Roger Schank:

End Run to the Goal Line

By Educom Review Staff

Sequence: Volume 30, Number 1
Release Date: January/February 1995

"Goal-directed" and "just-in-time" learning are the new battle cries for
many education reformists, and the Institute for the Learning Sciences
(ILS) at Northwestern University is at the forefront of the effort to
move those ideas from theory to practice. Under the leadership of Dr.
Roger C. Schank, the ILS has been developing innovative software
programs since its inception in 1989.

"In the long run," Schank told us, "what's most important about what
we're doing is that it's the only real way to change the educational
system. No matter how many people there are with how many good
intentions, change in education is very difficult because there are so
many vested interests against it. But, happily, there is a kind of
Trojan horse that exists in education, and that is a computer, and when
you put something on a computer, people have some sense that it's magic.
Which is fine, because in addition, it turns out that the computer
offers one-on-one instruction that potentially could be magic, and they
can see that the computer is interactive and will go in directions that
kids want to go, and it helps eliminate the lockstep curriculum that
requires that everybody has to be on the same page on the same day. Of
course, the not-so-happy news is that a lot of people use those very
same words to produce some really bad software, but what we're trying to
do is create fundamental change in the way people learn.

"Probably the most important thing is that learning be goal and interest
based. All too often, when we're learning something, we're learning it
only because someone told us we were supposed to learn it. But as you
begin to create new kinds of computer curricula, you offer choices, and
you make it possible for people to follow their own interests and to go
in one direction while others go in a different direction. Having the
ability to follow your own interest and curiosity is a critical need in

"Now inside that framework there are technical issues. For example, I'm
a strong believer in failure-driven learning, which means in essence
that there's no good reason to learn anything if you already know how to
do it--you learn when you don't know how to do it. People from very
young ages learn when they get into trouble, when they can't succeed at
something: that's when they want to know more. We have goals that we try
to achieve; when we can't accomplish them we get frustrated, and that
frustration should lead to more learning, should lead to inquiry. And
yet in schools it frequently leads to somebody sulking in the corner and
feeling bad about himself. In 101 situations there is nobody looking
over your shoulder, and that's an opportunity for a good learning
experience. For me, learning by doing is the issue. Everything in the
school systems has always been learning by memorizing. And that doesn't
really work."

Acknowledging that his work falls within the tradition of John Dewey,
Schank says he is confident that if Dewey were living today he would be
a computer scientist. "The computer offers the possibility of doing real
things and still staying close to academic subjects. For example, we're
building a program called Feed the World, which is a way of teaching
biology. You have a simulated world, and you have to learn various
agricultural facts and political facts and nutritional facts and God
knows what else in order to be able to understand essentially how to
feed the world. That kind of possibility wasn't there without a
computer. The computer helps you address the issue: how do you teach the
more intellectual subjects in terms of learn by doing? They really do
require radical revamping of curricula that only a computer can provide.
Dewey was just before his time."

One of the "issues" (or nonissues, perhaps) for which Schank holds scorn
is the distinction between education and training. "The folks who think
there's a difference between training and education have profoundly
misunderstood education. There is an underlying mind-set on the part of
people in universities and high schools which says that the goals of
education are to produce little miniacademics. So what happens when
you're a student is that you sign up for a psychology course because you
want to understand yourself and understand your friends, but then you
find yourself doing psychological experiments; well, when that happens,
that course has failed you. If all the course teaches is the history of
academic psychology, it has failed.

"College students are consistently frustrated by the notion that they're
supposed to become philosophers rather than being able to understand
philosophical issues within the nature of their own lives. I think
training people to be good thinkers, which should certainly be a goal of
a university education, ought to involve critical inquiry, which is a
learning-by-doing kind of affair that involves writing papers and making
good arguments. Folks who argue that there's a difference between
training and education just don't understand how to train people to
think--and yet that's what higher education ought to be about."

The work Schank is doing at the ILS is much different from the
artificial intelligence research he conducted at Yale, where he was
chief of Yale University's Artificial Intelligence Project from 1974 to
1989. "The work we are doing now is much more applied, in the sense that
we are trying to take the things we learned about case-based reasoning
and natural language processing and put them into programs which are
actually useful for people. So now we build high-quality educational
software. This place has a vague feeling of a software company, because
we build product and deliver--which is something you never did at Yale.

"But I don't think it's a bad thing to deliver product. I think in
some sense computer scientists were doing the world a disservice by
constantly talking about programs they might write someday or that might
work someday without actually doing it. So we have this whole army of
150 people, and 30 are full-time programmers, and 20 are knowledge
indexers who put knowledge into the machine--things you couldn't do in a
university setting typically, though we've been able to do it here.
Still, it isn't a software company, because we don't have to make money;
we can experiment, we can build things that no one's ever built before,
and we're not under the kind of pressure where you have to get things
out the door right now."

And how are Schank and the Institute for the Learning Sciences accepted
by the business and academic communities?

"Well, the business world thinks we're flaky academics anyhow; we're
weird for them, we don't behave as other people do; still, the fact is
that we do deliver things which they think are exciting. As for our
academic colleagues--well, there are some who've come in here and said,
'So, how many publications have you got?' Well, we are still publishing,
but I argue, 'Hey, that's not the right measure right now; right now the
measure is: what are we producing?' I think we've had a reputation of
being a pretty hot and exciting place, and the university is very happy
to have us."

