Keynote Address, International Conference of the Learning Sciences, London Festival of Learning
Transitioning Education’s Knowledge Infrastructure:
Shaping Design or Shouting from the Touchline?
Abstract: Bit by bit, a data-intensive substrate for education is being designed, plumbed in and switched on, powered by digital data from an expanding sensor array, data science and artificial intelligence. The configurations of educational institutions, technologies, scientific practices, ethics policies and companies can be usefully framed as the emergence of a new “knowledge infrastructure” (Paul Edwards).
The idea that we may be transitioning into significantly new ways of knowing – about learning and learners – is both exciting and daunting, because new knowledge infrastructures redefine roles and redistribute power, raising many important questions. For instance, assuming that we want to shape this infrastructure, how do we engage with the teams designing the platforms our schools and universities may be using next year? Who owns the data and algorithms, and in what senses can an analytics/AI-powered learning system be ‘accountable’? How do we empower all stakeholders to engage in the design process? Since digital infrastructure fades quickly into the background, how can researchers, educators and learners engage with it mindfully? If we want to work in “Pasteur’s Quadrant” (Donald Stokes), we must go beyond learning analytics that answer research questions, to deliver valued services to frontline educational users: but how are universities accelerating the analytics innovation to infrastructure transition?
Wrestling with these questions, the learning analytics community has evolved since its first international conference in 2011, at the intersection of learning and data science, and an explicit concern with those human factors, at many scales, that make or break the design and adoption of new educational tools. We are forging open source platforms, links with commercial providers, and collaborations with the diverse disciplines that feed into educational data science. In the context of ICLS, our dialogue with the learning sciences must continue to deepen to ensure that together we influence this knowledge infrastructure to advance the interests of all stakeholders, including learners, educators, researchers and leaders.
Speaking from the perspective of leading an institutional analytics innovation centre, I hope that our experiences designing code, competencies and culture for learning analytics sheds helpful light on these questions.
1. Transitioning Education’s
Knowledge Infrastructure
Shaping Design or Shouting from the Touchline?
Simon Buckingham Shum
@sbuckshum • http://utscic.edu.au
Keynote Address, International Conference of the Learning Sciences
London Festival of Learning, June 2018
5. 5
infrastructure
amid justified public concern about data/algorithm ethics,
and academic concerns about computational methods,
how do we design
WE TRUST?
16. 16
Learning Analytics is therefore building bridges with…
Many disciplines…
LAK conference keynote speakers include:
Learning Sciences Paul Kirschner
David Williamson Shaffer
Art Graesser
Psychometrics Robert Mislevy
Information
Visualisation
Katy Börner
Cristina Conati
IT Ethics Mireille Hildebrandt
Critical studies of
technology
Neil Selwyn
Writing analytics Danielle McNamara
17. 17
Learning Analytics is therefore building bridges with…
LAK conference and LA Summer Institutes offer:
Industry panels to explore the opportunities and tensions
with academia
Industry researchers on the Doctoral Consortium
Commercial sponsors and exhibitors
Collaborative projects with industry
Many disciplines…
LAK conference keynote speakers include:
LMS, Analytics Vendors & Publishers…
Learning Sciences Paul Kirschner
David Williamson Shaffer
Art Graesser
Psychometrics Robert Mislevy
Information
Visualisation
Katy Börner
Cristina Conati
IT Ethics Mireille Hildebrandt
Critical studies of
technology
Neil Selwyn
Writing analytics Danielle McNamara
18. 18
Learning
learning sciences
educational research
teaching practice
curriculum design
learning design
pedagogy
assessment…
Analytics
data
statistics
classification
machine learning
text processing
visualisation
predictive models…
…not a straightforward dialogue
22. Who are Learning Analytics for?
22
Learning Sciences researchers!
1. Theory
2. Experiment
3. Simulation
4. Data-Intensive Science
http://FourthParadigm.org
23. Educational eResearch infrastructure
Education meets Cyberinfrastructure, eScience, eSocialScience, Grid computing, etc…
23
Bill Cope & M ary Kalantzis (2016). Big Data Com es to School: Im plications for Learning, Assessm ent, and Research. AERA Open, 2, (2): April 1, 2016. Open
Access: https://doi.org/10.1177/2332858416641907
Lina M arkauskaite (2016), Digital M edia, Technologies and Scholarship: Som e Shapes of eResearch in Educational Inquiry. The Australian Educational
Researcher, 37, (4), pp.79-101. Open Access: https://link.springer.com /content/pdf/10.1007%2FBF03216938.pdf
Educational research methods could change radically when:
• “sample size” is less relevant: N=All
• statistically sig. patterns are easy to find in such big data
• qualitative textual coding may be automated
• social ties can be tracked at scale
• etc…
24. Who are Learning Analytics for?
24
“The new possibility is that educators and learners — the
stakeholders who constitute the learning system studied for so long
by researchers — are for the first time able to see their
own processes and progress rendered in ways that until
now were the preserve of researchers outside the system.”
