The Role of Learning Analytics

advertisement
The Role of Learning Analytics:
A personal Journey…
5/22/2013
Oded Meyer
Department of Mathematics and Statistics
Georgetown University
Some background
about myself…..
Learning Analytics
Data driven approach for the purpose
of understanding and improving
learning and the and the environment
in which it occurs.
Carnegie Mellon Open Learning Initiative
(OLI)
Scientifically-based
online learning
environments based
on the integration of
technology and the
science of learning
with teaching. OLI is
designed to
simultaneously
improve learning and
facilitate learning
research.
The OLI Statistics Course
Educational Mission of Funder (2002)
(The William and Flora Hewlett Foundation)
Provide open access to high quality post-secondary
education and educational materials to those who
otherwise would be excluded due to:
– Geographical constraints
– Financial difficulties
– Social barriers
To meet this goal:
– A complete stand-alone web-based introductory
statistics course.
– openly and freely available to individual learners online.
Important form of analytics:
The Science of Learning
General:
• Make the structure and “big picture” salient
Learning Science Principles - continued
• Immediate and targeted feedback  students achieve
desired level of performance faster.
Science of Learning - continued
Discipline specific principles:
• Hands-on activities
• Use real data
• De-emphasize calculations.
Start Analytics Early in the Design
process
After one module :
• Qualitative feedback
• Observe students
• Talk-allowed protocols
Instructor feedback loop analytics
• Feedback to the instructor about students’
learning  Learning Dashboard
- Presents the instructor with a measure of student
learning for each learning objective.
- More detailed information:
• Class’s learning of sub-objectives
• Learning of individual students
• Common misconceptions
Learning Dashboard Team led by Dr. Marsha Lovett
Instructor feedback loop analytics
Key:
• Clearly defined learning objectives.
• Tying each activity to a learning objective.
Benefits:
• Can reveal misconceptions
• Impacts how your spend class time
• Can reveal “expert blind spots”
Students who
achieved proficiency
in finding median did
poorly in the following
assessment item tied
to the same sub-skill:
Larger Scale Analytics:
Assessing the Effectiveness
of the Course When used in
the Blended (Hybrid) Mode
Larger Scale Analytics:
Assessing the Effectiveness of the Course
The Hewlett Foundation’s “Accelerate
Learning Challenge”:
Can students using the OLI course in the blended
mode learn the same material as they would in
the traditional course in shorter time and still
have equal or better learning gains.
Three accelerated Studies
#1
Small class, expert instructor (2007)
#2
Replication with larger class (2009)
– With retention follow-up 4+ months later
#3
Replication with new instructor (2010)
– Experienced statistics instructor
– New to OLI Statistics course and hybrid mode
Study 1: Method
~180 students
enrolled
68 volunteers for
special section
24 students,
adaptive/
accelerated
condition
44 students,
traditional
control
condition
Adaptive/Accelerated vs. Traditional
Two 50-minute classes/wk
Eight weeks of instruction
Homework: complete OLI
activities on a schedule
Tests: Three in-class
exams, final exam, and
CAOS test
<
<
?
Four 50-minute classes/wk
=
Tests: Three in-class
exams, final exam, and
CAOS test
Fifteen weeks of instruction
Homework: read textbook &
complete problem sets
Same content but different kind of instruction
Dependent Measure
CAOS = Comprehensive Assessment of Outcomes in a First
Statistics course (delMas, Garfield, Ooms, Chance, 2006)
– Forty multiple-choice items measuring students’
“conceptual understanding of important statistical ideas”
– Content validity – positive evaluation by 18 content
experts
– Reliability – high internal consistency
– Aligned with content of course (both sections)
– Administered as a pre/posttest
Study 1: CAOS Test Results
Chance
Adaptive/Accelerated group gained more (18% vs
3%)pre/post on CAOS than did Traditional Control, p <
.01.
These analytics got the attention of education
leaders in the U.S. who are facing the “cost
disease” in higher ed.
William Bowen (former President of Princeton)
replicated our study in the “Interactive Learning
Online at Public Universities” study.
The study further indicated that blended
learning offers the potential of more economical
and rapid pathways to mastery.
Analytics on Students’ Learning Habits
•
•
•
•
Students in both groups recruited to complete time-logs
Self-report for both groups
Analogous point in the course (2/3 through)
Six consecutive days: Wednesday - Monday
Study 1: Time Spent Outside of Class
No significant difference between groups in the time
students spent on Statistics outside of class
Study 2: Replication & Extension
• Same method, same procedure, same instructor
• Larger class (52 students in Adaptive /
Accelerated)
• Follow-up retention study conducted 4+ months
later
Study 2: CAOS Test Results
Chance
Adaptive/Accelerated group gained more pre/post on
CAOS than did Traditional Control, p < .01.
Using Analytics to Assess Retention
Follow-up Begins
Trad’l Ends
Adapt/Acc Ends
Jan Feb Mar Apr May Jun Jul Aug Sep Oct
Adapt/Acc Delay (13 Students)
Trad Delay (14 Students)
Study 2, Retention: Re-taking CAOS
Chance
At 6-month delay, Adaptive/Accelerated group scored
higher on CAOS than Traditional Control, p < .01.
Study 3: Further Replication & Extension
• Same method, same procedure
• New instructor
• Not involved in development of OLI course
• New to OLI statistics and hybrid teaching mode
• Instructor held constant for both Adapt/Acc and
Control conditions
• Larger class (40 students in Adaptive / Accelerated)
Study 3: CAOS Test Results
70
60
50
40
30
Pretest
20
Posttest
10
Chance
0
Adapt/Acc
Trad Control
Adaptive/Accelerated group gained more pre/post on
CAOS than did Traditional Control, p < .01.
Current and Future Analytics
• Continued “gap analysis”.
• Better alignment between learning objectives/subobjectives and activities.
• Student Dashboard  provide learners with insight into
their own learning habits and can give recommendations
for improvement.
• Learning-facing analytics  allows learners to compare
their own performance against an anonymous summary
if their course peers
Lessons Learned…
1. Pedagogy must drive technology and not the
other way.
2. Developing online materials is a collaborative
effort.
3. Developing online materials is an iterative
process.
4. Steep learning curve.
5. Growth as a teacher.
“Improvement in postsecondary education will
require converting teaching
from a ‘solo sport’ to a
community-based research
activity”
Herbert Simon,
Last Lecture Series, Carnegie Mellon, 1998
Download