richard baraniuk rdls rice center for digital learning and scholarship update
TRANSCRIPT
agenda
• OpenStax College update
• Personalized learning update
• RDLS update
• Arnold Foundation and beyond
the goal
create a library of free, open-source learning resources that greatly expands access to high-quality learning opportunities
– student debt surpasses $1 trillion (NY Times)– 7 out of 10 students forgo buying texts (PIRG)
benefits:– improve college completion– improve student learning– positive disruption
proof of concept(phase 1)
$4.15m in venture philanthropy
5 open college textbooksto reach 10% market penetration
will save 843,750 students$83.24m through 2017 (20x ROI)
turn-key course solution textbook, mobile apps, ancillaries, homework system, analytics, …
high qualityfocus on learningwritten by professionalspeer review and classroom testing
sustaining ecosystemsupport project long-term
proof of concept(phase 1)
digital open publishing platformfounded at Rice University in 1999
1200 open textbooks/collections20,000 educational Lego blocks40 languages
>1 million users per monthfrom 190 countries
STEM content used 97 million times since 2007
technology development• Connexions’ XML/HTML5 OER platform provides
scalable distribution channel in multiple formats– HTML, PDF, ePUB, mobile, print
• By Fall 2012, books also available via Amazon, Barnes and Noble, iBooks, iTunesU, …
concept provedon-budget
– 2 books published in June; 3 to follow in early 2013– increased collaboration with CMU OLI on A&P
high quality– content professionally developed and peer reviewed– editorial boards with 21 luminaries, including
2 Nobel laureates, 2 former NSF Directors
high impact and efficiency– positive media reaction– 35 adoptions to date – $680k saved so far– College Physics has already paid back its $500k investment
growing ecosystem– 10 for-profit and non-profit partners
media awareness
“Big Savings for U.S. Students in Open-Source Book Program,” New York Times
“Free College Textbooks: Wave of the Future?” Forbes
“Rice University And OpenStax Announce First Open-Source Textbooks,” TechCrunch
“Why Pay for Intro Textbooks?” Inside Higher Ed
scaling up success
venture is exceeding expectations, but urgency to:
• keep the production line running smoothly (lowers costs and latency)
• add analytics and personalized learning functionality to increase student learning
• seize the narrow window of opportunity to disrupt the publishing industry
complete the library
$15.92m in venture philanthropy
20 additional personalizable open textbooks to reach 10% market penetration
library of 25 textbooks will save 1.6m students $156.8m through 2017
once phased-in, library of 25 textbooks will save students $782.9m every 5 years (ROI 39x)
sustaining ecosystemdisruptive force for good
textbooks that learn
personalized learning system that closes the learning feedback loop
new learning analytics(machine learning)
what works? what doesn’t?
massive open laboratory to study how we learn and really change how we teach
learning challenges
one size fits all
open-loop– students treated as passive receivers
of information– students poor at monitoring
their learning and often choose ineffective strategies
cognitively uninformed– activities that speed learning often
do not promote long-term retention or transfer
technology provides hope
personalized learning– adapt to each learner’s background,
context, abilities, goals
closed-loop– students and instructors as active
explorers of a knowledge space– tools for instructors and students to
monitor their progress
cognitively informed– leverage latest findings from the
science of learning
learn
erscontent
data
a long way to go
today’s personalized learning systems are
– proprietary (especially wrt data)
– expensive (limits access)
– fragile (based on rules)
– not scalable (limits access)
– focused on tech, not learning(creates a chasm)
learn
erscontent
data
textbooks that learn
a modern personalized learning system
– open (content, code, data)
– free (greater access)
– robust (based on machine learning)
– scalable (performance improves with more usage)
– focused on learning, not tech(crosses the chasm)
textbooks that learn
tech:digital repositoriesmachine learning
cog-sci:how to
optimize learning
open:leverage global
community
balance technology with cognitive science
cognitive science team
Elizabeth Marsh, DukeAndrew Butler, DukeHenry Roediger, WashU
“A Personalized Learning System based on Cognitive Science,” funded by NSF Cyberlearning Program, 2011
learning principles
• Retrieval practice– retrieving information from memory is not
a neutral event; rather it changes memory – “testing effect” is robust and replicable
• Spacing– distributing practice over time produces better long-term
retention than massing practice – “spacing effect” is extremely robust and replicable
• Feedback– closes the learning feedback loop– must be timely
learn
erscontent
data
textbooks that learn
tech:digital repositoriesmachine learning
cog-sci:how to
optimize learning
open:leverage global
community
machine learning for education
learning analytics– assess and track student progress– help instructors become better teachers– study what really works, what doesn’t– state-of-the-art machine learning– exploit massive data,
not hand-coded rules
scheduling– close the learning feedback loop– propose optimal learning path
for each student
(Peter Norvig)
Grade 8 science• 80 questions• 145 students• 1353 problems
solved (sparse) • 5 concepts
Concept 1: Properties of Soil 52% Classifying Matter 26% Earth, Sun, and Moon 22%Concept 2: Evidence of the Past 57% Earth, Sun, and Moon 24% Properties of Water 19%Concept 3: Mixtures and Solutions 40% Alternative Energy 34% Changes to Land 26%Concept 4: Alternative Energy 37% Earth, Sun, and Moon 35% Changes from Heat 28%Concept 5: Properties of Water 54% Formation of Fossil Fuels 27% Earth, Sun, and Moon 19%
applicationsfor instructors
• Instructor dashboard to replace grade book– estimate and track student
concept mastery, on individual and class basis
• Automatic “concept map” – estimate problem difficulty and
identify good/bad problems
• Automatically group students into “eigenstudent” groups for remediation or acceleration
• Detect cheating and gaming
• Suggest what content student(s) should study next (scheduling)
studentsconcepts
applicationsfor students
• Student dashboard to replace grade book
• Feedback on individual problems(concepts involved, etc.)
