Big Data and Learning Analytics (Bichel)

2014 UBTech Big Data and
Learning Analytics SIG
Anthony Bichel, Ph.D.
Leading Edge Learning
Analytics: Student Success Science
“Big Things Have Small Beginnings”
What problems are we trying to solve?
What questions are we trying to answer?
What are the right elements to measure?
What level or mode of analysis works best for
our particular problem?
• How does this align with other institutional or
strategic priorities?
Learning Analytics Defined & (Re)Defined
• Learning analytics is the “use of data, statistical analysis, explanatory and
predictive models to gain insights and act on complex issues.”
Educause 2012
“Learning analytics refers to the interpretation of a wide range of data
produced by and gathered on behalf of students in order to assess academic
progress, predict future performance, and spot potential issues.”
Horizon Report 2012
“Learning analytics is the field associated with deciphering trends and patterns
from educational big data, or huge sets of student-related data, to further the
advancement of a personalized, supportive system of higher education.”
Horizon Report 2013
Analytics Drivers
• Institutional Compliance / Accountability
– We have to
• Institutional Performance
– Its good for us
• Student Benefit
– Its good for them
• Student Management
– Because we can
“All things are subject to
interpretation. Whichever
interpretation prevails at
a given time is a function
of power and not truth.”
Friedrich Nietzsche
Five Steps of Analytics
“Data is the foundation of all analytics efforts.”
• Capture
– Selecting and Organizing
– Policy Decisions
• Report
• Predict
• Act
• Refine
Examples of Learning Analytics Models
VLE Dashboards
Data Visualization
Social Network Analysis
Discourse Analytics
Predictive Analytics
Adaptive Learning
Disposition Analytics
Data Visualization
The Benefits of Analytics
“Learning analytics can help faculty improve teaching and learning opportunities for students”
(Hrabowski, Suess, & Fritz, 2011; Mattingly, Rice, & Berg, 2012).
The most valuable things about data are: (1) data about behavior and (2) changes in behavior
Ways in which analytics can help educational institutions improve student achievement:
Monitoring individual and cohort performance
Identifying outliers for early intervention
Predicting potential so that all students achieve optimally
Preventing attrition from a course or program
Identifying and developing effective instructional techniques
Analyzing standard assessment techniques and instruments
Testing and evaluation of curricula
Thematic Issues
• Institutional Success Factors
• Policy Concerns
• Privacy & Consent
• Faculty Evaluation
Background Research
• Surveyed north Texas ISD’s about analytics use
• Surveyed Association of College & University
Policy Administrators (ACUPA) membership
• Surveyed 24 institutions implementing BbA
and using either Banner or PeopleSoft SIS
Institutional Success Factors
• Organizational Capacity
• Leadership & Vision
• Tactical Readiness
• Faculty Development
• Learning Spaces
Organizational Capacity
Strategic Choke
Leadership & Vision
Office of the President
Business Affairs &
Student Affairs
Academic Analytics &
Inst. Effectiveness &
Faculty Development
Decision Support &
Student Services
Office of Records &
Faculty Affairs
Office of Academic
Deans Council
Client Services
Digital Teaching &
Center for Distance
Education (Bb)
Center for Teaching &
Learning Excellence
Business Services
Academic Programs &
Planning &
Studies & Dir.
University College
Link Lab
DA: Data Analysis
SE: Statistical Expertise
CE: Content Expertise
PE: Policy Expertise
Tactical Readiness
Educational institutions often lack the skills necessary to incorporate data into
their everyday workflows, such as administrators and staff who can inculcate
and support a data-based culture. Examples include:
Data analytics experts
Data visualization specialists
Data managers
Instructional designers
Graphic designers
Digital media specialists
App developers
Marketing manager for analytics and technology
In addition to faculty, how many staff will require assistance and what will the
nature of that help be?
How many IT support positions will be needed as analytics usage grows?
Models of Faculty Development
• Volunteer-driven
– Initiated by individual faculty member seeking specialized assistance
• Facilitator-driven
– Typical classroom style of group instruction/training
• Technology-driven
– Mediated support services delivered on-line via email, blogs, wikis, streaming
and interactive video, or some form of social media
• Analytics-driven
– Faculty receives robust, actionable, student and cohort performance profiles
(dashboards) that demonstrate learner performance and/or mastery in near to
real-time that result in dynamic adjustments to course content, delivery mode
or learning pathways.