One of the institute's many successes is the ASK system, which builds on
Schank's work in both expert systems and case-based reasoning. Expert
systems are built on a system of rules; case-based reasoning (an
invention of Schank's) is software that searches relevant "cases" (the
way a doctor or lawyer might look for relevant cases) and identifies the
case closest to the problem at hand. "A problem we had with case-based
reasoning," says Schank, "was that we had trouble building the case
base: we couldn't get people to tell us their stories. With the ASK
system we have now built, you can have conversations with people we've
captured on video; we've recorded and indexed all their answers and put
them into an indexing scheme, and we've actually gotten to the point
where we're in a production mode. I think we have the opportunity to
deliver information to people, and that's very exciting to me, though I
don't think that it's the most theoretically interesting system we've
ever built.

"On the theoretical front, we have two systems that I'm very proud of.
We have one that trains you to be a customer service representative in a
water utility and allows you to practice with simulated customers. It
looks and feels like the real thing, and when you make a mistake the
experts of the company pop up and help you out. It's my model of what
training ought to be in a straightforward situation, where there are
'right' answers.

"We also have a different group of programs which I'm excited about
called the GuSS systems--for Guided Social Simulation--which help you
with simulated characters to practice situations in which there isn't
necessarily one right answer. So, for example, we built one for
Ameritech, to tell somebody how to sell Yellow Page ads, and we built
one for Andersen Consulting to teach them how to manage people. Right
now we're working on one for ARPA [Advanced Research Projects Agency] to
teach intelligence officers how to do their job.

"The most important thing here is the development of the tools behind
these things; we're building a tool set, which is driving down the cost
of production. The major problem in educational software today is that
educational software is too damned expensive. It can take a million
dollars to produce a decent course for most people. Well, we've driven
down the cost of production by developing tools that replicate the kind
of educational situations that keep coming up. We are very concerned to
come up with tools that, first, are good enough to allow teachers who
know nothing about computers to build multimedia courses by themselves
and that, second, contain a built-in theory of education so people can't
do bad stuff. And these tools have both."

One thing Schank is adamant about is the importance of learning by
doing. The first thing he asks a potential client for a new course is
whether there's a standardized test at the end, say, some government or
industry test that a student would have to pass. "If there is, I won't
take the work," he says, "because those tests are based on teaching
people by memorizing, and I'm not interested in that. Tests pervert
education. If you have to memorize a lot of short answers or vocabulary
for multiple-choice tests, then you're into learning by memorizing, and
learning by doing won't help. So I wouldn't take that kind of work.

"But if it's really a case of learning by doing--like learning how to be
an underwriter--and we can be concerned with actual performance goals
rather than tests, then what we do is send our people to whatever
schools they have now so as to teach our people how to do this job. And
then we build a situation that looks just like the real job. You'd deal
with the problems of customers coming to deal with you; in dealing with
them, you'd make mistakes; and when you made a mistake, a videotaped
segment of just the right company expert would pop up to correct you. So
we'd have a library of, say, a thousand answers from the company, to
give to the student exactly when they're needed.

"Any domain that's a 'doing' domain is straightforward; we build
something that looks and feels just like the thing that you're going to
do. The complexity comes when we get into schools. When you're dealing
with schools and they want a course in some subject, typically the
course has no performance goals for it. Schools have competence goals,
not performance goals. So what we need to do is restructure the
situation. For example, people from another country came in last week
and said they want us to build a course to teach basic science issues
and biology. And somewhere in the middle of the conversation they said
they are very interested in environmental issues and water issues. So
we're building a course for them on how to build a water system.
Students are faced with a problem, with decisions to make and laws and
rules to apply, and they need to understand the biology to create them,
and we treat the students as though they were training to be head of the
water protection agency.

"I believe in goal-based learning, so to teach calculus I'd have
students build bridges. To teach biology I'd have them make health
decisions or nutritional decisions or who knows what, but I'd always
have students in an active role of having to figure things out on their
own and to come to some conclusion on their own. We translate everything
over to a goal-based scenario--a little playlet, in which you have a
role and are trying to achieve some goal.

"We built a geography program, which allows learners to travel anywhere
they want. One kid goes to New York, one kid goes to Los Angeles, and so
forth. Well, guess what? They don't learn the same stuff! But if you
object to that, you've kind of missed the point! Look, nothing will
change so long as there are same-stuff tests. Where there's a test at
the end which requires that everyone has to have learned the same stuff,
you can't have much innovation. This is true even without computers.
When teachers try to be innovative in classrooms, they often get beaten
down by multiple-choice tests at the end that are administered by some
organization or another.

"That's why you see innovation in schools at the junior high school
level, where you don't have those kinds of requirements, but as soon as
you get to high school it's a lot harder because all of a sudden
standardized testing takes over. To use a football metaphor, I think
that to innovate in the schools you don't run up the middle; you do an
end run. And so if you create an innovative course in Japanese, you're
probably not competing against anybody, but if you've got a course in
French you've got a problem. And I always use the example of third-grade
physics: Nobody is going to object to third-grade physics and say, 'Here
is the test.' So you can be creative and have a lot of fun and learn
some stuff.

"I don't think you should go plodding ahead replacing course for course.
Where innovation is most possible is at the college freshman level,
because faculty tend not to care much about freshmen--especially those
not in your major--or seniors in high school, because no one cares about
the kids after they've passed the major hurdles. So that's where you can
start putting in innovative programs. How about a course in political
science--which they don't have in high school--or psychology--which they
don't have in high school?

"Getting faculty in the universities to help is problematic, because
faculty in general don't like teaching and don't much care about it. So
to get faculty to spend time on a course, you've got to find unusual

Take me to the index