Step 9: prerequisites, or, ya gotta start somewhere

This entry is part 16 of 21 in the series Ten Steps to Complex Learning (the book).

Step 9 in Ten Steps to Complex Learning is “analyze prequisite knowledge.”  As with Steps 7 and 8, van Merriënboer and Kirschner are talking about procedural information, the knowledge that you apply in the same way to different problems.

Procedural Information

(You can click the image to see a chart with all four components and all ten steps.)

The prerequisites are the things you have to know in order to perform these recurrent skills.  vM&K argue that you have to analyze the cognitive rules (Step 8) in order to find out what the prerequisites might be.

Facts: three levels down

Not all concepts are on the same level.Way back in Step 6 (analyze mental models), Ten Steps discussed domain models for non-recurrent skills (the skills you apply differently to different problems).  One way to think of a domain model is that it’s a high-level abstraction.  As you move closer to the job, you can discern elements like conceptual, structural, and causal or functional models; moving closer still, you get to individual facts.

When you’re talking about recurrent skills, the middle and bottom levels are more relevant to the algorithmic nature of procedural information.  As with cognitive rules, this is a step you might be able to avoid if the rule-based information already exists in manuals, traing programs, and similar forms.

For that middle level, vM&K talk about concepts, plans, and principles. Concepts are descriptions and classifications, like the concept of a field in data entry, or the concept of offsides in soccer.

Plans (another peculiar vM&K term) “relate concepts to each other in space…or in time.”  What I think they mean by plan is a kind of map or arrangement, like an electrical circuit or a standardized format.

One of their examples is how you present the mean and the standard deviation in a research paper: M=x.xx; SD=y.yy.  The idea is that as you go about writing your paper, that arrangement is the plan for that particular item.

Plans can also be scripts, which I take to mean small recipes for accomplishing some task.  The order of sections for your research paper is a kind of script.

Finally, principles related concepts to one another in cause-effect or as part of an overall natural process.  (The difference is that a natural process is continuous and doesn’t have a fixed start point, like the evaporation/condensation cycle.)

vM&K say that learning a principle helps a person move from rote memory to understanding.  If you learn principles for performing subtraction, for example, you don’t have to memorize how to subtract every number from every other number.

Facts and features

Ten Steps takes a side trip into logic, or philosophy, or something.  The chapter suggests that one way to identify a concept is to “list all the facts that apply to its instances” in a “feature list.”  These will be “propositions” with a “predicate” and at least one “argument.”

For example, in word processing, a column is elongated.  “Column” is the argument (the subject), and “elongated” is the predicate (the relationship).

No, I didn’t quite get it, either.  In another example:

A text processor [word processor] may construct columns using the table function.  This is a proposition with three arguments (text processor, which is the subject; column, which is the object; and table function, which is the tool connected with the predicate “construct”).

vM&K say that these propositions are the smallest building blocks of cognition; “there are no facts that enable the learning of other facts.”

Let’s see–let’s take the fact that Paris is the capital of France.  As the authors say, in a sense a fact has meaningless, arbitrary links.  You could say “Paris is the capital of France” without knowing anything about Paris, capitals, or France.  And even if you do know about Paris and France, you’re not necessarily better able to learn other facts about Paris or about France.

Facts and concepts

Concrete concepts may require physical models like schematic diagrams, exploded diagrams, and similar tools to help learners develop better mental models.  Along with the concept of an electrical resistor (opposes the flow of electricity), examples of its physical form can help the learner.

Let me explain why this is intuitively obvious...There’s actually an interplay here, and this may be part of that confusing conceptual-model business.  Not all resistors use color-coded bands to indicate the amount of resistance.

As you’ve seen with conceptual rules, vM&K say that prerequisites should be aimed at the entry behavior of the target learners.   You’d think it’d be easy to go too far–think of computer-application training that insists people know what radio buttons are, not just know how to use them on the job.  The authors are also concerned that it’s all too common to overestimate the amount of prerequisite knowledge that learners already have–the familiar curse of knowledge.

