responding to the explosion of student interest in computer science

Tsunami or Sea Change? Responding to the Explosion of

Student Interest in Computer Science

Ed Lazowska Bill & Melinda Gates Chair in

Computer Science & Engineering University of Washington

Eric Roberts Professor of Computer Science and

Bass University Fellow in Undergraduate EducaJon Stanford University

Jim Kurose DisJnguished Professor

School of Computer Science University of MassachuseNs Amherst

NCWIT 10th Anniversary Summit, May 2014 CRA Conference at Snowbird, July 2014

(Updated 7/23/2014 following CRA Conference at Snowbird)

hNp://lazowska.cs.washington.edu/NCWIT.pdf or Snowbird.pdf

Today •  Short presentaCon of trends •  Open discussion

–  Best pracCces: how to respond? –  What’s different this Cme around, and how can we demonstrate it? –  A proacCve agenda: what can we and our disciplinary organizaCons (CRA,

CCC, ACM, NSF, CSTB, …) do?

GRIPING

0

100

200

300

400

500

600

700

800 Harvard

CS50

Introductory course enrollments are exploding

0 100 200 300 400 500 600 700 800 900 1000 University of Pennsylvania

total

CIS 110

CIS 120

0

200

400

600

800

1000

1200 MIT

total

6.01

6.00

0

100

200

300

400

500

600

700

800 Harvard

CS50

Not just at Harvard

0

500

1000

1500

2000

2500 Stanford

total

CS 106A

CS 105

CS 101

Not just at elite private insCtuCons

0

500

1000

1500

2000

2500

3000 University of Washington

CSE 142

0

200

400

600

800

1000

1200

1400 University of Michigan

EECS 280

0

500

1000

1500

2000

2500 UC Berkeley

CS61A/S

CS10

E7

These enrollments are blowing past previous highs

0

500

1000

1500

2000

2500

3000 University of Washington

CSE 142

CSE 143

Increasing proporCons of students are taking second courses

20%

30%

40%

50%

60%

70% University of Washington

CSE 142-‐>143 %

In at least some cases, female parCcipaCon is increasing

20%

25%

30%

35%

40% University of Washington

CSE 142 % female

0

50

100

150

200

250

300

350 Harvard

Majors

0

200

400

600

800

1000

1200 MIT

Total

6.3

6.2

0

100

200

300

400

500

600 University of Pennsylvania

Majors

Demand for the major is increasing

0

100

200

300

400

500

600

700

800 Stanford

Majors

2007-‐08 Economics Government Social Studies Psychology (PSSR) English & Amer Lit & Lang History Anthropology History and Literature Biochemical Sciences Applied MathemaCcs Molecular and Cellular Biology Human EvoluConary Biology Neurobiology Biology MathemaCcs Sociology Chemistry Physics Visual and Environmental Studies History and Science Computer Science Engineering and Applied Science (AB) Chemical & Physical Biology Environ. Sci & Pub Policy Fine Arts / History of Art & Arch

Top 25 concentraCons at Harvard 2013-‐14 Economics Government Social Studies Psychology (PSSR) Computer Science Applied MathemaCcs Neurobiology History English Human Developmental & Regener. Biol. Sociology History and Literature StaCsCcs Human EvoluConary Biology Organismic & EvoluConary Biology History and Science Chemistry MathemaCcs Physics Molecular and Cellular Biology Engineering and Applied Science (SB) Anthropology Fine Arts / History of Art & Arch Biomedical Engineering Visual and Environmental Studies

Dot-‐com peak: 3,237 Dot-‐bust trough: 1,732 Most recent year: 6,174 Target enrollment: 135 (+ ~40 upper-‐division transfers)

0"

1000"

2000"

3000"

4000"

5000"

6000"

7000"

1995"

1996"

1997"

1998"

1999"

2000"

2001"

2002"

2003"

2004"

2005"

2006"

2007"

2008"

2009"

2010"

2011"

2012"

2013"

2014"

Carnegie(Mellon(University(

SCS"Freshman"Applicants"

SCS"Freshman"Class"Target"Size"

CMU freshman applicants

Non-‐major demand for upper-‐division courses is increasing, also

0"

20"

40"

60"

80"

100"

120"

2010(11" 2011(12" 2012(13" 2013(14"

Harvey'Mudd'College'CS'70'

