Swarm-based sequencing recommendations in e-learning Colin Tattersall, Bert van den Berg, René van Es, José Janssen, Jocelyn Manderveld, Hub Kurvers, Francis Brouns, Wim Waterink, Catherine Bolman, Pierre Höppener, Rob Koper Onderwijstechnologisch Expertisecentrum

Introduction ­Life Long Learning ­Adult learners who work ­Learning takes mostly place in spare time ­Update knowledge ­Online and Distance Education ­Flexibility ­Full learning programs ­Short programs ­Learners guided by interest ­Shopping behavior

Overview: -Background -Simulation -Results -Questions

Learning Networks ­Facilities for life long learners ­Support seamless, ubiquitous access ­Activity nodes: ­Anything that supports learning ­Course ­Workshop ­Conference ­Web resource ­Providers and learners can create new ANs

Background I: Developmental research

Learning Networks II ­Self-organized, distributed e-learning systems, designed to facilitate lifelong learning. ­Refers to the connection of ­a large number of actors and learning artifacts into an organization, ­as well as to the physical ICT network and facilities that are responsible for the technical connections.

Goal Position Learning Network AN e AN n AN g AN k AN h AN wAN s AN x AN f AN t AN o AN v AN u   Actors

Issues in a learning network: ­Position: Where are you at this moment ­Goal: What do you want to reach ­To do list: Goal – Position ≈ A list of ANs that you still have to complete

Problem: How do we support life-long learners to determine their learning path We used the principles of self-organisation to provide Feedback What feedback?

Principles of Self-organisation

What feedback? Last completedBest Next

Feedback tool ­Navigation help for learners ­Advices best next AN ­Based on the behaviour of other learners ­Applied at an Internet course

Architecture for wayfinding support

Feedback mechanism ABCDE {}13245 A 4251 B2 213 C34 12 D424 5 E1253 E = {ABBCCCCCDDD}

Approach – different viewpoints: 1. Analysis of current situation: Case 1 st year of Psychology 2. Simulating the effects of feedback/stigmergy in a Learning network 3. Feedback tool developed and tested in experiment Hypotheses: 1.Increased efficiency in Experimental group 2.Higher path conversions in Experimental group

Simulation of a Learning network ­Why a simulation ­To save time ­To test techniques prior to tool development ­To make predictions

The underlying model Learner Goal Competencies Motivation Available studytime Activity node Objective & Level Quality Studyload Studying of a AN by Learner Matching error Select best alternative Time learner spends Attain or dropout

De Roma simulation ­The learning network consists of ANs ­Learners have to successfully complete all ANs Various factors: ­With feedback/without feedback ­Amount of ANs ­Amount of learners ­AN quality ­Rate of feedback pursuit

Exemplary outcome ­Without pheromones ­67% successful ­30% dropout ­With pheromones ­76% successful ­22% dropout

Convergence of learning paths ­Without pheromones ­Module-7,Module-8,Module-5,Module-9,Module-3,Module-11,Module-4,Module-2 ­Module-5,Module-7,Module-9,Module-3,Module-11,Module-2,Module-1,Module-6 ­Module-3,Module-5,Module-6,Module-4,Module-11,Module-1,Module-7,Module-10 ­Module-10,Module-11,Module-9,Module-2,Module-4,Module-5,Module-3,Module-6 ­Module-7,Module-5,Module-6,Module-1,Module-2,Module-3,Module-9,Module-4 ­With pheromones ­Module-1,Module-7,Module-5,Module-6,Module-2,Module-3,Module-8,Module-11 ­Module-4,Module-10,Module-7,Module-1,Module-8,Module-5,Module-3,Module-9 ­Module-1,Module-8,Module-4,Module-9,Module-11,Module-3,Module-5,Module-2 ­Module-7,Module-1,Module-8,Module-4,Module-9,Module-5,Module-10,Module-2 ­Module-1,Module-8,Module-4,Module-9,Module-11,Module-3,Module-2,Module-5

Internet course Two Moodle sites ­Internetcursus.ou.nl (control group 505 learners) ­Minicursus.ou.nl (experimental group 507 learners) ­Per group 1 helpdesk ­11 AN’s (modules) ­Beginning internet surfers ­2 hours studyload per AN ­Ended with 5 question test ­Started March 17 ­Ended July 17

Cursusoverzicht

Control group

Experimental group

Content example

Learning paths

Learning Paths

Preliminary results ­Base upon: ­Questionnaire in the course ­Male/FemaleMale/Female ­AgeAge ­Preliminary training/experience ­Computer skills ­Learning paths (logged) ­Questionnaire afterwards

Completion rate of AN’s

Summary of preliminary findings ­Convergence ­Course completion ­Experimental group 95 ­Control group73 ­Non-starters ­Experimental group 114 ­Control group128 ­No show246

Final results I

Period I Final results I Learner progress

Period I Period II Final results I Learner progress

Period IPeriod IIPeriod III Final results I Learner progress

? Period IPeriod IIPeriod IIIPeriod IV Final results I Learner progress

Period IPeriod IIPeriod IIIPeriod IV Final results I Learner progress Significant difference up to interference

Final results II (after correction) Goal attainment Completion of 11 ANs GroupTotal ControlExperimental No ,6% ,8% ,1 Yes ,4% ,2% ,9% Total ,0% Significant difference

Final results III ­Efficiency ­Avarage time needed to complete all 11 modules ­Control group: 38,9 days ­Experimental group: 36,5 days ­Not significant

Roma results - products ­Smiling Faces ­Articles en papers: ­Simulation ­Swarm-based adaptation: wayfinding support for lifelong learners ­'Self Organising Wayfinding Support for Lifelong Learners' published in 'Education and Information Technologies'. ­Swarm-based sequencing recommendations in e-learning ­

back

Male/Female back Female Male unknown