Design implications

As with other types of knowledge, learning about prerequisites can benefit from an analysis of misconceptions.  Think of language barriers and George Bernard Shaw’s observation that Americans and Britons are two people separated by a common language.

In the U.K., a “scheme” is a plan; the British government is proposing a scheme for regulating asset-backed securities.  In the U.S., the word “scheme” almost always has a connotation of trickiness or outright deceit, enshrined in “Ponzi scheme.”

The Ten Steps consistently recommends presenting prerequisite information just-in-time, and on a just-enough basis.  In addition, by focusing attention on misconceptions, well-designed materials can help people learn.  I haven’t owned a Volkswagen since I was an undergraduate, but still recall John Muir’s manual for keeping my Beetle alive.  Muir consistently put “front is front” into instructions for dealing with engine problems.

Beetle engines were in the rear of the car.  Muir learned as he helped people work on their cars that they often shared a misconception: “front” to them meant “closer to the outside.”  If he said, “change the front plugs,” they’d start changing the ones closest to the back of the car.

So, “front is front” was a prerequisite piece of knowledge: in this manual, when you’re working on the engine, “front” means “the front of the entire car.”

Other techniques to help learners master prerequisite knowledge:

  • Slower fading–in other words, maintain a relatively high level of support during the early learning tasks.
  • Multiple representations.  You might provide both a verbal or textual description and a diagram, photo, or illustration.
  • Compare/contrast.  As a design element, vM&K urge you to have learners practice applying the new skills.  Working with effective and ineffective concepts helps them better recognize the difference when it matters.

That’s it for Step 9.  Next time, the tenth step (design part-task practice).

CC-licensed images:  not-quite-level steps by R.B. Boyer;
normal distribution with standard deviations from Wikipedia;
automatic flusher sign by tico24.

Step 10: part-task practice (getting better at getting faster)

This entry is part 17 of 21 in the series Ten Steps to Complex Learning (the book).

In Ten Steps to Complex Learning, van Merriënboer and Kirschner say that “part-task practice is often pointless.”  That’s because Ten Steps emphasizes whole-task learning, both for the recurrent and non-recurrent aspects of complex skills.  Part-task practice, though, is one of the four main components.

So, why this Step 10?  Because some of the time, it makes sense to practice part of the complex skills–specifically, certain recurrent aspects that demand a high level of automaticity.  You can think of this as an exception to the whole-task approach.  The main idea is to make the application of those skills instinctive.

The right type for practicingIn my own case, I learned to type when I was 11; the recurrent aspects are now  so automatic that looking at a keyboard slows me down.  I also can’t tell you which fingers I’d use for a given word unless I type it on an imaginary keyboard.

Notice: typing is the recurrent skill; writing is non-recurrent.  What that means is that as I compose this post, I’m doing some things I’ve never done before.  I’m conceiving ideas and putting them into these particular words for the first time (hence, non-recurrent).  But the mechanics of typing are always the same (hence, recurrent).

Recurrent skills are ones that you apply the same way to each task that calls for them.  You can see the benefit of automaticity that comes from practicing musical scales, doing math problems, or completing drills for touch typing.

Some other reasons for part-task practice:

  • Make it easier to perform other skills.  vM&K give the example of learning the order of the alphabet so as to enable the use of dictionaries and other alphabetical tools.
  • Enable simultaneous performance with other skills.  Air-traffic controllers need to recognize dangerous situations while tracking new aircraft and communicating with pilots.
  • Build mastery for critical skills.  Some tasks demand automaticity but can’t reasonably be practiced in whole-task settings: emergency shut-down of a reactor, responding to cardiac arrest, controlling a fire at sea.

Perfect for practice

I once hear Jim Fuller say that practice doesn’t make perfect; it just makes permanent.  He was talking about practice without guidance or feedback. vM&K would agree.  Corrective feedback is an essential part of learning procedural tasks.

