Teaching Computer Science to All
- Articles, Blog

Teaching Computer Science to All


The session today is
about computing for all. And obviously it
has intersections with the last panel
and the panels earlier. And the way I interpret
it is to mean, teach computer science very
broadly across disciplines, across different
stages of education, and across geographies. So well, I guess I forgot
to tell you who I am. It’s not important. Dimitris Bertsimas. I have been a professor
here for 30 plus years, and particularly excited about
this important time at MIT. The last time I think MIT
had a new school or college was, I think, 1951. The Sloane School. So this doesn’t
happen every day. So 80 later is something
to celebrate and reflect. So we have a panel of very
distinguished educators and researchers. I will focus on the education
aspect, given the day. So I will introduce every
panelist separately. Our first panelist is
Professor Anant Agarwal, who has been a professor at MIT
for even longer than I have– not by a lot– and he is the CEO of edX. In learning about
him, he is remarkable. Something I discovered is
Anant taught the very first edX course on circuits
and electronics from IT, which drew 155,000
students from 162 countries. I would say he is
well qualified to talk about geographies and so forth. Anant? How many computer
scientists does it take to change a light bulb? OK. So thank you. I’m really delighted to
have a chat with you today and brainstorm a bit about
how we can increase access to computer science. But what is the problem here? To give you a quick
idea of the problem, we are trying to solve– there are 1.4 million CS-related
jobs available by 2020 and we are producing
a pitiful 400,000. So we are off by a
mere factor of 4. At the same time, we don’t
have place in our classrooms to hold students. The New York Times article
that you see a picture of here talks about how in all the
classes in computer science, people could not get
into those classes. I was talking to my dear
friend Regina Barzilay, and she was telling me that
in her machine learning computer science
class at MIT, there were so many students in
the classroom, overfull, that the police were called in. And they had to then they
had to take the students out because they refused
to leave her classroom. Look at this. This is what is happening. This is not a
third world country where people are fighting
for scraps of rice. These are people. These are people fighting
for bits of knowledge in computer science. And here, the
challenge on campus– our campuses simply cannot teach
the number of people you want to teach. And these numbers
are true everywhere. University of Texas
at Austin just launched a online master’s
degree on edX for $10,000. And they gave us the numbers. 1,300 students applied
for their masters in computer science on campus. They make 100 offers, and
30 people come to campus– 3%. This is insane. When everybody around the world
can go in and type a search up. I don’t need to have
a library in my house. I can go and type search. Digital technology has solved
so many problems for us where we all have ample
access, but we’ve not been able to solve all of
the fundamental accesses of education. And there are not many people
would argue that education is a human right. But then we have this
gatekeeper somehow, physical space in classrooms,
that do not allow us to teach as many people
as we want to teach. So one solution is digital
education– particularly modular digital
education– and let me talk a little bit about it. A digital computer science
learning can scale. So on edX.org–
as an example, edX is a non-profit founded
by Harvard and MIT, where up about 140
institutions– many of them represented here, Georgia
Tech and others, offer courses to people from Cornell
and others from people all over the world. And we have 800 plus
computer science courses and eight million
unique learners from all over the world taking
these computer science courses. And in case you’re
wondering, are they really completing them? Yes. 200– a quarter of a
million students just in the space of
about six years have earned a quarter of a
million certificates in computer science courses. So it is possible to teach
computer science at scale. We can teach at the
high school level. We have a number of
high school courses on edX, a number of them
high school AP courses. Harvard has a course on AP
Computer Science principles. Berkeley has a course on
beauty and joy of computing. Is Hal Abelson here? So Hal Abelson, in MIT
and part time at Google, done some amazing work in
creating the app Inventor, a way to teach computer
science and build computers, build applications for mobile. Trinity College has an amazing
course on edX and Purdue as well. And so we have 1.5 million
unique high school learners enrolled taking these
high school level courses in computer
science on our platform. MIT is doing an amazing job,
but have 90 plus courses on edX with 1.7 million
learners on our platform. And here i’m broadly
construing it as data science, computer
science, and related courses, with 25,000 certificates earned. And Eric Grimson,
who’s sitting here, probably single
handedly contributed to a large fraction of them. Eric and John Guttag teach the
MITx Introduction to Computer Science and Python on edX. And what I want to
stress to you is, not only is it a free
course, and people are earning
certificates, but this is accessible in the ADA sense. You see this learner
here, Bhargav. He’s from IT Madrassa in India. And someone want to guess
what’s unique about him? He completed this MITx Grimson’s
computer science course on edX. And this is the same as
six triple 0 up at MIT. And he completed this course. Anyone want to guess
what’s special about him? He’s blind. So the platform is
WCAG 2.0 complaint. The World Wide Web Consortium
that has the WCAG 2 standard, they themselves have courses
on edX, the W3Cx Front-End Web Developer. And here’s another
student, Larissa, who’s completed their program on edX. Online courses can also
help meet campus demand. I mean, isn’t it
insane that we’re not using digital technology to
solve the problem that involves teaching digital technology? It’s a bit meta-circular. We all remember meta-circular
evaluation and scheme and stuff, and so
we should be using the technology we create
to solve the problems of the same technologies. So Gerogia Tech– a couple
of my Georgia Tech colleagues or friends are here. So what is Georgia Tech doing? The enrollments in
the Introduction to Computer Science course at
Georgia Tech was overflowing. So what they did was,
they allowed the students to take the fully online
course on edX, 100% online, or they can take
the campus course. And our friend Zvi
Galil, the dean there, tells me that 80%
percent of them rate the online course as
better than the campus course. And today, 60% of
Georgia Tech students are self-selecting into
the fully online course. MIT has done the same
thing, although MIT did it with the circuits course– same idea. The circuits course, where half
the students took it online– 100% online, not blended
model, 100% online– and half took it on campus. And this was an experiment. And they even gave the
students the same exam. And the research papers written
by Anne Marshall on the study talked about how the
results from the exams are pretty similar. So online courses
have come of age. Universities like
Georgia Tech and MIT are giving campus
students credit for completely online courses. In addition to access– so far I’ve talked
about access and how we can use these courses to
solve a major access problem. How do more people
learn computer science at the high school level,
at the college level, and also how do we
solve the problem that we face in colleges that
our classrooms are simply not big enough? But again Irina just
told me that she had to cut down the
enrollment in the class to 90 because MIT did not have
a classroom big enough to hold the number of
people she wanted to hold. This is insane, folks. I mean, this is nuts. I mean, can you
imagine real estate? I guess Cambridge real
estate is a different thing, so that’s a real constraint. I know the rents we
are paying at edX. The other important
thing is that there’s another major challenge in the
world, which is this upskilling challenge where various
studies have shown that within the next 10
years, half of today’s jobs will have to change in
one way or the other. People will have to learn
radically new skills or lose their job because of
automation, and AI, and all the things that Regina
and company are doing. It’s all your fault. And so the whole planet
has to upskilling We all have to upskill. And how do we get these
skills out to people? How do we train people? How do they learn? Many of them are
working in companies. Many of them looking for jobs. No way they’re going
to come to a campus to get a master’s degree