• Identify strong/weak areas, including what to watch out for when studying
• Progress through the “course map”
• Relative standing in class
• Projected final grade
• Suggest content to study next (scheduling)
studentconcepts
applicationsfor admins
• Admin dashboard – tracks student progress– tracks and compares
instructor progress
• Estimate problem difficulty and identify good/bad problems (aids curriculum design)
• Predict scores on final exams/standardized tests
• Detect cheating and gaming
• Insights into higher-level demographic effects
experiments
Ongoing: Mturk with Algebra and OSC College Physics
Fall 2012ECE courses at Rice, GaTech,UTEP, RHIT
Rice Coursera courses (2)
STEMScopes (~700,000 students)
beta testing
ELEC301 Signals and Systems – homework replacement w/ cog sci
(feedback, retrieval practice, repetition, spacing)– no machine learning based personalization
preliminary findings1. better retention and transfer of knowledge on an end-of-
semester assessment relative to standard practice2. magnitude of the benefit was almost equivalent to one letter
grade considering completely accurate use of knowledge (no partial credit) and about half of one letter grade considering giving credit for partial knowledge
summary3. OST > standard practice4. effect size ≈ 1/2 to 1 letter grade
deploying atGaTech, RHIT,
UTEP, Fall 2012
impacts• Textbooks come alive!
– one size does not fit all in education– reinvent the entire process, making
it a continuous dynamic process of exploration
– close the learning feedback loop– open access for maximum impact
• A renaissance in computer-based learning– exploit the “unreasonable effectiveness of data”
– students will learn more effectively– instructors will become better teachers– everyone will better understand what works and what doesn’t– opportunity for cognitive science research at a massive, global scale
– the future of assessment?
learn
erscontent
data
RDLS• 3 Rice-based education projects
gaining momentumOpenStax CollegeSTEMScopesPersonalized learning
• Personalized learning broadens footprint from just outreach to cutting-edge research
machine learningcognitive scienceneural engineering (eventually)
• Impacts both outside and within Rice• Opens up new opportunities for fundraising from
“education minded” donors, especially K-12
arnold foundation
• Strong resonance with RDLS goals and activities
• “The Foundation works for transformational change in K-12 public education.”
• “Learning Systems: Developing and implementing innovative approaches to learning, including competency-based, digitized curricula with built-in assessments to permit students to learn anytime, anywhere and at any pace.”
• “Performance Management: Shifting the focus of accountability systems from compliance to performance; creating clear standards and transparent, accessible data to measure performance; and developing incentive and human resources structures that use this data to drive decision-making and improve quality.”
urgency• Know of other groups approaching Arnolds soon
regarding open textbooks and learning
• Proposal concept: Make Rice and RDLS the AF’s “research lab” for digital curricula, analytics, and personalized learning– short term
curriculum development (K-12, HE)– medium term
personalized learning system (OpenStax Tutor) massive open learning data archive
(first of its kind; can be a Rice/Arnold legacy)– long term
fundamental research in machine learning, cognitive science, neuroengineering, and beyond
budget thoughts• Star cognitive science chaired professor + startup $5m• Junior cognitive science faculty $1m• Star machine learning chaired professor + startup $5m• Junior machine learning faculty $1m• Nationally prominent Postdoc program $3m• Nationally prominent Grad student program $3m• Research funds $5m
• Personalized learning software tools $5m• Open data library $2m• STEMScopes/PL integration $1m• OSC/PL integration $1m• OSC library Phase 2 $10m• Support endowment $5m
• Total $47m• Compare to edX: $60m pledge from MIT and Harvard