If we can scale to 100 students, why not 100,000? Armando Fox, David - - PowerPoint PPT Presentation

▶
if we can scale to 100 students why not 100 000
SMART_READER_LITE
LIVE PREVIEW

If we can scale to 100 students, why not 100,000? Armando Fox, David - - PowerPoint PPT Presentation

Nov 10, 2022 266 likes •396 views

If we can scale to 100 students, why not 100,000? Armando Fox, David Patterson {fox,patterson}@cs.berkeley.edu 9 Background: new CS169 2009-2011: enrollment 35=>50=>75=>110=>175 (Fall12) Feb

slide-1
SLIDE 1

“If we can scale to 100 students, why not 100,000?”

Armando Fox, David Patterson {fox,patterson}@cs.berkeley.edu

9

slide-2
SLIDE 2

Background: “new” CS169

  • 2009-2011: enrollment

35=>50=>75=>110=>175 (Fall12)

  • Feb 2011: start textbook
  • Nov 2011: agree to offer first 5

weeks on Coursera

  • Same quizzes, HWs,

deadlines (lag by 5 weeks) as UCB

10

http://saasbook.info tinyurl.com/about-saas

slide-3
SLIDE 3

11

Whatʼs a MOOC?


(Massive Open Online Course)

Characteristic What we did A plausible alternative Content delivery 7-10 minute lecturelets 60-90 minute lectures Assessment Deep autograding Peer grading; self- assessment only Forum monitoring TA assigned to help Youʼre on your own Content capture Screencast of live lecture Studio + postproduction Pacing Synchronous deadlines Self-paced On-campus course Traditional lectures “flipped classroom”

slide-4
SLIDE 4

Key Changes for MOOC

  • Nontrivial autograders for programming

assignments (open source)

  • Adapting lectures to 7-10 min segment +

peer learning/self assessment question

– 7-10 min segment + peer learning question – 8-10 hrs/week ugrad to convert & format videos

  • TA support to monitor question forums
  • No final project
  • Non-change: same HWs, quizzes, deadlines

12

slide-5
SLIDE 5

0.4%1% 2% 3% 4% 8% 15% 28% 32% 7% 0.3%

0.0% 5.0% 10.0% 15.0% 20.0% 25.0% 30.0% 35.0%

1950 1955 1960 1965 1970 1975 1980 1985 1990 1995 2000

Year of Birth

Who are these students?

  • 12% female, 88% male
  • Median: 27 years old
  • 75% of class: 21 to 38
  • From 10 to 106
slide-6
SLIDE 6

33%

Grad degree

31%

BA or BS

12%

Some 
 college

12%

Profʼnl degree

8%

HS

Who are these students?

  • 75% Baccalaureate or higher; 7% instructors
  • 60% do SW dev/maint at job

3% Associates degree 1% < HS degree

Busy people, many with high expectations

slide-7
SLIDE 7

Funneling & Stratification

15

50,000 
 “registered” 25,000 
 watch ≥1 lecture 10,000 
 submit ≥1 HW 3,500
 “passed”

  • “I want to help with future
  • fferings”
  • “Better than any course

available at my university”

90% “attrition” confirmed by 3

  • ther MOOCs, including MITx
slide-8
SLIDE 8

Autograding Strategies

16

Submission Grading strategy Upload code file (s)

  • RSpec (correctness)
  • [soon] reek/flay (code style)

Upload test case files

  • Mutation testing (Amman &

Offutt): app with inserted bugs should fail tests Submit URI of cloud-deployed app (Heroku)

  • Remote (cloud-based)

integration test using Mechanize Interactive short-answer/ multiple-choice

  • Our tools emit both printed &

Coursera-compatible (online) quizzes

Grading strategy

submi ssion rubric feed- back

95 100

slide-9
SLIDE 9

Neutralizing direct costs

  • $0.30 Hosted download of large VM file

– Google & Microsoft donation: $20K credits

  • <$1 cloud-based autograding

– Amazon donation: $8K credits

  • $10 Cloud computing (AWS credits)

– Amazon donation: $500K credits

  • $20 Private GitHub repo for 90 days

– GitHub donation: $1M in account credits

  • $10 E-textbook (in our case)
  • Free but could improve with donation: app hosting
  • n Heroku, cloud-based integration testing

17

slide-10
SLIDE 10

MOOC Lessons for Classroom

  • Zero-config courseware works

– downloadable or EC2-deployable VM image – hosted dev tools (Tracker, Heroku, GitHub…)

  • Autograding works

– Demands bug-free assignments up front – Frontloaded work to create autograders, many improvements planned – Easier to create new autograding scripts

  • MOOC improved on-campus course

– and MOOC >> recording on-campus course!

18

slide-11
SLIDE 11

New Opportunities

  • Which students are making similar

mistakes?

  • Can we find exemplar of good solution and

use for hint?

  • How do ad-hoc communities impact learning
  • utcomes?
  • Can autograding technology also assist

manual grading?

  • Yes, weʼre open sourcing everything

19