or a second bachelors. No way. No way on earth. Nor do they have the time
to learn online and do a full one year master’s
or multi-year programs. We need modular credentials. We need modular approaches
to teach people. And the first picture I showed
you with the LEGO blocks, it has the answer, which is
we need modular education. And at edX, a number
of modular programs– MIT pioneered a new modular
program called MicroMasters. See masters– MicroMasters. Computers– microcomputer. So the MicroMaster is one
year or two year long. Micro Masters are about 25%
of a master’s degree and fully online. And people can take
these, and complete them, and get a campus credit
if they get admission into the university. So you can get a MicroMaster
from MIT, for, example in data science for
about $1,000 today. You can also do
radical, crazy things with the modular programs. You can stack. To my knowledge, it’s
the world’s first stack, multi-institutional degree. So MIT has a MicroMasters
program in Supply Chain Management on edX. So Arizona State– MITx is the number two ranked
program in supply chain. Arizona State is the
number three ranked. We said, hey, why do we need to
create a full master’s degree? Let’s take MIT’s MicroMasters. Let’s just create the
complementary part. MIT has four courses. They build the six courses,
and they launched a masters in supply chain
management on edX. Modular. So education would
become like LEGOs. It will go online. Campuses will have online
learning and campus physical learning
at the same time. And I see a world
in the future where this kind of modular
digital learning can really address
the kind of needs that our planet
faces in computer science and frankly,
other technologies. Thank you. [CLAPPING] Our second panelist is
Professor Regina Barzilay from our Computer
Science Department here. In addition to being a recent
winner of the MacArthur Award for her work of developing
machine learning methods that enable computers to
process and analyze vast amounts of
human language data, she’s a recent recipient
of the Jameson Teaching Award, one of the major prizes
of the Institute for Teaching. Regina? Thank you very much. I learned that there are
actually many other recipients here on the panel. OK. Let me start. So I’m not sure. What am I supposed to be doing? Oh. To do what? Suddenly six slides have
advanced, so watch out. Try it again. OK. So I arrive to MIT in 2003. And when I have term in 2003,
it was not a popular term. And we didn’t have like,
hundreds of students who wanted to get into the classes. And at the time,
machine learning classes were taught all
on graduate level. So there was no question. Teaching non major
machine learning? We rarely had, you know,
majors who were studying it and majors which were
graduate students. So clearly the
situation changed. And if you don’t need to go far,
you can just open a newspaper and you would see that
AI and machine learning does all different
types of miracles across all different
disciplines. And our students
read newspapers. And of course, their interest
in machine learning greatly increased. So around maybe six years
ago, Tony [INAUDIBLE],, is a professor here,
and myself, we said, we actually have to
start teaching this class in the undergraduate level. And that’s how 6036, one of the
very popular classes in machine learning, was born. And you can see for the few
years when I was teaching it, the enrollment steadily
increased in this class. And originally, we thought
that just once a year, and now we’re teaching
it twice a year. So what I want to tell
you, though, there is something else that happened. What we noticed that as we
continue teaching the class, the composition of the
class greatly changed. Originally, this class was
designed for non major computer scientists. And you can see in 2013,
83% of the students were in computer science. If we are looking at
the chart in 2017, you can see that almost half
of the students in the class actually non majors. And they cover almost all the
different majors across MIT. They’re not computer scientists. And this was a great news. [INAUDIBLE] with us who
really can see, with one class we can deliver the goods
for the whole community. However, the part that you
cannot really see on this chart is what happens to non majors
when they take this class. First of all, they
have much higher rates of chance of dropping
from the class, and second the grades are
much lower than our majors. So the question is,
if you still want the students to take these
classes and to benefit, what do you want to do? So one answer which is
really unsatisfactory answer is to say, you know what? We’re going to do
a remedial class. We’re going to take a normal
machine learning class, watered it down, and
make it for non majors so that they can succeed. And you can see now the
landscape of research totally change. Now we have all these
big building blocks. So we have all these different
machine learning packages, which are very high level. So maybe we can just teach
them how to use these packages, and that’s it. So I think it’s
a wrong approach. What we really need
to do is to enable people who are users of machine
learning technology– future users of machine learning
technology– really to think about modeling. What are the problem
formulation strategies? What kind of tools you
can use to solve it? Now let’s just look at the
curriculum of standard machine learning class that
we are teaching, that everybody are teaching
a variation of this class. So what you would see
here, across the lectures that I put on my slide,
the vast majority of it is classification. So for those of you
who are non majors and then take machine
learning classes, is one technology which is
very powerful technology, just one building block. And we’re looking
at different ways to build this block inside. We’re not talking about what
a beautiful things that you could build from these things. How do you need to select
the right type of blocks? We’re just focusing
on one technology. And to me, one of the biggest
surprises working with MIT undergraduates who are majors
who did very well in the class, that even they
don’t understand how to put these blocks together. How to decide which of this
technology can be useful and how to move forward? So the topics that
are really absent, which are really crucial
for using machine learning technology
across disciplines, are kind of core
pieces of technology, like making the
models interpretable. If your models
make the prediction and you’re using it
in health, you’re using in financial
industry, how can you explain what is it doing? And it’s particularly a
problem when we’re thinking about deep learning models. So we don’t teach it. And now the big question, which
is a very useful question, what do you do when you are
training your model on one type of data and
you need to apply it to another type of data? Very, very common scenario
across disciplines. We don’t talk about it. Another one is sparse data. For the vast majority
of applications, we don’t have huge data sets. What tools do we
give to students to address these questions? And of course, there are
all the different kind of scientific paradigms
that people are thinking about and talk about. Science like
causality and others– they are not addressed
in our classes. And even when we are
thinking about evaluation– when we want to make sure
that the student and the user understands how you actually
are using technologies and doing the right thing. We don’t have any material. And there are lots and
lots of omissions that are in our standard curriculum. So what we have done– I stopped teaching 6036, and we
started teaching a new class. I just came from teaching
the class with 90 people that we have in the classroom. It’s called Modeling
with Machine Learning– from Algorithms to Applications. And the idea is really, to
defocus from classification. We do teach a few
classification algorithms. The students understand
what are they doing. But to really, to focus on the
idea, how do you build from it? What do you need to be
understanding and thinking when you are applying and
creating new formulations? And the topics that I just
show you on the previous slide are really kind of
developed in this class. Another point that is important
is the teaching style. What we realized–
that when you’re talking about non
majors, you’re looking at the huge diverse audience. You have people who are maths
majors, physics majors, who are much more
advanced than others in terms of linear
algebra and probability. You have people from
other disciplines. I would name them, but– which maybe need more help. And they approach it
from that other college that was created before
computer science college. [LAUGHING] The one that I was trying to
make you realize– that, you, know teaching the
lecture, teaching the material, the core