Non(Majors"

0"

50"

100"

150"

200"

250"

300"

350"

400"

450"

500"

2008"Fall"

2009"Spring"

2009"Fall"

2010"Spring"

2010"Fall"

2011"Spring"

2011"Fall"

2012"Spring"

2012"Fall"

2013"Spring"

2013"Fall"

2014"Spring"

UC#Berkeley#Upper#Division#CS#Enrollment#from#L&S#Outside#of#CS##

L&S:"Undeclared"

L&S:"Math"&"Phys"Sci"Div"

L&S:"Social"Sciences"Div"

L&S:"Undergraduate"Div"

L&S:"Arts"&"HumaniGes"Div"

L&S:"Bio"Sciences"Div"

L&S:"AdminICS"

Upper-‐division class sizes are going through the roof

0"

50"

100"

150"

200"

250"

300"

CS"3410"*"Computer"Architecture"

CS"4410"*"Opera8ng"Systems"

CS"5150"*"So=ware"Engineering"

Cornell'Upper'Division'Courses'

Enrollment"

0"

50"

100"

150"

200"

250"

300"

CS"161"*"Op

era0ng"Systems"

CS"179"*"User"Interface

s"

CS"181"*"Mach

ine"Learning"

CS"171"*"Visua

liza0on"

CS"109"*"Data"Science"

Harvard&Upper*Division&Courses&

Enrollment"

Upper-‐division class sizes are going through the roof

0"

50"

100"

150"

200"

250"

300"

CS"3410"*"Computer"Architecture"

CS"4410"*"Opera8ng"Systems"

CS"5150"*"So=ware"Engineering"

Cornell'Upper'Division'Courses'

Enrollment"

760 students!

CS229 Machine Learning Autumn 2013

Why is this man smiling?

0"

50"

100"

150"

200"

250"

300"

CS"161"*"Op

era0ng"Systems"

CS"179"*"User"Interface

s"

CS"181"*"Mach

ine"Learning"

CS"171"*"Visua

liza0on"

CS"109"*"Data"Science"

Harvard&Upper*Division&Courses&

Enrollment"

We’re all familiar with cycles in demand

0"

10,000"

20,000"

30,000"

40,000"

50,000"

60,000"

70,000"

1966"

1967"

1968"

1969"

1970"

1971"

1972"

1973"

1974"

1975"

1976"

1977"

1978"

1979"

1980"

1981"

1982"

1983"

1984"

1985"

1986"

1987"

1988"

1989"

1990"

1991"

1992"

1993"

1994"

1995"

1996"

1997"

1998"

1999"

2000"

2001"

2002"

2003"

2004"

2005"

2006"

2007"

2008"

2009"

2010"

2011"

2012"

Computer)Science)Bachelors)Degrees)Granted)

But this Cme, it truly feels different •  Students are figuring out that every 21st century ciCzen needs to

have facility with “computaConal thinking” – problem analysis and decomposiCon (stepwise refinement), abstracCon, algorithmic thinking, algorithmic expression, stepwise fault isolaCon (debugging), modeling – driving introductory course demand –  Programming is the hands-‐on, inquiry-‐based

way that we teach computaConal thinking and the principles of computer science

•  Students are figuring out that fields from Anthropology to Zoology are becoming informa(on fields, and that those who can bend the power of the computer to their will – computaConal thinking, but also computer science in greater depth – will be posiConed for greater success than those who can’t

•  Students are figuring out that computer science is not Dilbert – it’s an intellectually exciCng, highly creaCve and interacCve, “power to change the world” field

•  Students are figuring out that all of the STEM jobs are in computer science

71%$

15%$

3%$

3%$

4%$4%$

Job$Growth,$2012.22$.$U.S.$Bureau$of$Labor$Sta8s8cs$Computer$Occupa8ons$=$71%$of$all$STEM$

Computer$Occupa3ons$

Engineers$(Aerospace,$Biomedical,$Chemical,$Civil,$Electrical,$Electronics,$Environmental,$Industrial,$Materials,$Mechanical,$Other)$

Life$Scien3sts$(Agricultural$&$Food$Scien3sts,$Biological$Scien3sts,$Conserva3on$Scien3sts$&$Foresters,$Medical$Scien3sts,$Other)$