Part-task practice is straightforward, since by definition you’re practicing a procedural skill or applying if-then rules.  You’re not performing that practice in a high-fidelity environment, though.  You (usually) don’t practice your piano scales in a concert hall with an audience; you don’t do batting practice in the middle of a game.

This chapter talks about “produce items,” as in “produce the answer to this problem.”  That’s the default form of practice, though vM&K mention some others to use in certain situations, such as when the task is long, complex, or prone to errors.

  • Recognize items ask the learn to select the correct procedure from two or more choices.  (“Which of these shows the right steps for long division?”)
  • Edit items ask the learner to correct a solution by identifying incorrect steps or solutions and supplying the correct ones.  This is a good method for dealing with typical errors.

The three approaches can appear in a fading strategy: begin with recognize items, move to edit items, and end with produce items.

Okay, Vince, so what should you do next?Because practice involves procedures that always yield the right solution–there’s not a lot of discussion regarding long division–there’s no need for problem-solving guidance.  At times, however, it’s helpful to constrain the performance situation to help prevent ineffective behavior.  This is the training-wheels approach.  You can see examples like hiding multiple toolbars as someone learns the basics of a word processor, or starting sky-diving practice by jumping while attached to an instructor.

Divergence of practice items means that you’re covering all situations to which the procedure or if-then rules apply.  For typing, that might mean drills that include upper-case letters, numbers, and punctuation marks.

How to sequence practice items

Back in the discussion of Step 2 (sequencing learning tasks), I tried to figure out what vM&K meant by “part-task sequencing.”  When the whole task is too complex for a learner to practice, you design learning around logical parts of the whole task (e.g., you work on documenting on-the-job performance, then on coaching an employee to improve performance).

Those are non-recurrent skills (you handle things differently for different employees), but this type of sequencing can apply to part-task practice as well.  One significant difference: usually for learning tasks, you use backward chaining.  Forward-chaining makes more sense for part-task practice. “The performance of each step, or the applicaiton of each rule…creates the conditions that prompt the next step or action.”

One technique is segmentation: you break a procedure into distinct parts.  If the whole task is repairing a flat tire, you might practice removing the tire, then repairing it, then remounting the tire.  A good technique when the performance steps are essentially linear.

Simplification makes sense for tasks with steps and decisions.  To learn subtraction, for example, you first practice problems that don’t require borrowing.  When that’s mastered, you work on simple borrowing–say, only one case of borrowing per problem–and finally on problems requiring multiple borrowing.

With fractionation, you break a procedure down into functional parts.  In typing, for example, practice problems involve only the index fingers, then the middle fingers, and so on.

* * *

This is another long chapter, so I’ll stop the post here.  Next time: part-task practice and procedural information; overlearning.

By the way, although this is the tenth step, it’s not the last chapter of the book.  Chapter 14 covers use of media, while 15 deals with self-directed learning.  There’s a short “closing remarks” chapter as well.  I’ve only skimmed these, so can’t say yet whether I’ll make a separate post for each one.

Keyboard image from Microsoft’s clip art.
CC-licensed skydiving photo by Dawvon.

Media’s role in complex learning

This entry is part 18 of 21 in the series Ten Steps to Complex Learning (the book).

Chapter 14 of Ten Steps to Complex Learning deals with the use of media.  Van Merriënboer and Kirscher make clear that they’re discussing the preliminary selection of media; media choices change as the design process continues.

They related media to the four main components of the Ten Steps; those components in turn relate to four key learning processes.  I’ve made this chart based on one of theirs:

ten_steps_media500

What I especially like is the way the chart summarizes the underpinnings of the Ten Steps. For example, when you’re dealing with non-recurrent tasks (and if you’ve read any of the previous posts in this series, you know that means “things you do differently each time you confromt a new task), you want to help the learner build cognitive strategies and mental models (see the overview for Step 4).

A case where training IS learning...

Keep it real?  Not always.

Because the Ten Steps so strongly advocate whole tasks, it’s no surprise that the ideal media to support induction would include real or simulated task environments.  You might use less elaborate simulations in the early stages of learning; novices are sometimes hindered by over-rich environments.