Faculty Development Modes
• Pre-Analytics
– Lower costs
– Improve efficiency
– Increase productivity
“It's like they say in the Internet world
- if you're doing the same thing today
you were doing six months ago, you're
doing the wrong thing.” - Bruce Feiler
• Post-Analytics
– Change the ways that faculty think about and use
– Challenge the assumptions and bias that faculty bring
to decision making
– Faculty use data to generate new insights and
contexts to serve students better
Faculty Development Challenges
• The 2nd law of Faculty Thermodynamics: Any project that requires
sustained faculty energy probably won’t succeed.
• Faculty development for using analytics in the classroom involves
data about faculty performance in the classroom.
• Dashboards are not the nightly news.
• Follow-up and follow-through is critical when dealing with
• Individual faculty are not the only players – faculty managers and
student support staff need to be involved.
Typical Faculty Profile (R1)
• Total Faculty = 1,621
– Tenured faculty = 446 (28%)
– Tenure track = 179 (11%)
– Non-tenured = 996 (61%)
• Faculty Profile by Rank
Professor = 221
Assoc. Professor = 234
Assist. Professor = 170
Other faculty = 992
Unknown = 4
• Dept. Teaching Assistants = 703
• Total Instructors = 2,324
• Current Faculty LMS Users = 30% (486) – 50% (810)
UTA Suggestions
• All Common Core (CC) faculty must use Bb
• All CC faculty must attend specialized training
for using analytics in the classroom
• All CC faculty, student support staff and
department chairs must attend a joint
meeting to discuss roles, responsibilities and
desired outcomes
Learning Spaces: Collaborative Classrooms
• The archetype collaborative classroom is defined by a technologically infused
and socially-networked space that can be arranged, rearranged or disarranged
for whatever purpose is required during a f2f class.
• Popular models include the MIT “Technology Enabled Active Learning
(TEAL)” and University of Iowa “Transform, Interact, Learn, Engage (TILE)
Challenges of Collaborative Classrooms
• Classroom control or lack thereof
• Much more time intensive
• Student-centered activities often take more time than
• Individual differences among students
• Students must prepare outside of class
• TILE / TEAL classrooms are not oriented for discussion
because students can’t see another
• Lack of a focal point in the room makes lectures
problematic, if not outright counterproductive
• Faculty development and support essential to success
McKinsey Global Institute
The classroom is where many of the
most valuable applications of data will
evolve. Improved instruction can be
enabled by developing:
• Personalized learning plans for
• Frequent feedback on teacher
performance, and
• Targeted professional development
programs for educators.
“We estimate the potential value from improved instruction to
be $310 billion to $370 billion per year worldwide.”
Policy Concerns
Policy Drivers
Creating a culture of evidence on campus requires clear policies and
processes with respect to the use of data.
Developing trust and normalized processes that transcend individual
personalities and arbitrary decision making are key. Issues include:
• Data Governance
• Roles and Responsibilities
• Data Politics
“Data access is perhaps one of the most important – and difficult – policy issues to clarify.”
• Data Privacy and Fair Information Practices
• De-Identification of Data
Policy Issues
“A key component of an analytics program is first to
identify the policy questions to be answered and then
to engage in a risk-management exercise, including a
cost/benefit analysis, to determine if analytics will
provide the answers or feedback needed.”
Rodney J. Petersen, Educause Review July/August 2012
Ownership of data
Best practices
“All things are subject to
interpretation. Whichever
interpretation prevails at a
given time is a function of
power and not truth.”
Friedrich Nietzsche
Case Study: UTA Policies/Materials Impacted by
Analytics Implementation
• FERPA Training for Faculty and Staff
• Handbook of Operating Procedures (HOP)
– Rights, Responsibilities and Duties of Faculty
– Annual Review and Comprehensive Evaluation of
• UT System Rules and Regulations of the B.O.R.
• UT System Documents
Privacy & Consent
Identity in the Analytics Age
By Nathan Jurgenson
Data Privacy & Fair Information Practices
• Notice / Awareness
• Choice / Consent
• Access / Participation
• Integrity / Security
• Enforcement / Redress
• Federal Policies (Privacy Safeguards)
Student Rights Under FERPA
The Family Education Rights and Privacy Act (FERPA) affords students certain rights
with respect to their education records, including:
The right to consent to disclosures of personally identifiable information contained
in the student's education records, except to the extent that FERPA authorizes
disclosure without consent.
