Learning Analytics: On the Way to Smart Education Dr. Griff Richards Athabasca University Moscow, 08 October 2012.

Learning Analytics: On the Way to Smart Education

Dr. Griff RichardsAthabasca University

Moscow, 08 October 2012

60oN

49oN

Distance LearningIn Canada

- Is growing, especially in K-12 public schools

-15% of 2008 high school graduates in BC took at least one on-line course

-Increase in “Blended Learning”, online activities for F2F courses.

-Increasing concern for course quality and student retention

Athabasca, Alberta

Learning Engagement

The more a learner interacts with the content and with their peers about the content, the more they are likely to internalize it, and remember it...

(Richard Snow, 1980)

Learning engagement promotes student retention and academic success.

(George Kuh,2001) • Engagement is “action based on internal

motivation”• Can not directly measure engagement• Must look for external traces, actions that

indicate interest, and involvement (NSSE)

Learning Engagement

CAUTION: Engagement

• Arum & Roska (2011)• Students who studied

alone had higher marks than students studying in groups.

• Engagement that is not task-focused is unlikely to improve learning.

Interaction Equivalency Theory(Garrison & Anderson,1995)

The quality of the interaction is more important than the source

CourseContent

Instructor

Other learners

4 places to improve learning

CourseContent

Instructor

Other learners

1. Clearer designs

2. CollaborativeLearning activities3. Less lecturing, more

mentoring

4. More study skills

We have insufficient data about which instructional strategies actually work best.

Hybrid or Blended Learning?

Face toFace On-Line

The use of online technologies to augment face to face delivery.-> replaces some F2F “face time” (forums save classroom space)-> uses LMS to track assessments and assignments-> uses technology in class e.g. pop quizzes, collaboration tools

Analytic Measures

Engagement is inferred from activity.• Student interaction with Online

systems leaves an “exhaust trail” of their activity

• Learning Analytics is not a statistical sample, it is all the data for all learners

• Questions: What patterns to look for? How to interpret them?

• LMS interaction data e.g. Moodle discussion

• Extract linked list• Plot interactions

in star map

Example: Snapp

SNAPP Visualization

SNAPP shot of Conference

Students “engaged”

SNAPP shot of Conference

Individuals with lower engagement, 3 or less messages

Does Activity = Engagement ?

Beer (2010) plotted total LMS interactions with academic grades. Does activity = engagement?

Is Blackboard more engaging than Moodle?

Limitations: Activity Outside LMS

• As learners become engaged, they move to more engaging channels: email, Elluminate, Skype

• This activity is not tracked by LMS.

• No data available.

LMS

Interpretation of Analytics

• Data patterns require investigation• Quantitative data requires interpretation

--> make and test hypotheses--> create useful models

• When we measure something we risk changing it.– e.g. If learners know we count hits they may

make meaningless hits to fool the system.

Analytics for Learners!

• The same analytics should be able to provide easy to understand information dashboards for students.

Analytics for Learners

SIGNALSArnold (2010) inserted a traffic light on each student’s course page to provide guidance on course success. A/B C D/F

Dashboard for Faculty

Arnold (2010) reported 14 % shift from D’s to C’s & B’s. Same number of withdrawals, but W’s occurred earlier before damaging student grade point averages.12% increase in students requesting assistance. N=220

Mesmots a dashboard for learners

(Richards & Sehboub, 2008)

Data for my class

My webquest data

How to start: Analytics for online & blended learning?

• Measuring something is the first step• “Better to measure something than to

measure nothing” (Scrivens)

• Need more data than just page hits. We also need to ask learners about their experience, what worked, what needs improvement.

Dynamic Evaluation Model (Richards & Devries,2011)

» Preparation Conduct Reflection

• Design

• Facilitation

• Learning

Analytics at the activity level

Dynamic Evaluation Model

» Preparation Conduct Reflection

• Design

• Facilitation

• Learning

Timely feedback enables quick fixes

If Analytics, Then What?

• If analytics show students are failing, is there a moral obligation to help them?

• If analytics show a course has weaknesses, is there a business obligation to fix it?

• If analytics reveal weak instruction, who is responsible to intervene?

• If analytics are inaccurate, who fixes them?• What about privacy & ethics? Who owns the

data? Who has the right to see it?

The Analytics Box?

•jo

anna

pary

pins

ki.c

om

Learning Analytics: On the Way to Smart(er) Education

Dr. Griff RichardsAthabasca [email protected], 08 October 2012