Knight, S. and Buckingham Shum , S. (2017). Theory & Learning Analytics. H andbook of Learning Analytics,
Society for Learning Analytics Research, Chap. 1, pp.17-22. http://doi.org/10.18608/hla17
Learners! Educators!
25. Who are Learning Analytics for?
25
“Data gathering, analysis, interpretation and even intervention
(in the case of adaptive software) is no longer the preserve of the
researcher, but shifts to embedded sociotechnical
educational infrastructure.”
Learners! Educators!
Knight, S. and Buckingham Shum , S. (2017). Theory & Learning Analytics. H andbook of Learning Analytics,
Society for Learning Analytics Research, Chap. 1, pp.17-22. http://doi.org/10.18608/hla17
26. 26
But… for many, we’re suspect, wielding
data, analytics and AI in ways that may
oppress rather than empower
tectonic shifts under way
in the educational landscape…
27. Justified concerns around privacy, surveillance and
data ethics are redefining the context for our work
27
28. Growing public literacy around the ethics
of data / algorithms / AI is to be welcomed
28
For more see
http://datasociety.net
29. Critical academic commentary on the
datafication of education (Ben Williamson)
29
“Datafication” at all levels of the educational
system, from government statistics to biometrics
Concern about the ownership of data and
analytics platforms by commercial entities
Worried that students will have opportunities
closed down rather than opened up by
algorithms
And much more… Justified?
30. Critical academic commentary on
Learning Analytics (Neil Selwyn, LAK18 keynote)
30
The Promises and Problems
of ‘Learning Analytics’
https://w w w .youtube.com /w atch?v=rsUx19_Vf0Q
https://latte-analytics.sydney.edu.au/keynotes
31. Classification schemes and metrics are
suspect, with good reason…
31
ideas now being
popularised…
incisive STS
scholarship into
classification
schemes…
32. Du Gay, P. and Pryke, M. (2002) Cultural Economy: Cultural Analysis and Commercial Life. Sage, London. pp. 12-13
“accounting tools...do not simply aid
the measurement of economic activity,
they shape the reality they
measure”
32
33. 33
A rapidly changing educational data/analytics ecosystem…
Venture Capitalists
Philanthropic Foundations
Publishers
as analytics
providers
Pearson
McGraw Hill
Wiley
etc.
Learning
Platform
Services
Blackboard
Canvas
D2L
Facebook
etc.
Adaptive/
Learning
Analytics
Services
SmartSparrow
Knewton
Unizen
CogTools
etc.
Data
Protection
Laws
GDPR National
privacy lawsetc.
Govnt. &
inter-
national
datasets
UK HESA Data Futures
OECD PISA
UNESCO Inst. for Statistics
US Institute for HE Practice
etc.
Learning Analytics
H um an
Factors
34. 34
A rapidly changing educational data/analytics ecosystem…
Learning Analytics
H um an
Factors
Publishers
as analytics
providers
Pearson
McGraw Hill
Wiley
etc.
Learning
Platform
Services
Blackboard
Canvas
D2L
Facebook
etc.
Adaptive/
Learning
Analytics
Services
SmartSparrow
Knewton
Unizen
CogTools
etc.
Data
Protection
Laws
GDPR National
privacy lawsetc.
Govnt. &
inter-
national
Statistics
UK HESA Data Futures
OECD PISA
UNESCO Inst. for Statistics
US Institute for HE Practice
etc.
Venture Capitalists
Philanthropic Foundations
36. “Knowledge Infrastructures” (Paul Edwards)
36http://knowledgeinfrastructures.orghttps://mitpress.mit.edu/books/vast-machine
Knowledge Infrastructures:
Intellectual Frameworks
and Research Challenges
A Vast Machine: Computer Models,
Climate Data, and the Politics of
Global Warming
37. “Knowledge Infrastructures”
37
“robust networks of people, artifacts, and
institutions that generate, share, and
maintain specific knowledge about the
human and natural worlds.”
Routine, well-functioning knowledge systems include the world weather forecast
infrastructure, the Centers for Disease Control, or the Intergovernmental Panel on Climate
Change — individuals, organizations, routines, shared norms, and practices.