Physical$Scien3sts$(Astronomers,$Physicists,$Atmospheric$&$Space$Scien3sts,$Chemists$&$Materials$Scien3sts,$Environmental$Scien3sts$&$Geoscien3sts,$Other)$

Social$Scien3sts$and$Related$Workers$(Economists,$Survey$Researchers,$Psychologists,$Sociologists,$Urban$&$Regional$Planners,$Anthropologists$&$Archeologists,$Geographers,$Historians,$Poli3cal$Scien3sts,$Other)$

Mathema3cal$Science$Occupa3ons$

Data from the spreadsheet linked at hpp://www.bls.gov/emp/ep_table_102.htm


57%$

25%$

5%$

5%$

5%$3%$

Job$Openings$(Growth$And$Replacement),$2012>22$>$U.S.$Bureau$of$Labor$StaFsFcs$Computer$OccupaFons$=$57%$of$all$STEM$

Computer$Occupa2ons$

Engineers$(Aerospace,$Biomedical,$Chemical,$Civil,$Electrical,$Electronics,$Environmental,$Industrial,$Materials,$Mechanical,$Other)$

Life$Scien2sts$(Agricultural$&$Food$Scien2sts,$Biological$Scien2sts,$Conserva2on$Scien2sts$&$Foresters,$Medical$Scien2sts,$Other)$

Physical$Scien2sts$(Astronomers,$Physicists,$Atmospheric$&$Space$Scien2sts,$Chemists$&$Materials$Scien2sts,$Environmental$Scien2sts$&$Geoscien2sts,$Other)$

Social$Scien2sts$and$Related$Workers$(Economists,$Survey$Researchers,$Psychologists,$Sociologists,$Urban$&$Regional$Planners,$Anthropologists$&$Archeologists,$Geographers,$Historians,$Poli2cal$Scien2sts,$Other)$

Mathema2cal$Science$Occupa2ons$

Data from the spreadsheet linked at hpp://www.bls.gov/emp/ep_table_102.htm


0" 500" 1000" 1500" 2000" 2500" 3000" 3500" 4000" 4500" 5000"

Computer"Science"

Engineering"

Health"Professions"(gap"exists"at"graduate/professional"level"only)"

Research,"Science,"Technical"(gap"exists"at"graduate"level"only)"

Washington*State*High*Demand*Fields*at*Baccalaureate*Level*and*Above*WSAC,*SBCTC,*WTECB,*October*2013*

Current"CompleGons"

AddiGonal"Annual"CompleGons"Needed"2016K21"

Data from Table 2 of the report linked at hpp://www.wsac.wa.gov/sites/default/files/2013.11.16.Skills.Report.pdf

It seems likely that the recent dramaCc growth will conCnue (although cycles are inevitable)

0"

10,000"

20,000"

30,000"

40,000"

50,000"

60,000"

70,000"

1966"

1967"

1968"

1969"

1970"

1971"

1972"

1973"

1974"

1975"

1976"

1977"

1978"

1979"

1980"

1981"

1982"

1983"

1984"

1985"

1986"

1987"

1988"

1989"

1990"

1991"

1992"

1993"

1994"

1995"

1996"

1997"

1998"

1999"

2000"

2001"

2002"

2003"

2004"

2005"

2006"

2007"

2008"

2009"

2010"

2011"

2012"

Computer)Science)Bachelors)Degrees)Granted)

How are we, and our insCtuCons, going to respond? •  10% of Princeton’s students are computer science majors

–  It’s a far greater percentage at, e.g., MIT

•  10% of Princeton’s faculty are unlikely to ever be in computer science! –  Dipo, proporConately, at MIT

•  And then there is –  Introductory course demand ++ –  Upper-‐division non-‐major demand ++ –  Graduate non-‐major demand ++

We have seen this movie before, and it wasn’t prepy! •  In the middle 1980s and the middle 2000s, student demand

increased tremendously •  UniversiCes did not respond adequately •  Kent CurCs report:

“80% of universiCes are responding by increasing teaching loads, 50% by decreasing course offerings and concentraCng their available faculty on larger but fewer courses, and 66% are using more graduate-‐student teaching assistants or part-‐Cme faculty. 35% report reduced research opportuniCes for faculty as a result … these measures make the universiCes' environments less apracCve for employment and are exactly counterproducCve to their need to maintain and expand their labor supply.”

http://cs.stanford.edu/~eroberts/Curtis-ComputerManpower/

Some possibiliCes … •  Restrict the size of the major

–  ImplicaCons for diversity?

