ACM Logo  An ACM Publication  |  CONTRIBUTE  |  FOLLOW    

What Can Be Taught: Part I

By Roger C. Schank / September 2009

Print Email
Comments (1) Instapaper

Not everything we would like to teach can be taught. Similarly, not everything we would like to learn can be learned, especially if we are taking the wrong approach to learning. In a previous column, I discussed things that can't be taught. Here I discuss what can be taught.

In this two-part article, I discuss the kinds of thing we can learn. I consider how we can best approach learning by listing 16 types of learning. There may be more, but those 16 will at least cover enough ground to describe how human learning looks. The types of learning are divided into four groups: 1) conscious processes, which I will cover here in Part I, 2) subconscious processes, 3) analytic processes, and 4) mixed processes (nos. 2, 3, and 4 are covered in the forthcoming Part II).

Notice first that all the types of learning are types of processes. All processes require practice in order to master them. You cannot master a process without practicing it again and again. Feedback and coaching help.

One problem in such a discussion is that we are used to, (because we went to school) thinking about what needs to be learned in terms of subjects: literature, algebra, biology, political science, and so on. We think this way because school was originally organized by academics who specialized in these subject areas. They set up the lower schools up on the basis of their areas of expertise.

When I was working in AI, I began to realize that what I needed to teach the computer to do in order for it be smart was a far cry from what people thought it needed to be taught.

People assume we needed to tell the computer facts about the world-similar to the kinds of information we typically believe children should learn in school-and that these facts would make the machine smart. But what computers lack is intelligent capabilities, not information.

It's simple enough to fill a machine with information, but when you're done, the machine is only able to tell you what you told it. If that were a child, we'd say he had brain damage.

Intelligence and the learning required to create useful knowledge are capabilities. If we wish to teach people, it's important to ask what capabilities we want them to have when we are done-not what we want them to know.

Conscious Processes
1. Prediction: Making a prediction about the outcome of actions.
Making a prediction is a kind of experiential learning about everyday behavior in its most common form. It includes learning about how to travel, eat, and get a date. In its complex form, it is how one learns to be a battlefield commander or a horse race handicapper. One learns how to make predictions through experience by trial and error.

quote from article

The cognitive issue is building up a large case base and indexing it according to expectation failures as described in Dynamic Memory.

We learn when predications fail. When they succeed, we fail to care about them because most of the predictions we make are uninteresting: "I predict the room I just left will look the same when I return."

Learning to predict what will happen next requires repeated practice in each domain of knowledge. There is some transfer across domains, but not that much. (Learning to buy an airplane ticket is somewhat related paying the bill in a restaurant, but not that much. You might use a credit card in each, for example.)

2. Judgment: Making an objective judgment.
There are two forms of judgments, both involving decisions based upon data.

In the first, there is no right answer, such as deciding if you prefer Baskin Robbins or Ben & Jerry's ice cream. We make judgments and record them for use later. We find ways to express our judgments: "Ben & Jerry's is too sweet," for example. We learn what we like by trying things out. A sommelier learns about wine by drinking it and recording his reactions and thoughts so he can compare his notes about one wine to a different wine later on.

The second form is reasoning based on evidence. Judges learn in this way, as do psychiatrists and businesspeople. They collect evidence, form a judgment, and later they may get to see if their judgment is correct. When asked, they can state clear reasons why they decided the way they did. The sommelier can give reasons as well, but the evidence for taste is not really all that objective. (Of course, the evidence may be found after the judgment is made. People are not always entirely rational.)

To learn to make objective judgments, one needs constant feedback either from a teacher, or colleague, or from reality. One needs to think about what was decided and why. People who are good at this are good at it because they have analyzed their successes and failures and can articulate their reasoning. Learning requires repeated practice.

3. Modeling: Building a conscious model of a process.
We need to learn how things work. A citizen knows, presumably, how the electoral process works. Someone looking for venture capital should know how fundraising works.

Processes need to be learned for people to effectively participate in them and propose changes to them. Building a conscious model of a process matters a great deal if you want to make the process work for you. If you want to get into college, you need to understand the application and admittance process. This cannot be learned from experience in a serious way because an college applicant does not actually experience the entire process. Having it explained may not work that well either because hearing an oral explanation does not create a complete understanding of a process. Designing it, modifying it, and participating in simulations of it work much better as learning methods.

4. Experimentation: Experimentation and re-planning based on success and failure.
This is probably the most important learning process we engage in while living our lives. We make life decisions and we need to know when we need to change something.

There are big decisions, like getting married or deciding how to raise a child or whether to change jobs, and little decisions, such as changing your eating or sleeping habits.

We make decisions on the basis of what has worked before and what has failed to work. But we tend to make life decisions without much knowledge. We don't know how our bodies work all that well, and we don't really know how the world works or what it has in store for us.

Thinking about these issues and learning from failure is a pressing need all through life. Learning to analyze what has worked what has not and why is part of living a rational life. These things are learned by living and talking about our experiences-hearing stories from others' as well as hearing our own stories as we construct them-thus creating a database of stories that we can rely upon later.

We can learn about life through reading, watching movies, and other media, too. We like stories in all these forms precisely because they focus on life issues. Most conversation depends upon story exchange. The more emotional a story is, the more likely it is to be remembered. The cognitive tasks are story creation, comparison, indexing, and modification.

5. Describing: Creating and using conscious descriptions of situations to identify faults to be fixed.
When problems exist in any situation we need to be able to describe and analyze those problems. We need to be able to describe them in order to get help from people who may know more about the situation than we do. We learn to focus on critical issues.

In order to do this, we also need to be able to analyze these situations to see what was supposed to happen and why it isn't happening. Consultants who fix failing businesses do this sort of thing all the time, as do doctors when consulting on difficult cases. Learning to create a careful description of a situation is a skill which only be learned through practice.

6. Managing: Managing operations using a model of processes and handling real time issues; case based planning.
There is a big difference between learning how a process works and managing it. As we gain more responsibility, we tend to have to learn to manage the processes that we are part of. We may become managers of groups we belong to or we may want to start up our own processes. Either way we need to not only know why the process worked the way it did when we arrive, we also need to know how it improve the process. This means building up a series of cases (indexed in terms of their role in the process) about faults in a process and known (or invented) solutions to rely upon when suggesting changes. The cognitive strategy here is called case-based planning.

Part II
In Part II of this article, which will be published on in the coming days, I will continue discussing the 16 types of learning by looking at the three remaining processes: subconscious, 3) analytic, and 4) mixed processes.

About the Author
Roger C. Schank is one of the world's leading researchers in AI, learning theory, cognitive science, and the building of virtual learning environments. He is President and CEO of Socratic Arts, a company whose goal is to design and implement low-cost story-based learning by doing curricula in schools, universities, and corporations.


  • Wed, 28 Apr 2010
    Post by Terrence Gargiulo

    Thank you for rich article. I believe we make too much of Beginning, Middle, End forms of story telling. Of course most stories have these elements...and when they do not we usually build some story scaffolding that gives the appearance of clean BME's. We require these handles to working with the story to generate sense making and attach other stories, and stimulate conversation. I see story as more fractal in nature...a beginning is a trigger - the middle - are tributaries ripe for exploring and branching into lots of different directions, and end - are like tunnels - our perception is narrowed through the middle of an hour glass where our imaginations are awakened to encounter or consider new ideas, insights, constructs,etc... in other words ends - are tunnels leading to a door of new triggers that lead to new stories. So the super-imposed - necessity/comfort of BME coexists with a sort of sub-atomic layer of stories where triggers, tributaries, and tunnels.