Paul N . Edw ards, , Steven J. Jackson, M elissa K. Chalm ers, Geoffrey C. Bow ker, Christine L. Borgm an, David Ribes, M att Burton, Scout Calvert (2013). Know ledge
Infrastructures: Intellectual Fram ew orks and Research Challenges. Report from N SF/Sloan Fndn. W orkshop, M ichigan, M ay 2012
38. “Knowledge Infrastructures”
38
“Infrastructures are not systems, in the sense of fully
coherent, deliberately engineered, end-to-end processes.
…ecologies or complex adaptive systems[…]
made to interoperate by means of standards, socket
layers, social practices, norms, and individual
behaviors.”
Paul N . Edw ards, , Steven J. Jackson, M elissa K. Chalm ers, Geoffrey C. Bow ker, Christine L. Borgm an, David Ribes, M att Burton, Scout Calvert (2013). Know ledge
Infrastructures: Intellectual Fram ew orks and Research Challenges. Report from N SF/Sloan Fndn. W orkshop, M ichigan, M ay 2012
39. “Knowledge Infrastructures”
39
“Infrastructures are not systems, in the sense of fully
coherent, deliberately engineered, end-to-end processes.
…ecologies or complex adaptive systems[…]
made to interoperate by means of standards, socket
layers, social practices, norms, and individual
behaviors.” I think we can see the
educational ecosystem
herePaul N . Edw ards, , Steven J. Jackson, M elissa K. Chalm ers, Geoffrey C. Bow ker, Christine L. Borgm an, David Ribes, M att Burton, Scout Calvert (2013). Know ledge
Infrastructures: Intellectual Fram ew orks and Research Challenges. Report from N SF/Sloan Fndn. W orkshop, M ichigan, M ay 2012
40. Knowledge Infrastructure concepts
40
Models, models, models…
“Everything we know about the world’s climate — past,
present, and future — we know through models.” (p.xiv)
“I’m not talking about the difference between “raw” and
“cooked” data. I mean this literally. Today, no collection of
signals or observations […] becomes global in time and space
without first passing through a series of data models.” (p.xiii)
Paul Edw ards (2010). A Vast M achine: Com puter M odels, Clim ate Data, and the Politics of Global W arm ing. M IT Press
41. Knowledge Infrastructure concepts
41
Models, models, models…
“Everything we know about the world’s climate — past,
present, and future — we know through models.” (p.xiv)
Today, no collection of signals or observations […] becomes
global in time and space without first passing through a series
of data models.” (p.xiii)
Machines ‘see’
learners only through
models
“Raw data is an
oxymoron”
(Geof Bowker)
Paul Edw ards (2010). A Vast M achine: Com puter M odels, Clim ate Data, and the Politics of Global W arm ing. M IT Press
42. Knowledge Infrastructure concepts
42
infrastructural inversion
“The climate knowledge infrastructure never disappears
from view, because it functions by infrastructural inversion:
continual self-interrogation, examining and reexamining its
own past. The black box of climate history is never closed.”
Paul Edw ards (2010). A Vast M achine: Com puter M odels, Clim ate Data, and the Politics of Global W arm ing. M IT Press
43. Knowledge Infrastructure concepts
43
infrastructural inversion
“The climate knowledge infrastructure never disappears
from view, because it functions by infrastructural inversion:
continual self-interrogation, examining and reexamining its
own past. The black box of climate history is never closed.”
We must keep lifting the lid
on learning analytics
infrastructures
We must equip learners and
educators to engage
critically with such tools
Paul Edw ards (2010). A Vast M achine: Com puter M odels, Clim ate Data, and the Politics of Global W arm ing. M IT Press
44. Knowledge Infrastructure concepts
44
metadata friction
“People long ago observed climate and weather for their own
reasons, within the knowledge frameworks of their times.
You would like to use what they observed — not as they used it,
but in new ways, with more precise, more powerful tools.
[…]
So you dig into the history of data. You fight metadata friction, the
difficulty of recovering contextual knowledge about old records.”
(p.xvii)
Paul Edw ards (2010). A Vast M achine: Com puter M odels, Clim ate Data, and the Politics of Global W arm ing. M IT Press
45. metadata friction
“People long ago observed climate and weather for their own
reasons, within the knowledge frameworks of their times.
You would like to use what they observed — not as they used it,
but in new ways, with more precise, more powerful tools.
[…]
So you dig into the history of data. You fight metadata friction, the
difficulty of recovering contextual knowledge about old records.”
(p.xvii)
Knowledge Infrastructure concepts
45
cf. Reanalysis of educational data
(your own and others’) using
computational methods
Paul Edw ards (2010). A Vast M achine: Com puter M odels, Clim ate Data, and the Politics of Global W arm ing. M IT Press
46. Epistemic Infrastructure taxonomy for professional knowledge
(Markauskaite & Goodyear)
Partic contributions at the “Micro-KI” level: how professionals construct their EI
46
Markauskaite, L. & Goodyear, P.