techniques, in the class is a bad idea. Because for some,
it’s really easy. For some, it’s really hard. So we recorded technical
lectures and the students can go using MITx. Students can go
listen to lectures. They have exercises,
which they can solve. And then we’re using the
classroom time, really, to make the discussion
and to think about different
application scenarios. And the biggest surprise to
me– that actually the class that we’re teaching– we
advertise it for non majors. 80% of students in this class
are majors of core six majors. But you know what
was a surprise to me? That these majors many
of them took 6036. They know kind of
basic algorithm. I was surprised to the extent
that they cannot understand how to apply them. And why do you select technique
A, versus technique B? So I think that it’s useful
both for majors and non majors. And through this
discussion, we actually kind of learning
how to move forward. And finally, we’re
really focusing on connecting to
other disciplines like chemistry,
biology, linguistics. And what we are expecting
moving forward– that we’re going to
create specific sections for the subdisciplines,
so that the faculty and part of our new
college can help them to do modeling which
is specific to their area of study. Thank you. [CLAPPING] Our third panelist is
Professor Marie desJardins, and she’s the Inaugural Dean of
the College of Organizational Computational Information
Science at Simmons University. She has done extensive research
on artificial intelligence and computer science education. As well, she’s well-known for
her leadership in broadening participation in computing. I know also having
talked to her, she is particularly passionate
about K12 education. And I think she will be talking
to us about this matter now. Yeah. So I’m sorry I wasn’t
here this morning to hear Tony DeRose say that
he was the only one who was– I don’t even know who that is– the only one talking about K12
computer science education, but me too. I think it’s really important. And I think it’s
essential, actually. And I’m going to talk
about why I think that. So I’m not going to talk so much
about the content of computer science or what we should
teach to K12 students, although I think
that’s really, really important to be thinking about. But I want to talk about how
we could get to this place where K12 students
are actually being exposed to computer science
throughout their education. And to do this, as I think
was mentioned earlier today, is very challenging
in the United States because we have a very
distributed educational system. And what that means is
that every single state has to develop its own
solution for computer science in the curriculum. And in some states, every single
school district has to do that. So it’s really,
really a challenge. And it’s really about leading
transformational change, and how you get people to
work together to really change a system that’s
kind of entrenched in a current way of thinking. And why do I think
this is so important? I think this is so
important because if we want to change the
face of computing, if we actually
want to have gender representation,
racial representation, socioeconomic representation
that reflects our society, we have to get different kinds
of people into the field. And the best way to do that
is to expose those people who are not currently coming into
computing to computer science early in a way that lets them
think of themselves as computer scientists. Because they’re not
seeing the role models out there because they don’t exist. So we have to solve this
chicken and egg problem. And I’ve become really
convinced that we have to do this by starting early– as early as elementary school– in changing the way that
young people think about, what does a computer scientist
look like and what did they do? So I just want to
talk about what we did in the state of Maryland
to address this problem. And it’s kind of–
looking back on it, it was a five step procedure. That makes it sound like
we knew what we were doing. But we really
didn’t know what we were doing until
we got to the end and we said, woah, that worked. So I want to talk
about what worked and how we fumbled
our way to something that I think was really
successful that that can be leveraged to
develop similar solutions in other states. So I kind of have
this five step thing. Identify the problem,
gather momentum, garner resources,
solidify infrastructure, and build for the future. And so I’m going to talk about
each of those a little bit with a few lessons
learned along the way. So first, identify the problem. It’s easy to say, oh, we need
more students and K12 computer science education. And that’s a very abstract goal. But to actually
understand how to do that, you have to gather
more data about what the current situation is
in order to address it. Like, what are the
nuances of what’s happening in your local
region in the schools? Who are your teachers? Who are your students? What’s the curriculum? What’s the state? So how did this
work in Maryland? Well, in summer of 2011, I
got some funding from Google through a program called
CS4HS to run a summer workshop for high school teachers. And we were doing
this because we wanted to get more women
into our undergraduate degree program. And we thought, well, where
would we go to get more women? High schools. That’s where the students are,
so let’s go to high schools. Well, who should we
go to in high schools? How about computer
science teachers? Who are the computer
science teachers? Uh. We don’t know. We don’t know. It turns out part
of the reason we don’t know is because computer
science in most schools– at least in Maryland
and most states– is not its own department. Who’s teaching computer science? Math teachers, maybe science
teachers, maybe a gym teacher here and there, social studies. You know, whoever we can find. And especially in 2011,
that was the case. So we brought about a
dozen teachers together at this summer
workshop, and they were so excited to meet
each other that we wanted to keep that momentum going. And that was kind of the
beginning of this whole seven year long process. And we started– to
try to keep it going, to create some
infrastructure– we decided to start a local chapter of
the Maryland of the Computer Science Teachers
Association, which is the national professional
association for computer science K12 teachers. And we’re like,
how do we do that? I don’t know. Let’s pull up the website. We have to do an application. We’ll make an application. So we just did it. We didn’t know what we’re doing. We just thought that might
be a good way to get started. And then a group of us
kind of got together. We’re doing some activities
over the course of that year and decided to try to get
some money to do more. We ended up with
a planning grant from NSF’s Computing for
the 21st Century program, and did a landscape survey. So that got us to
the place where we understood what
we were facing, and we could start to
think about solutions. So that’s the first lesson. Jumping ahead a little more
quickly because I see my time– we used to some of that money to
start bringing people together. So we had a couple
of statewide summits, and we got people talking about
the challenges and the problems and the gaps. And the thing that
I realized was, we don’t know what
we’re trying to do. I mean, we know what it’s like
writing a PhD dissertation. We know we’re trying to
write a PhD dissertation. But what are the steps
along the way to that? You have to break
it down into pieces. And especially when you’re
doing something collaborative, you have to find a
starting point that you can get a lot of momentum behind. And a lot of people
wanted to say, let’s require computer science
for everybody in high school. And that was a really
daunting challenge. We didn’t think we would
get there very soon. So we started talking
about what could we do that everybody in this room– 120 people at the summit– can agree is possible
and important. And the goal that
we came up with was every Maryland
public high school should offer a high quality, rigorous,
college preparatory computer science class. Everybody agreed that that
was a step along the way to the big thing that
we all wanted to do. And so we spent
the next five years focusing on how do we do that? So you have to have a shared
goal to bring people together. And you can’t just
take your goal and make it
everybody else’s goal because that’s not
how people work. So resources– now we’ve got
this goal, what do we do? You get money. So we wrote another grant to
NSF under their CS10K program– Jan Cuny’s program that
was mentioned earlier– to develop an AP Computer
Science principles curriculum and train teachers to teach it. So we now have a curriculum
called CS Matters, which is one of the College Board
endorsed curricula for computer science principles. Now there’s lots out there. And we led a collaborative
