier et al., 2001]

§

Virtual Reality meta-model based on UML for designing the semantic of the IVE

§

Domain and pedagogical concepts are explicit

Agent Knowledge Base (MASCARET) Environment Structure Entities’ Behavior Users & Agents activities and interaction

taoum@enib.fr

4/15

▰

Natural realization of pedagogical actions

INTELLIGENT TUTORING SYSTEM

Student Model Tutoring Model Domain Model Interface User- Learner

ITS four component architecture

[Nkambou et al., 2010]

taoum@enib.fr

Embodied Conversational Agent (ECA)

5/15

▰

Interface is represented by ECA SAIBA Compliant

IINTERFACE

Multi-modal Signals Pointing Smiling Frowning Tutor (ECA) Learner (Human user)

GRETA

[Niewiadomski et al., 2009]

MARC

[Courgeon, 2011]

VHT

[Gratch et al., 2013]

PRIMITIVE ACTIONS

Verbal communication:giving an information, … Non-verbal communication: multimodal signals, … Actions on the environment: manipulating an object, … Navigation: observing, moving, …

PEDAGOGICAL ACTIONS

On the virtual environment: highlighting an object, … On the user’s interaction: changing viewpoint, … On the structure of the system:describing the structure, …

taoum@enib.fr

In Mascaret, the tutor and the learner are considered as Agents

5/15

INTERFACE

taoum@enib.fr

5/15

▰

Adaptation of the pedagogical scenario based on the learner’s knowledge

INTELLIGENT TUTORING SYSTEM

Student Model Tutoring Model Domain Model Interface User- Learner

ITS four component architecture

[Nkambou et al., 2010]

taoum@enib.fr

Instructions encoding in memories

6/15

▰

General theoretical framework of human memory [Atkinson and Shiffrin, 1968]

▰

Main contribution:

§

Formalizing the content of memory

§

Implementing the execution flow of memories

STUDENT MODEL Sensory Memory Working Memory Long-term Memory

Encoding Storing Incoming Retrieving Rehearsal Information

taoum@enib.fr

6/15

Example

Open the tube of Neoplastin Encoding Storing Retrieval Visual Information Auditory Information

7 chunks

taoum@enib.fr

Auditive register Visual register

7/15

Example

Open the tube of Neoplastin Encoding Storing Retrieval Visual Information Auditory Information

7 chunks

taoum@enib.fr

Auditive register Visual register

7/15

Example

Open the tube of Neoplastin Encoding Storing Retrieval Visual Information Auditory Information

7 chunks

taoum@enib.fr

Auditive register Visual register

7/15

Example

Open the tube of Neoplastin Encoding Storing Retrieval Visual Information Auditory Information

7 chunks

taoum@enib.fr

Auditive register Visual register

7/15

Example

7 chunks

Encoding Storing Retrieval Visual Information Auditory Information

taoum@enib.fr

Auditive register Visual register

7/15

▰

Adaptive tutor behavior based on Domain and Student Model

INTELLIGENT TUTORING SYSTEM

Student Model Tutoring Model Domain Model Interface User- Learner

ITS four component architecture

[Nkambou et al., 2010]

taoum@enib.fr

Adaptive Tutor Behavior

8/15

TUTOR BEHAVIOR

Adaptive Behavior Inputs unexpected expected Action realization Pedagogical scenario Remediation (Pedagogical action) Domain Model Working Memory Long-term Memory

taoum@enib.fr

8/15

Long-term Memory Transfer Working Memory Modification

Based on cognitive psychology hypothesis. Behavior to be defined using UML activity diagrams.

TUTOR BEHAVIOR

taoum@enib.fr

8/15

Evaluate the impact of our adaptive virtual tutor on the learner’s learning performance

Objective

Two groups of 22 participants each, 14% female, mean age 22,7

Participants

Procedural learning for blood analysis in a virtual environment laboratory

Application

EVALUATION

taoum@enib.fr

9/15 Objective Performance Measure

Time, errors and Help request

Same first trial for both groups “Systematic Assistance”

Experimental Protocol

Group I Group II

Non-Adaptive Tutor No pedagogical actions One type of help request Feedback when incorrect action (Goal + Action + Object) Adaptive Tutor Pedagogical actions based on the learner’s memories content Different types of help request Feedback when incorrect object and/or action and inaction

EVALUATION

taoum@enib.fr 10/15

MEAN OF REALIZATION TIME

Significant results only for the second trial:

Learner’s in adaptive condition takes less time to realize the procedure

50 100 150 200 250 300 350 2 3 4 5

Realization time (sec) Trial’s number

RESULTS

Statistical analysis applied starting trial 2 taoum@enib.fr 11/15

MEAN OF HELP REQUEST

Significant results for the second trial and for all trials:

In general, adaptive condition participants require less help

taoum@enib.fr 12/15

1 2 3 4 5 6 7 8 9 10 2 3 4 5

Number of help request Trial’s number

RESULTS

Statistical analysis applied starting trial 2

taoum@enib.fr 13/15

MEAN OF INCORRECT ACTIONS

Significant results for the second trial and for all trials:

In general, adaptive condition participants make fewer incorrect actions.

1 2 3 4 5 6 7 2 3 4 5

Number of incorrect actions Trial’s number

RESULTS

Statistical analysis applied starting trial 2

“

Generally, interaction with an adaptive virtual tutor improves learner’s performance

Specially, during the first executions of the procedure

DISCUSSION

taoum@enib.fr 14/15

CONCLUSION AND PERSPECTIVES

The proposed model can also be applied in use cases of cognitive impairments, Alzheimer disease: modification of the memory flow.

We propose an adaptive virtual tutor based on Intelligent Tutoring Systems and Cognitive Psychology (Theory of Human Memory).

taoum@enib.fr 15/15

THANK YOU!

Any questions? You can contact me at taoum@enib.fr

An Adaptive Tutor to Promote Learner’s Skills Acquisition during Procedural Learning

Joanna TAOUM

Workshop “eliciting Adaptive Sequences for Learning” ITS, Montréal (Canada), June 2018