One exception which permits disclosure without consent is disclosure to school
officials with legitimate educational interests. A school official is a person
employed by the University in an administrative, supervisory, academic, research,
or support staff position or a person or company with whom the University has
FERPA was written before the Internet
Privacy Concerns
“Someday we'll laugh about the way we used to worry about a "credit
score," because the data will be so much deeper, more intrusive, and
more deeply hidden than credit scores ever were.”
Blogger Tim_Sims
Students shed streams of data about their academic progress, work
habits, learning styles and personal interests as they navigate educational
websites. All that data has potential commercial value:
• It could be used to target ads to students or their families, or
• To build profiles on them that might be of interest to employers,
military recruiters or college admissions officers.
The law is silent on who owns that data.
Kathleen Styles, the Education Department’s chief privacy officer,
acknowledged that much of the data is likely not protected by FERPA –
and thus can be commercialized by the companies that hold it.
“The purpose of notice and consent is that the
user assents to the collection and use of
personal data for a stated purpose that is
acceptable to that individual... this framework
is increasingly unworkable and ineffective.”
“Notice and consent fundamentally places
the burden of privacy protection on the
individual – exactly the opposite of what is
usually meant by a “right.”
“As a useful policy tool, notice and
consent is defeated by exactly the positive
benefits that big data enables: new,
non‐obvious, unexpectedly powerful uses
of data. It is simply too complicated for
the individual to make fine‐grained
choices for every new situation or app.”
“Ultimately the vision for Knewton is that
everyone should have their own learning
profile that’s free, secure, hosted in the cloud,
and just follows them around forever.”
“…We’re extremely cognizant that we hold a data set that is the most important data set
in ones’ life, other than maybe your healthcare data.“
“…In 30 years, the human race will be totally dominated by data science…In terms of
education and healthcare, I don’t think there will be an invasion of privacy because
there’s not going to be any marketing of it.”
Jose Ferreira, CEO, Knewton
Just Because it is Accessible Doesn’t Make it Ethical
It may be unreasonable to ask researchers to obtain consent from
every person, but it is unethical for researchers to justify their
actions as ethical simply because the data is accessible.
Privacy doesn’t mean the same to everyone and the end of it will
affect people of different social and economic classes differently.
Faculty Evaluation
Sloan-C Presentation
There are many areas of data that could be used to create a faculty
performance dashboard:
Student surveys
Faculty surveys
Peer surveys
Faculty issues
Professional development
GPA and grading
Defining what the organization deems important for data collection, is
the first step in creating faculty performance dashboards.
Facets of Faculty Evaluation
Traditional Educational Setting:
(1) number of assigned courses
(2) development of course curriculum
(3) number of drops
(4) GPA distribution
(5) number of refereed students in comparison to number of graduates
(6) timely grading
(7) timely final grade submission
(8) peer review
Online Environment:
(1) using the defined curricula
(2) number of drops
(3) GPA distribution
(4) weekly online course activity
(5) student responsiveness
(6) timely grading
(7) timely final grade submission
(8) faculty development opportunities
(9) peer review process
Analytics Infused Instruction
(1) Individual student learning outcomes
(2) Cohort performance profiles
(3) Adaptive curriculum
(4) Optimized modalities
Mapping Faculty Concerns
Instructional Concerns:
What are faculty obligations?
How do we handle the increase in students seeking assistance?
Who contacts the students, under what conditions, and for how long?
Should faculty know what resources are available to students needing help?
What do faculty need to know about how FERPA applies?
Are faculty obliged to create new material or to update materials based on evidence
(data) that indicates consistent failings or misunderstandings?
Professional Development Concerns:
– Will there be an orientation program that clarifies roles and responsibilities?
– What resources are available to assist faculty with course redesign?
– What expertise is needed to respond to data-driven alerts?
Faculty Evaluation Concerns:
– Is there an opt-out option?
– Analytics can reveal which teaching techniques are more effective than others, are faculty
obligated to adapt accordingly?
– What added time constraints might be incurred by using analytics?
– How will this new evidence of teaching effectiveness be incorporated into faculty
Academic Freedom