curriculum development process. So teachers were
actually working together to write the curriculum,
which I think was a really good way to do it. And we started
training teachers, and doing that work together
created a really strong team. And that’s what kind of let
us launch into the next phase. Along the way, we also
started thinking about, who are our partners? And we developed some
really close contacts at the Maryland State
Department of Education. I will say, we also had people
who were not as supportive. But we found the people
there who were supportive, one of whom eventually became
Interim State Superintendent, and essentially signed an
executive order saying computer science can count for the
state technology education requirement. Which the hard
part to believe is that previously,
computer science didn’t count for the computer
science, technology, education requirement. So there was a tech ed
graduation requirement that didn’t have any
computer science in it. So that was a big barrier. Because it’s like, well,
we’ve already got technology. We don’t need more technology. Well, we do. We need different technology. And we found it was
not going to work to go through the
legislation, which is how that requirement
had been put into place. But we could do this
end run to just add another option for students. Well, the students really
wanted to take computer science because by this time,
computer science was starting to
explode in interest. And so a lot of factors came
together to make this all work. But we had this
grassroots effort, but the reason
that things really came together and
took off in Maryland is because we also
laid down roots. You know, we didn’t just kind
of get people together and talk. But we found the key thought
leaders and the people who controlled resources who
we could get on our side and start to think
about bigger change. So the last step– and this is really exciting
because I left UNBC last year to move to Boston to become this
new Dean at Simmons University, which is a great opportunity. But I was really worried
that when I walked away, everything was going to
fall apart with CS education in Maryland because we don’t
have any infrastructure. We just had some people
who were working together, and I was kind of at
the hub of all of it. And I thought, well, if you
pull out the hub of a wheel, it doesn’t keep
rolling very well. So I spent the last couple
of years really thinking about building infrastructure,
and we established the Maryland Center for Computing
Education, which originally was a cubicle
and a sign on the wall. Like, it had zero
resources but USM agreed we could have a thing. And they wrote a memo
saying, there’s a thing. And then we went to
the state legislature, and we said, well, you
should really give people lots of money for computer
science education and train teachers and all of that. And we think you should
give it to the Center. Because the Center,
that’s its mission. And so, it all kind of
was like smoke and mirrors that coalesced into $5
million in state funding in this fiscal year’s budget
for teacher preparation and advocacy
through the Maryland Center for Computing Education. So that’s how things happen. And the people who
were involved in it are now the leaders of that. So the one last thing– I want to just put a plug in, is
if anybody is in Massachusetts and is interested in getting
involved with a nascent CS for Massachusetts activity,
please reach out to me. I’d be happy to put you
in touch with that group, and I think Massachusetts should
be at the forefront of this. And they’re not. So maybe the MIT Schwarzman
College of Computing can also help with that. So thank you. [CLAPPING] So our fourth panelist is
professor Eric Grimson, who is MIT’s Chancellor
of Academic Achievement and also Professor of
Medical Engineering at MIT. You heard about Eric
and John Guttag’s class from Anant, how
successful he was. Eric has taught more than
10,000 MIT undergraduates and has served as a thesis
supervisor for about 50 MIT PhDs. He is also the recipient of
the Bose Award for Excellence in Teaching in the
School of Engineering, which I understand is the
highest award for teaching at the School of Engineering. Eric? So I’m going to
describe experiences, lessons learned from teaching
MIT’s introductory CS course. It’s a class created
by John Guttag. I gave him a little help. John, Annabel, and I
continue to teach it. I’m here because John’s actually
across campus in lecture right now doing the right
thing, teaching undergraduates. And I’m talking to you. To set the context,
it’s a subject that’s taken typically by
about 900 students a year. An undergraduate class at
MIT is about 1,100 students, so almost all students take it. It’s a class intended for
students who know very little, or have very little or no
prior experience, in computing. An interest link
to a point that I think Jeanette raised
earlier– that number is dropping because we’re
seeing more students coming in prepared. And those students who
come in prepared taken advanced standing exam and
get credit and simply move on. The class started out
as a one term class. We recently split
it into two halves. The first half focuses on
computational thinking, and the second half talks about
data analysis and computational experiments. And I’ll say a little bit
about that in a second. We also provide two
online versions. We have put up two postings of
the course on Open Courseware. One, as it says, in 2008. One in 2016. The first lecture has been
viewed 4.3 million times, which is a frightening thought
that many people have heard my bad jokes. And we taught the
second MIT MOOC. Anant did the first one. We have a course that
came up in 2012 on edX. It’s been offered 17 times and
has had 1.2 million registrants around the world– also a frightening thought. So goals of the course? As I said, it’s intended
for students with little or no prior experience
in computation. We, in fact, actively discourage
students who have experience from taking the class. We don’t want them to
intimidate people in the class. And I’m going to borrow from
two great colleagues before me. One of the goals is really,
while the students do learn a programming language–
in this case, Python– the goal is to focus on
computational thinking. Thank you, Jeanette,
for creating that term. At least, I credit you
with creating that term. Computational thinking,
computational methods, not the specifics
of programming. This is not a programming class. It’s how to problem solve,
how to think a little bit like a computer scientist. And that’s what we
want to do, is change the student’s way of thinking. And then I’m going to borrow a
comment from Maria [INAUDIBLE].. Because in addition to focusing
on algorithmic approaches to problems, we want to
show students examples in other disciplines. So in our problem sets
and our lecture examples, we heavily borrow from biology,
physics, urban planning, economics, whatever. We can let them see
things in context. We think that really
helps them think about what they want to do. So the primary objectives,
the 30,000 foot view, are very simple. We want students
to learn approaches for decomposing problems
into logical steps. We then want them to
learn standard approaches for converting those
decompositions into algorithms. And finally, we want
them then to convert those algorithmic descriptions
into actual programs, especially understanding how to
make efficient implementations. But notice we start with
the high level view of this. How do we get them to
think computationally? You can read this. I’m not going to go through it. We use, basically, the
things you would expect, the standard tools of any good
introductory computer science class. But I’ll highlight that
in addition to abstraction and modularity– which are
in every good intro class– we really talk a little bit
more about recursive thinking, and how you can use that to
design algorithmic solutions in a wide range of areas. We want students, whenever
they see a new problem, to say, how can I break
this down into a simpler version of the same problem? How do I approach
it algorithmically? I want to talk briefly about
the campus version and then the online version. On campus, we mix online methods
with traditional methods. During lecture,
we’ll take stops. And we’ll do what we
call finger exercises, a short little question
that the students do online. We look at the responses. We also assign those finger
exercises during the week, as well as weekly problem sets. And a big factor
for us is to have those problems sets initially
graded automatically by an online tutor. Our experience is that instant
feedback is really valuable. I know we have lots of
computer scientists here. Looking for that missing close
param is totally annoying, and the fact that your code
doesn’t run is irrelevant. Instant feedback helps
