Slides for a talk given at the Institute of Physics Higher Education Group meeting on Concept Inventories and Learning Analytics, Tue 4 April 2017, Open University, UK
Trains and Balloons: An Introduction to Learning Analytics
1. Trains and Balloons:
An Introduction to
Learning Analytics
Doug Clow, Institute of Educational Technology
The Open University
HE Group meeting #IOPCI17
Open University, 4 April 2017
5. Learning analytics …
… the measurement, collection, analysis and
reporting of data about learners and their
contexts, for purposes of understanding and
optimizing learning and the environments in
which it occurs.
solaresearch.org
5
6. Cloud Chamber at the German Electron Synchrotron DESY
Photo public domain: http://commons.wikimedia.org/wiki/File:DESYNebelkammer.jpg
- Erik Duval
http://erikduval.wordpress.com/2012/01/30/learning-
analytics-and-educational-data-mining/
“collecting traces
that learners leave
behind and using
those traces to
improve learning”
44. Image (cc) Darwin Bell http://www.flickr.com/photos/darwinbell/296553221/
“The predictive model is
used as a trigger for
intervention emails to
the learner.”
44
45. 45
Photo (CC)-BY-SA Dudva https://commons.wikimedia.org/wiki/File:Number_4468_Mallard_in_York.jpg
From:
DONOTREPLY@mail.example.com
You are in trouble. The
computer predictive model
gives you a 87.4322% chance
of failing this course.
46. 46
Photo (CC)-BY HiggySTFC https://flic.kr/p/3hp761
Hi Alex
Are you Ok? I noticed
you haven’t logged on
this week, and I know you
struggled with the last
assessment. We can
work through this
together - let’s have a
chat as soon as possible.
Pat.
48. Your journey:
a) Where are you now?
b) Where do you want to get to?
c) How will you get there?
49. What do you want
to do with learning
analytics?
Discuss in your tables
49
50. Your journey:
a) Where are you now?
b) Where do you want to get to?
c) How will you get there?
51. 51
• Review of current practice in UK HE & FE
• Code of practice for learning analytics
• Evidence, Literature review
• Readiness questionnaire (onboarding guide)
• Learning analytics service
https://analytics.jiscinvolve.org/
Effective Learning Analytics
58. Test what you do
• Does it work?
• You’ll have data!
Photo (CC)-BY Kevin Dooley https://www.flickr.com/photos/pagedooley/6613526021/
Towards Evidence-
Based Practice
59. 59
Organisations are built to resist change
• Rock and the river?
• Big lever
Vernal Falls, Yosemite
Photo CC BY Mary PK Burns https://flic.kr/p/6zUzqh
66. What data do we have about learners?
• Demographics
• Previous educational experience
• Grades, scores, achievements, struggles
• Attendance, location, gaze
• Software logs
• Online tracking
• Other online activity (tracking)
• … more every week.
66
Photo (CC)-BY-SA AJ Cann https://www.flickr.com/photos/ajc1/15574010080/
67. What can we do with that data?
• Identify learners who need help
– Simple or predictive
• Trigger interventions
– Via teacher, or direct
• Learn which interventions work
• Build a complete cognitive learning system
• Suggest resources or source of help
– Learners like you found this helpful
– This person might be able to help you
67
Photo (CC)-BY-NC Pulpolux https://www.flickr.com/photos/pulpolux/8735428280
Tour: I will try to show you the main sights, but I am a not unbiased.
MOSTLY WHITE MEN
Many definitions. Slippery.
I like ‘Changed capacity to act.’
Optimising learning!
Big Data in education
More learning online. More sensors making offline, online.
Without interventions: still good stuff: computer science, educational research, business intelligence
But only LA if fed back.
What good teachers have always been doing, but more data, and better techniques.
What are you optimising?
Assessment crucialOptimising out.
Many definitions. Slippery.
I like ‘Changed capacity to act.’
Optimising learning!
Laser cutter
Followed instructions on the sheet
Load the shape file, change, load plastic in to machine, cut shapes. Glue.
Sensors: temperature, sound, light. Outputs: motors, LEDs, display
Program board to control.
Three groups. Input a signal, output a signal.
Train – assess output, learned to use laser cutter.
Balloon – harder. Softer skills. But multimodal data.
Two visions of learning.
Train learning
If we’re making our lives easier, teachers’ lives easier, use the space for better learning
If we’re making our lives easier, teachers’ lives easier, use the space for better learning
Balloon learning
Balloon learning
Balloon learning
Neither is better
Robots are taking our jobs!
There are loads of studies/papers/arguments that robots will take our jobs. Not just assembly line workers.
What is hard to automate?
Soc intel: dishwasher vs PR
Creativity: legal clerk vs biologist
Manipulation: telesales vs surgeon
Balloon view is the future.
Train is the pressure now. Must do it, so do it.
But look up! Make space.
Many definitions. Slippery.
I like ‘Changed capacity to act.’
Optimising learning!
Data Warehouse, pilot projects, VLE stats, dashboards?
Who has it?
Deanery, IT / Computing Service, Registry, Library
Talk to neighbour, one minute each way, I’ll referee.
Everybody wants a dashboard
Almost every Univ s/w product has a dashboard or analytics
Illusion of control and mastery
Vs Making right data visible to people who can do something about it
Dashboard analogy: keep looking out the window. Cruise control vs look at speedo vs appropriate speed
Doesn’t tell you where you should go!
Class 142 cab.
Very simple
Concorde cockpit. Much harder. Engineer’s block to the right.
Don’t use the data just because you have it.
Look out of the window!
Not just the data you’re given
Cohort dispositional analytics.
Building critical self-awareness.
Correlations with success measures, but complex relationship.
Learning power goes down over time in school!
Do not reply. Computer. Will fail. Self-fulfilling.
Have seen worse! At least it’s clear what it’s saying.
Human relationship
You are monitored vs I noticed
M4 motorway – share your data with students
Tell your neighbour. One minute each way, I’ll referee.
Big procurement exercise for an analytics infrastructure, data warehouse plus viz/analytics suite
To running a small exercise in a couple of your lectures (starting small is Ok)
Developing an institutional analytics strategy
To making sure students get their marks on time
Niall Sclater
Niall Sclater
LASI UMich
Interviews with LA experts, and me
We can help!
Talk to your neighbour, two minutes each way.
It could be a tiny thing: see what analytics your VLE can produce already,
It could be big: Develop an internal funding pitch for a comprehensive analytics plan
Make it concrete and specific.
Make the first step do-able today or tomorrow morning.
Idealised
You’re scientists!
Organisations are built to resist change
Solid granite, wears away over time
Do both. Train and balloon.
This is Google’s self-driving car!
If you can automate it, automate it, and use the time for balloon learning.
Computer programming.
M4 is a motorway that goes from London, England, all the way to South West Wales.
Wales is a different country from England. Separate, like Scotland. Roads are managed by a different organisation.
In England, surveillance. Who is looking at me? Why are they looking at me?
In England, surveillance. Who is looking at me? Why are they looking at me?
In Wales, cameras visible online. I can see when a junction is busy.
Now I feel sorry for the person who has to watch all these cameras for traffic jams.
Transparency helps. And Traffic England have now done the same!