(2017). Epistemic Fluency and
Professional Education:
Innovation, Knowledgeable
Action and Actionable Knowledge
(Springer, 2017), p.376
47. 47
Arguably, education is in transition
to a new KI
The seams of the old KI
are under stress
Rapid tech. change stresses the systems
with more inertia. How can we manage this?
48. What can we expect to see when a KI is in transition?
48
“social norms, relationships, and ways of thinking, acting,
and working” are impacted
“…when they change, authority, influence, and power are
redistributed.”
“new kinds of knowledge work and workers displace old
ones”
“increased access for some may mean reduced access for
others”
49. KI helps explain contemporary concerns…
49
outsourcing of student feedback to machines
alarm over the political agendas driving the collection of
educational data
concern over the commercial drivers/owners behind data
worries that use of analytics/AI = dated pedagogy
50. KI helps explain our current immaturity…
50
poor interoperability between platforms
early products with simplistic dashboards that don’t count what
really counts in learning
growing recognition of the data illiteracy in educational
institutions
universities discovering they have limited control over how they
can access their students’ data from cloud platforms
51. KI helps explain our current immaturity…
51
use of machine learning with no awareness of data and
algorithmic bias
poor grounding of learning analytics in theories of learning
an early fascination (borrowed from business) with predicting
student failure (assumes the past can and should predict the
future)
excessive weight placed on computational performance with less
attention to educational outcomes
52. How to respond as researchers?
52
“social scientists cannot remain simple bystanders
or critics of the current transformations, which will
not be reversed;
“instead, we need research practices that can help
innovate, rethink, and rebuild.”
53. How to respond as researchers?
53
“treat it as a design opportunity, create a cadre at the
interface between scientists and software, and use
participatory design techniques
So what does that look like?
56. Donald Stokes: Pasteur’s Quadrant
• 56
• Edison: invent commercially
viable electric lighting
• Early, pre-theoretical
taxonomising of phenomena
• Bohr: atomic structure;
quantum theory
• Pasteur: understand and control
bacteria in order to prevent disease
• Manhattan Project: atomic bomb
• John Maynard Keynes: economics
57. Locating Learning Sciences & Learning Analytics
57
Learning Sciences
research into the
foundational constructs
and processes
underpinning learning
Data Science/AI research
into new approaches to
machine learning,
algorithm optimisation,
mitigating bias, etc.
BOHR’s Quadrant PASTEUR’s Quadrant
EDISON’s Quadrant Learning Analytics to improve outcomes in
specific contexts (may or may not draw on current theory,
but doesn’t feed back to foundational research concepts)
Learning Sciences:
research-based
educational intervention in
specific contexts, reflect on
implications for theory,
establish sustainable
practices
(e.g. DBR; DBIR; RPPs;
Improvement Science;
Collaborative Data-intensive
Improvement)
58. Locating Learning Sciences & Learning Analytics
58
Learning Sciences
research into the
foundational constructs
and processes
underpinning learning
Data Science/AI research
into new approaches to
machine learning,
algorithm optimisation,
mitigating bias, etc.
BOHR’s Quadrant PASTEUR’s Quadrant
EDISON’s Quadrant Learning Analytics to improve outcomes in
specific contexts (may or may not draw on current theory,
but doesn’t feed back to foundational research concepts)
Learning Sciences:
research-based
educational intervention in
specific contexts, reflect on
implications for theory,
establish sustainable
practices
(e.g. DBR; DBIR; RPPs;
Improvement Science;
Collaborative Data-intensive
Improvement)
59. Locating Learning Sciences & Learning Analytics
59
Learning Sciences
research into the
foundational constructs
and processes
underpinning learning
Data Science/AI research
into new approaches to
machine learning,
algorithm optimisation,
mitigating bias, etc.
BOHR’s Quadrant PASTEUR’s Quadrant
EDISON’s Quadrant Learning Analytics to improve outcomes in
specific contexts (may or may not draw on current theory,
but doesn’t feed back to foundational research concepts)
Learning Sciences:
research-based
educational intervention in
specific contexts, reflect on
implications for theory,
establish sustainable
practices
(e.g. DBR; DBIR; RPPs;
Improvement Science;
Collaborative Data-intensive
Improvement)
Learning Analytics: design and deploy
analytics that demonstrate how
theory can inspire models,
algorithms, code, user experiences,
teaching practices, and ultimately,
learning.
The ability to formally model
theoretical concepts, and shape
learning outcomes, advances theories
(as in other fields)
60. The analytics innovator’s dilemma…
60
PASTEUR’s Quadrant — USE inspired foundational research
The innovation gulf facing learning analytics (much ed-tech) research:
1. Researchers develop novel forms of student-facing
analytics, but…
2. USE-inspired research requires USERS.
There’ll be few users without robust infrastructure
3. So, how to create a KI that accelerates the transition
from analytics innovations to embedded infrastructure?