students very quickly identify when they don’t
know something, and go back and look at the text. So we use that a lot. But in addition to
that, we also require students to come
to an office hour with an undergraduate
or graduate TA and explain what they did. So they have to get a check off
where they have to talk about, why did they make
those decisions? Where do those
solutions come from? A little bit to catch the
people that are getting too much help from others,
and a little bit to actually get them to explain. One of things I want to do– I want to talk very
briefly– because I still got a few minutes here– about
the young woman in the middle. Because in addition
to them hearing John or I tell bad jokes,
they go to tutorials. And that’s taught by
graduate students. It’s a great thing to do. But the young woman in that
middles image is, I think, a great story. She took this class many years
ago as a freshman, convinced that she hated computer science
because her father was an MIT alum and he was a
senior person at IBM. Absolutely convinced
she was not going to be a computer scientist. Took the class, discovered
she loved computer science, finished a major in
computer science. Became one of our best TAs. We actually have recorded her
on video– which you see here– going through tutorials. And today, she’s
the Chief of Staff to Drew Houston at Dropbox. So you know, you can discover
that computer science is a cool thing to do. In addition to on campus, we
do have an online version. And we offer it through OCW,
which is really a static snapshot of the course,
and more importantly, a dynamic version
on– thank you– the great platform edX. We break the lectures
into eight minute chunks. We separate them with what we
call finger exercises, very quick question that
gives you a sense of, did I understand what
was in that chunk? If not, I go back
and look at it again. The one difference we use
here is, we do not have– no matter how many times
I’ve begged Rafael– the budget to staff this. And so we use what we
call Community TAs. It’s an interesting
experience, which I’m not I’m guessing other
people on edX do. These are people who’ve
taken the course in the past. They volunteer to
serve as online TAs. And they do it purely
for the glory, no money. And they will do
it multiple times. We’ve had Community
TAs who’ve done it eight or 10 different times. And it really helps the
students get a sense of how to ask their questions. I want to give you one
quote from a student who took this class. Sent me a note a
couple of weeks ago. She said quote, “Your courses
really changed my life. I have a degree in
civil engineering, but was having trouble finding
work in that field where I am.” Which happened to be
the state of Utah. “And I ended up as a program
manager for software developers for a few years. I decided I wanted to pick up
development in your courses, where I ended up from a
recommendation I found online. I finished it in April and
with no further education, was hired into a back-end Python
development internship in May, and just converted
that internship into a full-time
position as of Monday. It was largely thanks
to the knowledge I gained in your course,
a real life changer.” Now, I’m sure I’m going to get
nasty messages from people who didn’t like my class, but it
is nice to hear that somebody can take one class and discover
that they’ve got a way to go. So the take home message? We want to expose students
without prior experience to computational thinking. Some will go on to be CS majors. We’re leveling the playing
field and that’s important. But many of them will use
that computational thinking in other domains. And thus, we want
students to learn how to think computationally,
design systems that leverage modularity and abstraction,
use them to complement theoretical explorations
and physical experiments with computational
analysis and exploration. But we also believe
that exposure to computational thinking
provides a different mode of communication. You talk differently when
you’re thinking algorithmically. And as President
Reif has said, we think every student here
should be bilingual– speak computation and something else. And this course,
for many students, is a first step
in that direction. Thanks. [CLAPPING] Our fifth panelist
is Dr. Jim Kurose, who is currently the Assistant
Director in the Computer Information Science and
Engineering at the National Science Foundation, on
leave from the University of Massachusetts at Amherst. He himself also has
multiple awards in teaching. In fact, it was difficult
to find what to say. So he’s a nine time recipient
of the Outstanding Teaching Award from the National
Technical University, and the recipient of
the Outstanding Teacher Award from the College of
Natural Science and Mathematics at the University
of Massachusetts, among many others. Thank you. Thanks very much. I noticed you counted that
introduction against my time. I guess that’s OK. [LAUGHING] It could’ve been much longer. It’s always tough being the
last speaker of the day, but I’m so pleased to be here. And I wanted to start by
congratulating you all. And those congratulations
come not just from me, but everybody in
the CISE directorate at the National
Science Foundation. And when I was running out
last night to catch the plane, France Córdova, our
director knew I was coming. She said, make sure you say
congratulations from me too. So congratulations
from all of us. I mean, what you’re launching
here is really historic. And actually want to come
back to that at the end. So one of the things about
being the last speaker is, a lot of people have said
things that you wanted to say. And in fact, is rob here? Rob had my slide. I saw one of his slides. I’m like, I’ve got
that slide, Rob. But what I want to talk about
are really, three things. And then I want to end up with
some high level comments here. This is probably the only
place I can go and say, there’s an interesting
right hand rule here. But I want to talk
about computer science for all, what we call CUE– computing and
undergraduate education, and broadening
participation in computing. And I’m going to focus a little
bit more on the programmatics from the NSF side,
but Jeanette’s already done some of the
background for that. So I can go through
that pretty quickly. Oh. Sorry. Seems like everybody’s having
trouble with the clicker. OK. Wow. No, no, no, no. Man, it seems like everybody’s
better than me too. Can we try again? OK. All right. I’ll just go home with my tie. I’ll just talk to the slides. OK. Well, this is actually not how
the right hand rule works here. But I’ll start with broadening
participation in computing. We’ve talked about that
some, about equity and access for all. The BPC alliances were actually
launched at the National Science Foundation in 2007. People have mentioned Jan
Cuny’s name multiple times. It turns out that’s
when she came. Peter Freeman. Jeanette Farnam Jahanian, who
was the Assistant Director, between Jeanette and myself. We’ve all supported
this tremendously. I’ll only point out
that also in 2017, we’re launching a pilot
where we’re asking– we want this to be part
of the entire community’s responsibility, broadening
participation in computing. We have a Dear Colleague
letter and a pilot where we’re requiring our PIs
to do something meaningful. And there’s a lot
of background there. But we want to– it’s really
part of the whole community. It’s something that’s been part
of the strategic plan for CISE for now for quite a long time. OK. So I really wanted to start
with Computer Science for All. And I want to go on record
as being the fourth person now to talk about computer
science education and education broadly at the K
through 12 level. So as you’ve heard before, the
idea is to enable all students. This is about
equity and access– all students to have
access to a high quality educational experience
in computer science in K through 12. Developing the knowledge
base, developing the capacity, as Marie said, for
teaching that rigorously. That also includes teacher
professional development. NSF committed in 2016– there’s Randy– so working
with OSTP and the Department of Education and nonprofits,
putting in $120 million over five years to
actually build this. And so I feel like this is going
to be– this is a home run. It’s a home run already. I feel coming in– I joined NSF in 2015, so
I was born on third base as far as this is
concerned, right? That again starting in 2010– I think Marie mentioned CS10K. And so that was a project
that that actually Jan Cuny and others started in
2010, to bring computer science to 10,000 students by 2016. 10,000 teachers. Right. We didn’t quite make it, but
we actually came pretty close. And CS for All sort
of grew out of that. And as it’s been mentioned
before, one of the things– two things actually– in CS for All– computer