•  Exclude non-‐majors from upper-‐division courses •  Retreat to “the core” – turn over many of our courses to other

departments •  Have enormous class sizes and/or enormous teaching loads •  UClize vast numbers of lecturers •  Have a beer while the students use Coursera

A concrete reason for the concern about diversity •  Introductory courses are parCcularly important “apracCon

waters” for members of under-‐represented groups •  When introductory courses become “weed-‐out” courses, this

disproporConately impacts members of those groups

0%#

5%#

10%#

15%#

20%#

25%#

30%#

35%#

2014#2013#2012#2011#2010#2009#2008#2007#2006#2005#

Computer)Science)%)Female)Degrees)Granted)

UW#CS#%#Female#

0%#

5%#

10%#

15%#

20%#

25%#

30%#

35%#

2014#2013#2012#2011#2010#2009#2008#2007#2006#2005#


UW#CS#%#Female#

US#all#depts.#CS#%#Female#

0%#

5%#

10%#

15%#

20%#

25%#

30%#

35%#

2014#2013#2012#2011#2010#2009#2008#2007#2006#2005#


UW#CS#%#Female#

US#all#depts.#CS#%#Female#

US#Ph.D.>granBng#depts.#CS#%#Female#

0%#

10%#

20%#

30%#

40%#

50%#

60%#

%#of#female#majors# %#of#male#majors#

Propor%on'of'UW'CSE'majors'who'did'not'intend'to'pursue'the'major'when'they'enrolled'in'the'introductory'course'

Our field is at a criCcal juncture!

•  Best pracCces: how to respond? •  What’s different this Cme around, and how can we document it? •  A proacCve agenda: what can we and our disciplinary

organizaCons (CRA, CCC, ACM, NSF, CSTB, …) do?

(The ideas that follow come from audiences at NCWIT, the NSF CISE Advisory Commipee, and the CRA Conference at Snowbird)

ObservaCons

•  Administrators are going to want to address this problem in the least expensive and least permanent way: using TAs, contract instructors, faculty from other departments …

•  Even if CS faculty numbers expand, CS research funding may not

•  If we don’t teach the necessary courses, other units will; they have teaching resources due to the decrease in their own majors, and their students have the need

•  We believe that compuCng is fundamental, but many others remain to be convinced

•  Today, most of the majors in fields such as Psychology, Economics, and History are not intending to “pracCce” in that field. This may become true of CS; it would require changing our approach

ObservaCons (cont’d.)

•  There are now 55 iSchools; we need to ensure complementarity and collaboraCon. (InteresCngly, the undergraduate gender balance of iSchools is not significantly beper than that of CS programs!)

•  The local environment is important – there is no “one size fits all.” Does tuiCon revenue flow to the unit doing the teaching? Are the rewards for majors vs. minors comparable? Etc.

Best pracCces: how to respond?

•  UClize undergraduate TAs – they scale with the demand!

•  Women tend to drop out of compuCng industry careers. Might industry grant 2-‐year teaching sabbaCcals to women?

•  Automate grading in a variety of courses – this would relieve much of the drudgery of being a TA

•  Partner with departments that can teach CS-‐relevant courses, spreading the load while maintaining a modicum of curricular control. (If we’re growing, someone else must be shrinking; there are spare cycles somewhere!)

•  Create a “culture of teaching”: –  Value lecturers/instructors:

reasonable salaries, 12-‐month appointments, mulC-‐year contracts

–  Encourage graduate students to pursue teaching careers – e.g., insCtute “CS EducaCon” reading groups

–  Like Biology, introduce “teaching postdocs” where the postdoc co-‐teaches with his/her faculty mentor – providing teaching cycles, another mentoring opportunity, and career value

Best pracCces: how to respond? (cont’d.)

•  Think carefully about what students don’t need to know, as well as what they do need to know – CS is an ever-‐expanding sphere, and students cannot possibly learn all there is to know –  Can we reduce the Cme required

for students in our major to less than 4 years?

–  Is there an efficient version of Stanford’s “CS+X”?

•  “Take the long view”: –  Expand our view of what

consCtutes CS – suppress our egos regarding what consCtutes “the core.”