61. Hybrid analytics innovation + service centres
61
University of Technology Sydney
Connected Intelligence Centre
University of Michigan
Digital Innovation Greenhouse
EDUCAUSE Review, Mar/Apr 2018
https://er.educause.edu/articles/2018/3/architecting-for-learning-analytics-innovating-for-sustainable-impact
63. CIC skillset
Board Room
VC/DVCs/Deans/Directors
Common Room
Academic staff
Server Room
IT Division
Interpersonal skills
+
Education, Learning Design, Interface
Design, Programming, Web Development,
Text Analytics, Machine Learning,
Statistics, Visualisation, Decision-Support,
Sensemaking, Creativity & Risk,
Participatory Design
64. Advantages that this org structure brings
Operating within the DVC’s Office
enables close coupling with student
services and teaching innovation
Baseline funding provides invaluable
stability for planning projects and staff
Reporting directly to a DVC, and talking
directly to other operational directors,
gets stuff done
Operating outside a faculty provides
agility for decision-making, and
helpful neutrality
65. 65
2 design lenses:
Analytic Accountability Cycle
zoom in on the analytics design cycle
Design Practices
zoom in on the material practices of analytics design
66. Expertises/stakeholders and key transitions
in designing a Learning Analytics system
Educational/Learning
Sciences Researcher
Learning Theory
Educator
Learner
Learning Outcomes
Educational Insights
Programmer
Software, Hardware
User Interface
Data
Algorithm
Learning Analytics
Researcher
IF…
THEN…
72. 72
So how can we frame the
theory-analytics relationship?
73. Why are the Learning Sciences missing so
often from Learning Analytics?
• 73
April 28 2016: LAK16 Keynote
Paul Kirschner - Learning Analytics: Utopia or Dystopia
https://youtu.be/8Ojm nOiM IKI
“Put the learning back
into learning analytics”
The Learning Sciences
much to offer (helpful
examples)
Ignore us at your
dystopic peril…
74. One of the most mature fusions of
qualitative and quantitative methods
74http://w w w .quantitativeethnography.org https://w w w .youtube.com /w atch?v=LjcfGSdIBAk
LAK18 Keynote Address
75. The quant/qual distinction has dissolved.
Each has methods to enrich the other
75
“In the age of Big Data, we have an opportunity to expand the tools of
ethnography — and history, and literary analysis, and philosophy, and any
discipline that analyzes meaning — by using statistical
techniques not to supplant grounded understanding,
but to expand it. To use additional warrants to support
the stories that we tell about the things people do,
and reasons they do them.”
David Williamson Shaffer, Quantitative Ethnography, p.398
76. • 76
Machine learning ≠ atheoretical empiricism
Rosé, C. P. (2018). Learning analytics in the
Learning Sciences, invited chapter in F. Fischer,
C. H m elo-Silver, S. Goldm an, & P. Reim ann
(Eds.) International H andbook of the Learning
Sciences, Taylor & Francis.
“The strong emphasis on empiricism grounded in big data
advocated by data mining researchers can sometimes be
misunderstood as an advocacy of atheoretical
approaches.”
“I caution against a bottom-up, atheoretical
empiricism. In contrast, I would stress the role of rich
theoretical frameworks for motivating
operationalizations of variables”
“[strive for] intensive exchange between the
Learning Sciences and neighboring fields of data
mining, computational linguistics, and other areas of
computational social sciences.”
Carolyn Rosé, 2018
77. • 77
Data science opens new ways to observe learning
“Computational tools, which include machine learning approaches,
can serve as lenses through which researchers may
make observations that contribute to theory,
as machinery used to encode operationalizations of
theoretical constructs, and as languages to build
assessments that measure the world in terms of these
operationalizations.”
[…]
“They are more limited in the sense that in applying them, a
reduction of the richness of signal in the real world occurs as a
necessary discretization takes place. However, they are also
more powerful in the sense of the speed and ubiquity of the
observation that is possible.”
Carolyn Rosé, 2018
Rosé, C. P. (2018). Learning analytics in the
Learning Sciences, invited chapter in F. Fischer,
C. H m elo-Silver, S. Goldm an, & P. Reim ann
(Eds.) International H andbook of the Learning
Sciences, Taylor & Francis.