science principles and also exploring computer science. CSP is the new– it’s the second AP exam. It does not replace the
programming AP exam. You see the big ideas there. Programming is part of it,
but just a small part of it. it’s all the
computational thinking that we’ve been talking about. it’s about creativity. It’s about the use and
application of computing and computational thinking
and data and algorithms instantiated through programs. OK. So the challenge we faced
was, how do we scale this? And Jeanette gave an
example of school districts. I live in Massachusetts,
in Northampton. We have 404 school
districts in Massachusetts. I spent two years when I was
an assistant professor working with the Northampton
school system to try to get computer
science into– helping a math teacher there,
as it turns out, getting it in. You can’t do that, right? There are just too many
school districts everywhere. How do you scale really good
ideas to make a national– move the needle nationally? And this was done
through the introduction of this new computer
science principles exam. So to me, that was just like– that’s the light
bulb that went off. And in Jan or somebody’s
head, you know, circa 2011 12. OK. So I just want to
show you some data. Everybody seems to
love data, so this is data from the AP
Computer Science exam. And I want to talk
here about diversity. So what you see in the three
graphs as a function of time, the number of female students
who are taking an AP Computer Science exam from–
what does that start?– 2007 to 2018. The number of black and
African-American students in the top right hand corner,
and in the bottom right hand corner, the number of
Hispanic and Latinx students who are taking that. So to me, this is a real home
run, and a really good thing for the community. And people have been working
on this for 10 years. And I think we’ve got
a lot to be proud of. And it’s going on. The little table over there– just in terms of the
raw numbers– going up you see in the graphs, but
seeing the percentages changing in the number of women and
underrepresented minorities taking the CS AP
exam there also. So Jennette actually was kind. She put up a whole bunch
of CISE solicitations. So I’m only going
to put one up here. The one you missed,
Jeanette, because it only came out two weeks ago. And that’s OK. And this is all I want to say. This is about– well,
it’s not just about– let me show you. There’s two quotes from
the solicitation here. This is about rethinking
the role and the positioning of computer science education. There’s something about X+CS. We don’t put CS first anymore. I’m just telling you that
when we talk about that. But really thinking more broadly
about a holistic restructuring of interdisciplinary
degree pathways, right? And so, I think proposals
are due June 9th. So if you’re from an
institution that’s thinking about things like
computing and undergraduate education, I want to say this
was built for you with exactly the kind of things in
mind that we’re talking about here at this workshop. So I’m going to finish up. This is my last slide. This is the slide you saw
Ro put up, although he had Illinois data there. And I have Taulbee
Survey data there for the number of
computer science, computer engineering,
and informatics students who are taking classes. Newly enrolled majors
as a function of time. So I think Marie mentioned
the snowbird meeting. This was a meme picture of
the tsunami of students that were about to come in 2014. And actually, of course,
it’s come to happen. So it’s not news that we see
this increase in students. I guess the other thing– and
Rob didn’t have this on his slide, so I want to
talk about this some– we’re seeing multiple tsunamis. And actually, this gets to Greg
Morrisette’s question, right? So yeah, we’re seeing all
these students come in. But we’re also seeing
students come in now, I think, because of some of the
AP exams that we’ve been doing. And we’re seeing
students come in knowing and having broader interest in
the application of computing. And so the question is, how are
we going to respond to that? And I think that’s going
to be our challenge. And when we think about
the diversity of students, we’ve talked about gender. We’ve talked about
our interests. We should also think about age. We should think
about background. We should think
about preparation. Maybe since I get to have
the very last say here, I’ll go back to Farnam’s
talk at the very beginning. Nothing is going to be more
important over the next 30 or 40, 10, 20, 30, 40
years than education in a technologically
disruptive time. That’s what Farnam was
talking about, right? So it’s our responsibility
as a community to take the lead there. Now, you’re starting
a new college here. You’ve got an
opportunity here, right? Who are your students? Who are you designing
curriculum from? What’s the model? Is it students who are 18 to 22
who are going to be on campus? You’ve got to do that, right? But what about lifelong
learning for students? What about students from
diverse backgrounds? What about students
with diverse interests? What is this
college going to do? I’ll just throw that
out as a question. What is this college
going to do about this? Because there’s a very
small handful of colleges in this country,
or in this world, that people look to for
intellectual leadership in education. And people have mentioned
the thread models, so I’ll give a shout out. I don’t see Charles
here anymore, but I’ll give a shout out to
Georgia Tech for the thread models. I think that really moved the
needle in computer science. And we’ve got the
opportunity– got the need– to do that now. So I want to say
congratulations to all of you. It’s like, when I
had a colleague when we got a big DARPA grant. They said, OK. It’s like, the best
thing when you launch, thinks it’s the best, and then
it’s all the hard work ahead. You got a lot of hard
work ahead and a lot of challenges to face, but
just a tremendous amount of opportunity. So maybe I’ll close
off simply by again, saying congratulations
to all of you. It’s just phenomenal
what you’re doing here, and I think we’re all going
to be watching very carefully. Thanks. [CLAPPING] We have about 15
minutes for questions. I would like to open it up. If I could– I
actually have to leave. Sorry, I have to leave
in about five minutes because it turns out that
I’m giving the opening keynote at 60 tomorrow because
the opening keynote speaker has the flu. So I have to get to the airport. Just at the risk of
being presumptuous, if there were any questions
particularly for me, I’d love to hear those. Or not. That’s fine too. [INAUDIBLE] [LAUGHING] That’s my brother. He’s a show. So if we succeeded in
dramatically expanding access to computer science
in K12, is that going to exacerbate the problem
that others have talked about if you want to
study science, or does it connect with– is it one big picture
of [INAUDIBLE] computer science majors who use computer
science and more joint degrees absorbing that? I mean, I think
the answer is yes. It is going to
exacerbate that problem, and it’s a good problem to have. What Carla said
before, it doesn’t stop the people who
were already going to be coming into the major,
which is continuing to grow. It’s not going to stop them from
coming if other kinds of people come. So that means that the
growth is potentially going to be even
larger, and I think who are the from Illinois? Rob. Yeah, Rob, sorry, who
is talking about that. This has to happen. It has to happen, and so
we have to deal with it and get out ahead of it. And and it’s a moral
imperative that we do this in an equitable way. And I think partnering
across disciplines is exactly the way that we
can potentially do this. Because I know at
my institution, there is pushback from
the humanities worrying about the number
of students who are majoring in computer science. And I don’t want
pushback from them. I want collaboration with them. Because these students who
we’re training, a lot of them are going to go
out and influence what happens in the
humanities in the future. And we have to make sure
we’re working with them and educating students
in a broad way, and thinking about the
implications of the computing that they’re going to be doing. Yes. This is on online
education, as far as the quality of
online education, we get to a point in
which we have confidence that the assessment in terms
of how well the education goes through as far as testing. The student online
is reaching the level of performance expecting
to come one of two ways. How do we convince cooperation? Who is going to
hire these people that the quality is there? This question has to do
with online education and how can we be sure
in computer science that the quality of the
education, and the value of the credential, and the
competency of the students going through it match
those of campus education. I think we’re well past that. A number of top 10 ranked