Other advantages of simulations over actual environments:

  • Control of sequence: the designer and the learner aren’t restricted to whatever real-life situation happens to come along.
  • Learner support: unlike the real world, a simulation can provide both supportive and procedural information.
  • Safety: simulations reduce or eliminate dangers to the learner (or to others, like coworkers who might suffer from a beginner’s mistakes)
  • Control of time: a simulation can speed up or slow down processes to help the learner attend to detail or see the consequence of his actions.

Some media make it easier to provide dynamic task selection–proving a unique set of problems to each learner.  This aids adaptive training, which in turn tends to help learners do better in training and transfer skills more readily to the job.

How you did and what to do

This dynamic selection requires continuous assessment against objectives and criteria.  vM&K point out, though, that assessment is inadequate for diagnosis.  Knowing that someone can’t do X doesn’t tell you why, nor what to do about it.

For routine tasks (right, the recurrent ones), it’s relatively to analyze errors and give corrective feedback.  These tasks have correct procedures or interpretations, and often what I think of as expected wrong answers.  In the problem 3 ( 2x + 5 ) = 9 , if your answer is x = 2/3, you almost certainly multiplied 2x, but not 5, by 3.

It’s still hard for computer-based systems to diagnose well when it comes to errors in non-recurrent tasks (problem solving, reasoning, and so on).

[T]he teacher or instructor will typically provide this type of cognitive feedback, or the learners will be invited to critically compare and contrast their own problem-solving and reasoning processes with those of others–including expert task performers.

Along the same lines, learners could be involved in the assessment of their own performance…or the performance of other learners….

Beyond the learning task

The Ten Steps sees a real or simulated task environment as the primary medium.  After that, secondary media support the other three components of vM&K’s learning blueprint.

For non-recurrent tasks, supportive information helps the learner connect new information to what she already knows.  Media that allows her to interact with and explore models and examples encourages elaboration.  In this view, a simulation of a conceptual domain is not to practice a skill–for that, the simulation would involve whole tasks.  Rather, simulations or case studies related to the domain help the learner construct and improve her mental models of that domain.

Hypermedia allows the learner to move from one informational element to another, in a connected way, and thus explore relationships in that field.  The Ten Steps offers three principles to help stimulate deep processing:

  • The redundancy principle says that presenting redundant information typically has a negative effect on learning.
  • The self-explanation principle holds that learners benefit from trying to connect new items to each other and to existing knowledge.  Learning tasks can prompt them to, for example, identify underlying principles.
  • The self-pacing principle says that giving learners control over the pace of a presentation can facilitate elaboration and deep processing.  The learner benefits from being able to pause, rewind,  or replayaudio or video, for example.  This allows them to pause and reflect on the new information

For recurrent tasks, vM&K say that procedural information must be available when it’s needed, and in small enough units to be helpful.  Step 7 touched on the problem of split attention.  Other principles for online help include signaling (e.g., explaining a process step-by-step; highlighting parts in an engine repair) and modality (using two modes for presenting information, such as an audio explanation of a detailed blueprint.

Finally, part-task practice benefits from media that support small-step, drill-and-practice techniques that lead to automaticity.

CC-licensed image of BVE train simulator by Alan_D.

You? Auto? Practice.

This entry is part 18 of 21 in the series Ten Steps to Complex Learning (the book).

My previous post in this series ended by discussing various techniques for sequencing part-task practice.  Ordinarily, Ten Steps to Complex Learning recommends that you design learning tasks so they have a “high contextual interference.”  In other words, within a class of learning tasks, the individual tasks differ from each other as much as they do in the real world.  The idea is that this helps the learn build mental models and cognitive strategies.

But part-task practice doesn’t need those models and strategies.  So the practice items in a given set resemble each other quite a bit–there’s low contextual interference.  According to van Merriënboer and Kirschner, this helps the learner automate particular recurrent skill being practiced.