78. Learning Sciences meets Learning Analytics:
Helpful accounts of justified concerns, and demonstrable synergies, e.g…
78
Wise & Cui (ICLS 2018): 7 principles for Learning Sciences-aware analytics
Jivet, et al. (LAK 2018): 3 ways to ground learning dashboards in the
Learning Sciences
Marzouk, et al. (AJET 2016): grounding automated student feedback in
Self-Determination Theory
Rummel, et al. (IJAIED 2016): AI+CSCL could be great, or end very badly
Wise and Schwarz (IJCSCL 2018): 8 provocations for CSCL, inc. 2 debating
the role of computational methods
79. 79
How do we bridge
“from clicks to constructs”?
80. Proxies for
“Conscientiousness”?
Shute, V. J. and M. Ventura (2013). Stealth Assessment: Measuring
and supporting learning in video games. Cambridge, MA, MIT Press.
Figure 5 fromreport to The John D. and Catherine T. MacArthur
Foundation Reports on Digital Media and Learning
http://myweb.fsu.edu/vshute/pdf/Stealth_Assessment.pdf
81. Proxies for
“Conscientiousness”?
Shute, V. J. and M. Ventura (2013). Stealth Assessment: Measuring
and supporting learning in video games. Cambridge, MA, MIT Press.
Figure 5 fromreport to The John D. and Catherine T. MacArthur
Foundation Reports on Digital Media and Learning
http://myweb.fsu.edu/vshute/pdf/Stealth_Assessment.pdf
82. Proxies for
“Conscientiousness”?
Shute, V. J. and M. Ventura (2013). Stealth Assessment: Measuring
and supporting learning in video games. Cambridge, MA, MIT Press.
Figure 5 fromreport to The John D. and Catherine T. MacArthur
Foundation Reports on Digital Media and Learning
http://myweb.fsu.edu/vshute/pdf/Stealth_Assessment.pdf
83. Proxies for
“Conscientiousness”?
Shute, V. J. and M. Ventura (2013). Stealth Assessment: Measuring
and supporting learning in video games. Cambridge, MA, MIT Press.
Figure 5 fromreport to The John D. and Catherine T. MacArthur
Foundation Reports on Digital Media and Learning
http://myweb.fsu.edu/vshute/pdf/Stealth_Assessment.pdf
84. 84
Milligan, S. and Griffin, P. (2016). Understanding learning and learning design in MOOCs: A measurement-based
interpretation. Journal of Learning Analytics, 3(2), 88– 115. http://dx.doi.org/10.18608/jla.2016.32.5
From clicks to constructs in MOOCs
Defining a C21 capability of “Crowd-Sourced Learning”
89. 89
From clicks to constructs in
MOOCs
Defining a C21 capability of
Crowd-Sourced
Learning
(Part of a
larger map)
Milligan, S. and Griffin, P. (2016).
Understanding learning and learning
design in MOOCs: A measurement-based
interpretation. Journal of Learning
Analytics, 3(2), 88– 115.
http://dx.doi.org/10.18608/jla.2016.32.5
90. 90
From clicks to constructs in
MOOCs
Defining a C21 capability of
Crowd-Sourced
Learning
(Part of a
larger map)
Milligan, S. and Griffin, P. (2016).
Understanding learning and learning
design in MOOCs: A measurement-based
interpretation. Journal of Learning
Analytics, 3(2), 88– 115.
http://dx.doi.org/10.18608/jla.2016.32.5
91. 91
From clicks to constructs in
MOOCs
Defining a C21 capability of
Crowd-Sourced
Learning
(Part of a
larger map)
Milligan, S. and Griffin, P. (2016).
Understanding learning and learning
design in MOOCs: A measurement-based
interpretation. Journal of Learning
Analytics, 3(2), 88– 115.
http://dx.doi.org/10.18608/jla.2016.32.5
92. 92
From clicks to constructs in
MOOCs
Defining a C21 capability of
Crowd-Sourced
Learning
(Part of a
larger map)
Milligan, S. and Griffin, P. (2016).
Understanding learning and learning
design in MOOCs: A measurement-based
interpretation. Journal of Learning
Analytics, 3(2), 88– 115.
http://dx.doi.org/10.18608/jla.2016.32.5
95. 95
Can we learn from the history of HCI theory?
From narrow experimental science to rich, timely design input
The definition of “theory” has evolved.
Matured from a focus on cognitive
psychology’s concepts and experimental
methods, to broader, richer theories
• from modelling individual mental states, to
cognition as embodied, social, distributed
• handle the complexity of real use contexts
• inform design on realistic timescales
96. The diverse contributions of “theory” in HCI
research and design In what senses do
Learning Analytics use
theory — and what forms do
the Learning Sciences offer
to designers?