–  Grow to Schools or Colleges. (Among other things, this provides a degree of budgetary control – our success will be less likely to be used to subsidize others)

–  CS should grow to be the size of Engineering – it is of at least equal impact. Plan for this!

What’s different this Cme around, and how can we demonstrate it?

•  At many leading insCtuCons, a very large proporCon of students are now choosing to take one or more CS courses (even though this is not required for many majors) – this trend is new

•  Enrollment by non-‐majors in upper-‐division CS courses is rising rapidly – this trend is new

•  Demand for the major is increasing very substanCally – at some bellwether insCtuCons, this demand is dramaCcally exceeding previous highs

•  Computers, computaConal thinking, and computer science are ubiquitous – every field is becoming an informaCon field. Students aren’t blind to this! –  CompuCng today is part of life

and part of popular culture; it’s not just about “compuCng industry jobs”

–  Today’s students are interested in using compuCng, not merely in advancing the core. (Note: this student interest is not reflected the balance of our faculty!)

What’s different this Cme around, and how can we demonstrate it? (cont’d.)

•  Workforce demand in compuCng is dramaCcally greater than in all other fields of STEM combined – BLS data clearly shows this

•  Heads of other units are now asking us to “teach our students computer science” (vs. “teach them programming”)

•  Looking at company parCcipaCon in recruiCng events, CS majors are apracCng the full spectrum of industry – the full cross secCon of the economy

•  The pervasiveness of compuCng in the economy suggests that, in the future, student demand for the major, and industry demand for graduates, will be cyclical with the health of the overall economy, rather than with that of the compuCng industry

•  Today, increasing CS size/strength increases ins(tu(onal strength – it’s an investment that pays broad dividends

A proacCve agenda: what can we and our disciplinary organizaCons (CRA, CCC, ACM, NSF, CSTB, …) do?

•  Obtain broader, more authoritaCve data; what we have at present is “a clear sense” and “many stories” but not a compelling naConal case

•  Understand the reasons for student behavior. In other words, the trends are clear, but the reasons are not. E.g., what are the specific reasons for the dramaCc increases we are seeing in upper-‐division enrollment? Who are these students, and why are they there?

•  Document that upper-‐division growth is something that’s new and different

•  IniCate serious conversaCons with leaders in fields such Math, Biology, and Economics – they have scaled, and they teach many who will not “pracCce directly” in the field. Study how the huge majors do it. (Economics is perhaps the best model – it’s a big major and the faculty are well compensated)

A proacCve agenda: what can we and our disciplinary organizaCons (CRA, CCC, ACM, NSF, CSTB, …) do?

•  Argue, based on structural changes in the naCon’s economy, that we are observing a fundamental shiy in the role of CS

•  Advocate for graduate programs focused on CS educaCon: students care about this, and they need a Ph.D. track and a career track that’s respected

•  Call on funding agencies to support flexible postdocs – research + teaching with a flexible proporCon –  Perhaps create a CIFellows-‐like

postdoc program but with a teaching orientaCon

•  Celebrate the situaCon! We are no longer merely the toolsmiths! We are now the soluCon providers!

VOLUNTEERS to dig into this more fully? •  Send email to Jim Kurose and Ed Lazowska!

Thanks for providing data! •  Colorado School of Mines: Tracy Camp •  Harvard Univ.: Greg Morrisep, Margo Seltzer •  Harvey Mudd College: Ran Libeskind-‐Hadas •  Johns Hopkins Univ.: Greg Hager •  MIT: Saman Amarasinghe, Anantha Chandrakasan, Eric Grimson, Laura Moses, Victoria

Palay, Daniela Rus, Jacob White •  Rochester InsCtute of Technology: Andrew Sears •  Stanford Univ.: Eric Roberts, Claire Stager •  UC Berkeley: David Culler •  Univ. Michigan: H.V. Jagadish •  Univ. Pennsylvania: Sue Davidson •  Univ. Rochester: Henry Kautz •  Univ. Texas: Tiffany Grady, J Moore •  Univ. Utah: Ross Whitaker •  Univ. Washington: Raven Alexander, Crystal Eney, Ed Lazowska, Jen Pesicka •  Wellesley College: Takis Metaxas

http://lazowska.cs.washington.edu/NCWIT.pdf or Snowbird.pdf

responding to the explosion of student interest in computer science

Documents