Presenting Procedures for Part-task Practice

I'm getting the point about part-task practice.Step 7 (design procedural information) advocates a whole-task approach toward procedural information.  Part-task practice is by definition procedural, and so much of the advice from Step 7 applies, even though we’re now talking about specific procedural skills for which additional practice makes sense.

Take the complex skill of becoming a chef.  Plenty of non-recurrent skills there, but as Jacques Pépin knows, lots of recurrent ones (procedural skills) as well.

For novice chefs, then, you might demonstrate knife-sharpening or cutting vegetables into julienne strips in a whole-task context.  To  practice how to sharpen or julienne, though, you’d then demonstrate the specific procedure in isolation.

Other instructional methods for difficult aspects of a recurrent task:

  • Subgoaling: have the learner describe the goal (the result) reached by a particular process.  (“Now that you’ve sliced the onions…?”)
  • Focusing attention: highlight steps that are difficult or dangerous.  (E.g., in a graphic demo, color-code or otherwise emphasize the dangerous steps.)
  • Multiple representations: use more than one format to present difficult procedures and rules.  (For example, a real-life demonstration by an expert along with a simulation that the learner can control.)
  • Matching: compare and contrast correct demonstrations with incorrect ones.  (Use correct and incorrect examples of carrying [in addition] to show why carrying is necessary.)

Do this?  Don’t do that?  How’m I doing?

vM&K use “continent tutoring” to mean monitoring the learner’s performance and presenting procedural information just-in-time.  Part-task practice is focused on particular skills, and so it’s easier for an instructor or a computer system to provide that JIT knowledge as needed.

Likewise, corrective feedback is easier for part-task practice, because the range of behavior is narrower.  Ten Steps discusses a “model tracing” approach, where the observer or a computer program monitors the learner’s progress on each step or rule.  Suppose the practice item involves solving the math problem:

3 ( 2x + 5 ) = 9

If a step’s correct, there’s no feedback.  If incorrect, feedback could look like this:

  • You tell the learner “you’ve made an error.”
  • You explain the error.  “You multiplied 2X by 3.  In a problem like this, you also need to multiply the 5 by 3 .”

Practice to overlearn

Part-task practice has two effects: it helps combine small rules into larger chunks (“knowledge compilation”), and it strengthens the application of those rules through deliberate, repeated practice.

Put down the tentacle, carry the claw...vM&K cite the “power law of practice,” another side trip into math ( “the log of the time to complete a response [is] a linear function of the log of the number of successful executions” ).

Here’s the idea.  Let’s say you take 3 seconds to add two digits.  You practice 100 times, and lower the rate to 2 seconds.  Here’s the improvement from further practice:

  • 1,000 items gets you to 1.6 seconds
  • 10,000 items gets you to 1.3 seconds
  • 100,000 items gets you to 1.0 seconds

Key point here: first,  compilation (acquiring basic skill with a procedure or rule) takes much less time than strengthening.  Second, strengthening continues over a long period of time.

I think there’s a connection with the notion that it takes 10,000 hours to become an expert (though I see “10,000 hours” as a metaphor, rather than as a strict standard).  At age 93, Pablo Casals was asked why he continued to practice scales every day, more than 80 years after he began studying the cello.  “Because I think I’m making progress.”

Building skill

Part-task practice typically goes through three stages:

  • Accuracy: the goal is to reach an acceptable level of accuracy.  In typing, for example, that might mean typing the letter C and using the correct finger to do so.
  • Speed: the learner performs the skill under moderate speed stress, like a typing drill.  The goal is to maintain current accuracy under stress.  One effect is to make it impossible to follow the steps in a procedure consciously (“move the middle finger of the left hand down to the C key, then press”), forcing automatic.
  • Integration: the learner performs the skill along with related skills, still striving to maintain speed and accuracy.  An example would be moving from typing drills (“cog cog cog tic tic tic can can can”) to full text.

atcSome skills can’t sensibly be practiced sufficiently in real time to develop automaticity.  Simulations can compress such time.  Ten Steps gives an example of air-traffic controllers.  Usually there’s a gap of about 5 minute between telling an aircraft to where to turn and seeing the result on a monitor.  A simulation can compress that to a few seconds, and the novice air-traffic controller “can practice more items on one day than in one year of normal training.”