97. 97
2 design lenses:
Analytic Accountability Cycle
zoom in on the analytics design cycle
Design Practices
zoom in on the material practices of analytics design
99. Co-designing an automatic feedback tool for nurses,
with students and academics
Carlos G. Prieto-Alvarez, Roberto Martinez-Maldonado, & Anderson, T. (2018). Co-designing learning analytics tools with learners. In Jason M. Lodge, Jared Cooney Horvath, & L.
Corrin (Eds.), Learning analytics in the classroom: Translating learning analytics research for teachers (Vol. 1). London: Routledge.
Card-sorting exploration
Sketching Learner/Data Journeys to show where analytics might help
Giving voice to students, teachers and
designers by adapting techniques
from Co-Design
Helping stakeholders understand
data privacy, collection and use in
learning analytics tools
103. Automated formative feedback on writing (Civil Law)
Knight, S., Buckingham Shum , S., Ryan, P., Sándor, Á. and W ang, X. (2018). Designing Academ ic W riting Analytics for Civil Law Student Self-Assessm ent.
International Journal of Artificial Intelligence in Education, 28, (1), 1-28. DOI: https://doi.org/10.1007/s40593-016-0121-0
104. Law academic annotates automated feedback in Word
Knight, S., Buckingham Shum , S., Ryan, P., Sándor, Á. and W ang, X. (2018). Designing Academ ic W riting Analytics for Civil Law Student Self-Assessm ent.
International Journal of Artificial Intelligence in Education, 28, (1), 1-28. DOI: https://doi.org/10.1007/s40593-016-0121-0
105. Law academic annotates automated feedback in Word
Knight, S., Buckingham Shum , S., Ryan, P., Sándor, Á. and W ang, X. (2018). Designing Academ ic W riting Analytics for Civil Law Student Self-Assessm ent.
International Journal of Artificial Intelligence in Education, 28, (1), 1-28. DOI: https://doi.org/10.1007/s40593-016-0121-0
106. Align the assessment rubric with the textual features (i.e. rhetorical
moves) that the tool can identify
Knight, S., Buckingham Shum , S., Ryan, P., Sándor, Á. and W ang, X. (2018). Designing Academ ic W riting Analytics for Civil Law Student Self-Assessm ent.
International Journal of Artificial Intelligence in Education, 28, (1), 1-28. DOI: https://doi.org/10.1007/s40593-016-0121-0
107. Evaluate with students: what worked well?
Knight, S., Buckingham Shum , S., Ryan, P., Sándor, Á. and W ang, X. (2018). Designing Academ ic W riting Analytics for Civil Law Student Self-Assessm ent.
International Journal of Artificial Intelligence in Education, 28, (1), 1-28. DOI: https://doi.org/10.1007/s40593-016-0121-0
…but it was far from perfect: see the paper for detailed evaluation results
108. Automated feedback on reflective writing
Reflection is critical to the integration of academic +
experiential knowledge
This is where you disclose what you’re uncertain
about, and how you’ve changed, in the first person
Scholarship clarifies the hallmarks of deeper
reflective writing
Gibson, A., Aitken, A., Sándor, Á., Buckingham Shum , S., Tsingos-Lucas, C. and Knight, S. (2017). Reflective W riting Analytics for Actionable Feedback. Proceedings of 7th International
Conference on Learning Analytics & Know ledge, M arch 13-17, 2017, Vancouver, BC, Canada. (ACM Press). DOI: http://dx.doi.org/10.1145/3027385.3027436
109. Learning reflective writing:
Distillation of theory and pedagogy into a framework
Gibson, A., Aitken, A., Sándor, Á., Buckingham Shum , S., Tsingos-Lucas, C. and Knight, S. (2017). Reflective W riting Analytics for Actionable Feedback. Proceedings of 7th International
Conference on Learning Analytics & Know ledge, M arch 13-17, 2017, Vancouver, BC, Canada. (ACM Press). DOI: http://dx.doi.org/10.1145/3027385.3027436
110. Learning reflective writing:
Distillation of theory and pedagogy into a framework
Gibson, A., Aitken, A., Sándor, Á., Buckingham Shum , S., Tsingos-Lucas, C. and Knight, S. (2017). Reflective W riting Analytics for Actionable Feedback. Proceedings of 7th International
Conference on Learning Analytics & Know ledge, M arch 13-17, 2017, Vancouver, BC, Canada. (ACM Press). DOI: http://dx.doi.org/10.1145/3027385.3027436
111. Learning reflective writing:
Simplification of framework à a visual language
Gibson, A., Aitken, A., Sándor, Á., Buckingham Shum , S., Tsingos-Lucas, C. and Knight, S. (2017). Reflective W riting Analytics for Actionable Feedback. Proceedings of 7th International
Conference on Learning Analytics & Know ledge, M arch 13-17, 2017, Vancouver, BC, Canada. (ACM Press). DOI: http://dx.doi.org/10.1145/3027385.3027436
112. Information design + Interface design
The key to automated annotations of the reflection
114. Example design problem: initial detection of affect poorly calibrated:
red-lining words that clearly don’t reflect author affect/emotion in their writing
http://heta.io/how-can-writing-analytics-researchers-rapidly-codesign-feedback-with-educators
115. Participatory prototyping builds trust in the NLP
http://heta.io/how-can-writing-analytics-researchers-rapidly-codesign-feedback-with-educators
Learning Analytics researchers work with
academics (3 hour workshop)
Goal: calibrate the parser detecting affect in
reflective writing, working through sample texts
Rapid prototyping with a Python notebook, then
integrated into end-user tool for further testing
116. Transparency in the analytics infrastructure:
Academic Writing Analytics platform now open source
https://utscic.edu.au/open-source-writing-analytics
Higher Ed. Text Analytics Project: http://heta.io
Demo: http://acawriter-demo.utscic.edu.au
118. 118
how do we handle inherent
imperfection in analytics for
complex competencies?