universities around the world and beyond are routinely
doing fully online learning for campus. Georgia Tech, for example, for
this freshman computer science class is giving the
students a choice. Do the fully online class
on edX from Georgia Tech or do the campus class. 60% of the students do that,
and they get the same number on their transcript. Similarly, UT Austin has
launched a computer science masters on edX. Georgia Tech has a computer
science degree, data analytics degree on edX, a
cybersecurity degree on edX from computer science
and other departments, where the degree
that the students get is identical to
the campus degree. So if you just see a student
running around with a master’s diploma– this piece
of paper from some of these universities– it is
no different from the degree students get on campus. And so I think they’re
well past the point where we have to worry that
the quality of the assessments, the quality of
virtual proctoring, and other techniques. We are well past those issues. From the corporate
standpoint, I think the corporates are clamoring
for more computer science graduates. And not only are they looking
for graduates come out of the degrees, many of them
are funding the students to go get the degrees. They’re also clamoring
for more microcredentials. Many of the corporates are not
able to send off their students for a masters. That want microcredentials. And I think the
challenge is there. How do we get the corporates
to appreciate and recognize your microcredentials,
as opposed to the older form, large degrees. I think that
remains a challenge. But as far if
something’s a degree, there’s no difference in the
piece of paper that they get. The corporate does not know
it’s online or on campus today. So Dimitris, if I
could add to that– partly agree and partly disagree
with my great colleague, and I got Regina in between
to keep him from hitting me, so thank you. [LAUGHING] And a chair. She’s a kickboxer. I know. I’m worried. Look. I agree with Anant
in terms of what you acquire from the
classes you take online. I mean at MIT, we talk
about, our online classes are MIT hard. They’re not any different
than the things we do. It’s really important that
that quality is there. And I really agree
that that acquisition, which is why I think the
MicroMasters works so well. The place where I think many
of us so are still uncertain is, an undergraduate education
is also about other things you learn in other places. It’s on an athletic field. It’s in a drama club. It’s in a dorm room
or, you know, wherever. And that’s the
place where I think universities have
to wrestle with, what is it that that is
adding to the degree? And how do– for online,
the credential is important, but that’s about the knowledge. There’s this other part
of an undergraduate degree that I don’t think
we want to lose, and we’ve got to
think about that. And the reason I
raise it is in online, I think there’s a
lot of progress. But the notion of forming
communities to talk– you do that in a fraternity
or a sorority or a dorm room if you’re on a campus. Online, it’s not yet
quite the same experience. And we have to think
about how to provide that. I would like to continue
the discussion asking all of you the following question. Oh, you wanted to add
something to that? I actually want to ask
Eric question then. Rather than turning the
challenge around about saying, well, what about
off-campus students, what about now the
on-campus students? What about the challenge
of providing them as a brick and mortar place
where students are just there. What is it that we do that makes
it better than consuming it online or sitting in a lecture? Or, you know, is
it active learning? Is it– you mentioned the
co curricular activities. But what about things that
happened in the classroom? Maybe a different classroom. Yeah, so I mean, my reaction
is I think in the classroom, some of it is mentoring. Some of it is, again,
is connections. I mean, when I
teach a big class, I hate it when they’re
talking in class. So I’d rather they
do it outside of it. But I think on campus,
it’s two things. One, it’s being able to
build the connections to have discussions afterwards that
you get by being on campus. The other one– and I
feel guilty saying it, because I don’t mean
it to apply to me– but one of the advantages
of doing it on campus, and one of the advantages
of still having lectures– I mean, I’m mixed on this is,
you get to see role models. I’m not one, but you
get to see role models. You get to see mentors that
you say, oh, that’s cool. I mean, I think when
it comes to diversity, that’s an important part of
who are those role models? And my experience is, it’s
not the same as watching a video of somebody as
actually being in the classroom going, wow, that’s a Nobel
laureate that’s giving that lecture at the school. I think I agree with
most of what Alex said, but I’ll just give
you a statement as to today’s generation
of students is different. My daughter, the
sophomore at MIT? Her biggest superstars,
her biggest role models– I wish it was me– are not physical people. How many of you heard
of the Green brothers? These are spectacular YouTubers. They can put any professor–
they’re spectacular creators. She ended up going for
some major Green brothers conference or
something, just so she could be in a crowd of
20,000 people screaming. It wasn’t a rock star concert. It was educators. And so you can have people
motivate you and inspire you online as much as
in person, I think. The physical reality, I
think, sometimes is not as good as the virtual reality. So you knew it was going to
be a discussion here, right? And Anant and I are
doing this really well. You had experience with both. You had flip classrooms. What was your experience with
using both media, both online as well as the? So I think that it
actually depends how good is the our
experience in classrooms. One of the reasons that
they stop teaching 6036 is because we had like,
400 students in the class. So given this huge class when
you don’t know your students, and you really cannot
have a discussion, I’m not sure that we’re
really delivering the goods. That’s why. And also, I kind of felt that
I am teaching to everybody, but there were people who
were bored in the classroom because they knew it,
and there were people who really couldn’t follow. So I think that the future is
actually in combining the two and finding certain things
that can be done online, especially teaching
very technical stuff where people who need
different, kind of personalized approaches. And then find a way to
break up these big classes, and really give the
value to our students of having the discussions. And I was so happy. I just came teaching
my new class. And it was really fun, and we
had many things in discussion. But I know that next year when
there’s going to be instead of 90, it’s going to be 250,
we cannot have the discussion. So we really need, as MIT
and other institutions where computer science
grows, to think, what is the most
effective way of combining the technology so we do the best
utilization of human resources? So maybe instead of me teaching
26 or 27 lectures, maybe I will teach less lectures. Less are recorded. But this lecture
just will be divided to subgroups where we really
get to know our students. Dimitris, really quick comment. I agree with Regina. And one of the disadvantages
of a big lecture is that I don’t know of
any undergraduate who in front of 600 of
their closest friends are going to say Professor,
I didn’t understand that. So if they miss something,
there’s no reset button. But they’ll do it
when there are 60? Sorry? But they do it
when there are 60? They’ll get closer at 60. But my point about it is, one
of the advantages about flipping the classroom and putting
the lectures online is that a student will go
back and re-listen to it. The other advantage–
quick story. My older son and my
wife took my MOOC, which was a really
bad dinner discussion. Why the hell did you– Can you tell us
what grade they got? Yes, I will separately. [LAUGHING] By the way, my wife is the
Head of the Computer Science Department at
Wellesley, an MIT PhD. And so I don’t know
what the hell she was doing taking this class. You inspired her. The reason I want to
tell you the story was because one of the things I
discovered from both of them– which I think is
interesting– about putting the lectures online. They would– thanks to edX– mostly listen to the
lecture at one and a half speed, which they also
thought was hilarious. You can do that. Until they got to a hard
part of the class, and then they slowed down to 3/4 speed. The point is, you give
the student more control of what do I listen to, how
quickly do I go through it? And I think the flipped
classroom– that’s an interesting opportunity. Then when you have
the discussion, it’s a discussion