119. “Embracing Imperfection in Learning Analytics”
119
Kirsty Kitto, Sim on Buckingham Shum , and Andrew Gibson. (2018). Em bracing Im perfection in Learning Analytics. In Proceedings LAK18: International Conference on
Learning Analytics and Know ledge, M arch 5–9, 2018, Sydney, N SW , Australia, pp.451-460. (ACM , N ew York, N Y, USA). https://doi.org/10.1145/3170358.3170413
Cognitive dissonance when feedback violates the student’s expectations:
“…as D’Mello and Graesser [15] demonstrate, it is when the
student experiences dissonance because the analytics
fail to match their expectations that they are likely to
reflect on why they think the machine is wrong. We
believe that this form of critical questioning is more likely to happen if
the student has been given an underlying reason to be a little distrustful
of the classifier.”
120. “Embracing Imperfection in Learning Analytics”
120
Kirsty Kitto, Sim on Buckingham Shum , and Andrew Gibson. (2018). Em bracing Im perfection in Learning Analytics. In Proceedings LAK18: International Conference on
Learning Analytics and Know ledge, M arch 5–9, 2018, Sydney, N SW , Australia, pp.451-460. (ACM , N ew York, N Y, USA). https://doi.org/10.1145/3170358.3170413
Mindful engagement with technology:
“Salomon et al. [42] are concerned that students move beyond
mindless use of potentially powerful cognitive tools, and instead
employ “nonautomatic, effortful, and thus
metacognitively guided processes” [p4].
This is precisely the role that we have been arguing
that “imperfect analytics” can help to facilitate.”
121. “Embracing Imperfection in Learning Analytics”
121
1. Robust learning design
ensures that the activity
involving automated feedback
is meaningful whether or not
the technology always works
(Knight et al ICLS 2018
crossover paper)
2. Explicit encouragement
— in student briefings, and in
the user interface — to push
back if they disagree with the
feedback
Kirsty Kitto, Sim on Buckingham Shum , and Andrew Gibson. (2018).
Em bracing Im perfection in Learning Analytics. In Proceedings LAK18:
International Conference on Learning Analytics and Know ledge, M arch
5–9, 2018, Sydney, N SW , Australia, pp.451-460. (ACM , N ew York, N Y,
USA). https://doi.org/10.1145/3170358.3170413
122. Macro-level
Critical infrastructure studies reveal how
KI is inherently political, social and
technical – an evolved system of systems
…but there is a risk that unless you have skin in the game,
critics of Learning Analytics/AI will be perceived as just
‘shouting from the touchline’ (from Bohr’s Quadrant)
123. Micro-level
We need ‘insider’ accounts of how design
practices can bring the different disciplines
and stakeholders together, with integrity
124. 124
“What would data science look like if its key
critics were engaged to help improve it?
…and how might critiques of data science improve
with an approach that considers the day-to-day
practices of data science?”
Gina N eff, Anissa Tanw eer, Brittany Fiore-Gartland, and Laura Osburn (2017). Critique and Contribute: A Practice-Based Fram ew ork
for Im proving Critical Data Studies and Data Science. Big Data, Volum e 5, N um ber 2, 2017. https://doi.org/10.1089/big.2016.0050
127. How do we ensure that LS and LA
are on the field, on the same team,
and shaping the game?
Are we equipped to shape the new
knowledge infrastructure?
formalisable theory?
sufficiently agile methods?
suitably skilled professionals?
128. Take home message:
Learning Analytics – with the help of the
Learning Sciences – must develop design
practices that bring the different disciplines
and stakeholders together, with integrity
This is an exhilarating time to be shaping
educational research and practice!