in a smaller group. I mean, given that we
don’t have a lot of time and maybe I ask a very
speculative question. So we have talked
about Arizona using portions of MIT online, and
the other, and so forth. Let’s consider 10
years from now. You know, Georgia Tech have the
College of Computing from 1990. MIT is doing it. Other places are doing it. It’s going to be
massively available. So do you see a future in
which all this material in the end of the
day, do we need to have– it’s like calculus. Does the nation need
3000 courses on calculus? Similarly, does the
nation need 3000 courses in something and something,
like x plus computer science, for example, whatever the x is? Do we see a convergence
that in the end, as far as undergraduate
education is concerned, that we converse
in an environment that maybe the vast majority of
education is by few instructors online, and then we use other
aspects that you mentioned? Maybe very speculative but
something I wondered about it. I’m certain you have
opinions, all of you. Maybe we’ll go this way. And I might put it differently. My dream is– and I believe
this is likely to happen– is that education will be
like the slide I put up, will become like LEGOs. Where 10, 20, 30 years from
now, universities often creating stovepipe programs for
just their students will create these
modular programs, set up some microprograms,
MicroMasters, and so on. And for the campus
students, put them online. And various students in various
universities around the world will create this
networked digital economy where they’ll
borrow microprograms from various places– MicroMasters, or others at
an undergraduate level– and they’ll compose
customized degrees. People want to be
multidisciplinary. And this is already
happening today. I’ll give you a quick example. The IT university in
Pakistan, new university. Computer science masters. They wanted to offer
data science masters. They can’t hire
a data scientist. They’re all making
too much money. So what they did was,
they had six courses– six courses taught by the
computer science professors– and access the Data Science
MicroMasters from UC San Diego. And what they did was, they
incorporated four courses, the entire MicroMasters
that’s completely online. And so the campus students
pay campus tuition and they do six courses
on campus, four courses online from UC San Diego. They get a degree from IT
University in Pakistan. I think in the future, I
want you to think LEGOs. When you get a degree,
you will be asking people, where is the degree from? Can I see the provenance
of each of your modules? Where do they come from? And I think people will be
stitching things together– internships, programs. Some people who teach
computer science– Eric is a spectacular teacher. Regina’s spectacular. I would take those programs. And I want to do
calculus from there. I want to learn Latin. I don’t think MIT teaches Latin. Heck, I’d love to take Latin
from Wellesley College. I think the future
is going to be very exciting for the learner. Let’s remember, it’s all about
the learners, not about us, the teachers. Any other different views? James really wants to talk. No, no, no. You have the floor, James. Really quickly. All right. I agree a lot with
Anant’s vision. But I think one of the
great opportunities is to combine two things. Somebody who’s a great
lecturer on some topic– Samuelson teaching economics. I would love to take
a class with Paul. I never had a chance to do that. Pick your favorite person. But then locally, to let the
local instructor amplify that. So you watch the lecture from
the really talented instructor, and on campus then
you have the dialogue. And that nicely blends
the more interactive piece with somebody who’s
a great lecturer James? I was going to say exactly
what Eric just said. And actually, I was going to
use Hennessy and Patterson as examples, rather than
Samuel, since maybe we can all relate to that. But I think that it’s
absolutely the case. And I think of it
actually more– I think of them as disembodied
from universities, right? Think about really great
textbook authors and people who think about how to
thread common themes through a series of ideas and
hopefully that will resonate. And there’ll be needs for
multiple ways of approaching that, but I think that people
who spend the time to think deeply about it, who
have a particular style, who are good writers– because I
still think we’re going to have text– who are really good and
inspirational speakers. I’ve never seen– Feynman in physics, right? I mean, we now have the
technology to do that. The question I was going to
ask Regina is if you know– I wonder not too
far down the line that we’re going to be embedding
those kinds of materials with, let’s say, smarter systems
that have models of learners and try to understand
better what learners are learning and not learning. And so, you, know the Jill stuff
that’s going on at Georgia Tech right now is just basically
simple question and answering. But once we have software that
now has models of learning and understands how individuals
are learning and looking at how they’re going
through material, I think combining
that is going to be great for online students. And then again, the
challenge comes back. What are we going to do on
brick and mortar campuses? You’ve started to
touch a little bit, but I want to bring that
last discussion back to the perspective that
James ended up with, which is the diversification
and equity issue. Except for Anant, nobody–
and this somewhat is addressed at Eric’s– I mean, I’ve been fortunate
to have been attached to several universities
and colleges with wonderful
on-campus experiences. But the truth is that
now 50% nationally of our undergraduates are
not traditional students. 50% are not four to
six years, and not finishing when they’re
something like 23 to 25. So how do all these
different models address that part of the
pathways, not pipeline issue? And one other little
piece of that– I’m surprised they haven’t
heard anybody mention anywhere, boot camps. I think they’re mostly– I’m not an expert,
but I assume they’re mostly effective for
coding and not really for the computational thinking. But how does that
stuff all fit in? Anybody volunteering? James? I might start off by saying
just in terms of student backgrounds, I heard Eric say
that the online courses are MIT hard, right? And honestly not everybody can
take MIT hard courses, right? And if you’re going to teach
MIT hard courses, and that’s going to be the folks
that you address, great. that’s really important. So there is need for lots
of other institutions to be addressing
those kinds of things. I mean, exactly the diversity
issue that you brought up. So I don’t see that
it’s just going to be some number
that we can count on, a couple of hands and feet. That this is going to be
the number of professors, numbers of
universities, that are going to thrive because we do
have such a diverse population. But then it’s up to the
individual brick and mortar campuses to now define how
they’re going to do that and who their students
are going to be. And I would just add– I didn’t mean to come
across as arrogant, but MIT knows how to
teach MIT kinds of people. I know how to
teach MIT students. I’m not skilled at
teaching K12 students. We need partners to create
parts of that pipeline. But I’d just simply give you
a number– and my apologies for using it– but I’ve
had 1.2 million people show up for my class. I’ll occasionally get
stopped in airports, which is a really weird experience. But of that 1.2 million,
the youngest student that successfully completed
my class, I think, was 10. And the oldest was in his 90s. And so there is a
whole range of learners that really want to do this. I’m not perfect for
everybody, but we need partners to
think about filling in the other pieces of this. You know Wall Street talks about
how many millions you have. As educators, we talk about
how many millions we have. Different millions. I would love to go
for further, but given that there are other
events, I would like to end the discussion. Right now I would like
to thank our panelists, and there is continuation later.

About James Carlton

Read All Posts By James Carlton

15 thoughts on “Teaching Computer Science to All

  1. I’m convinced that this panel discussion will be to education as Feynman’s plenty of room at the bottom was to nano-science. A pleasure to hear such visionaries

  2. Do you teach biology or physics to all? Then what's up with the obsession to teach people computer science? Let us do our thing without that much competition!

  3. Dear MIT, please reply to the message of the Science show (YouTube channel), they'll make a Russian dubbing of your videos. the author of this channel sent you an e-mail with the indicated offer.

Leave a Reply

Your email address will not be published. Required fields are marked *