Part-task practice seems to work better when it’s distributed–relatively short periods of practice, spaced over time–than when it’s done in long, concentrated periods of drill.  In addition, alternating practice sessions with other learning activities, according to vM&K, helps make the practice more effective.

Making practice work (in a good sense)

The Ten Steps emphasizes two principles for part-task practice:

  • Learners must be able to related the practice to whole-task performance.
  • Part-task practice should be distributed, and should alternate with learning tasks.

vM&K cite a study in which college students were learning Boolean functions.  Practice with 8,000 items before the whole task (troubleshooting logical circuits) had no effect; practice had a positive effect after exposure to a simple version of the whole task.

…And with that, I’m done with the details of the Ten Steps. Two remaining chapters deal with media and with self-directed learning; a final one has the title, “Closing Remarks.”  So the series isn’t quite done, but I think eventually it will be.

CC-licensed images: chef-knife photo by Sara Björk;
addition problems by Arenamontanus;
USAF air traffic controller by Lietmotiv.

Self-directed learning: stepping out on your own

This entry is part 19 of 21 in the series Ten Steps to Complex Learning (the book).

Chapter 15 of Ten Steps to Complex Learning looks at self-directed learning.  Van Merriënboer and Kirschner’s ten steps are part of a blueprint for programs for learning–a way to structure learning for complex real-world tasks.  Since the Ten Steps include things like dynamic selection of learning tasks based on the learner’s interest and abilities, there’s a natural fit with self-directed learning.

Successful self-directed learning requires skills like:

  • Monitoring progress (possibly in too much detail)Orienting: What do I want to learn?  Where could I learn it?  How can I use it?
  • Planning: What should I do?  How much time and effort do I need?
  • Monitoring: Have I learned enough from this task?  Am I paying attention to context?
  • Adjusting: Do I need to make changes?  Do I need some help?
  • Assessing: Have I reached my goal?  What do I need to work on next?

These skills are beyond the scope of the Ten Steps, but many principles contained in the steps support the skills in action.

vM&K see three levels of self-directed learning.  Independent part-task practice is the simplest.  Within an overall program, a learner can choose practice items.  That’s because part-task practice applies best to well-defined items that fit well into individual practice and benefit from such things as automated drill-and-practice.

Just-in-time open education goes beyond the procedural aspects of part-task practice.  Think of it as resource-based learning that includes supportive information–in other words, this level of self-directed learning includes problem-solving and reasoning aspects of the learning tasks.

With on-demand education, learners decide not only on the resources but on the learning tasks as well.  Challenges for the learner include:

  • How to choose an appropriate level of difficulty
  • How to obtain an appropriate level of support and guidance
  • How to get enough variability (so that learning tasks vary as much as they do in the real world)

Reading those points reminded me of my own experience with Head First HTML with CSS, a book that teaches you, well, HTML and CSS (my post about the book).  Head First absolutely worked for me–I liked the presentation, I liked the level of detail, I learned lots of stuff.  But it might not work for everyone.  The authors themselves say the book’s not for you if you’re completely new to computers, or if you’re a skilled web developer looking for a reference book, or if you’re afraid to try something different.

(Though you’ll find this on page xxvi:
Note from marketing: this book is for anyone with a credit card.
)

How’m I doing with self-directed learning?

This chapter has a highly detailed discussion about assessing performance in self-directed learning.  Keep in mind that the Ten Steps assume that performance standards remain constant throughout a learning program.  What differs from level to level is the complexity of the task and thus the relevant portion of the standards.

On-the-job assessment

A program for trauma specialists would have highly detailed standards for diagnosis, for example.  A subset of those skills might be “take vital signs,” which make up one aspect of the much larger skill.

Assessment methods with a high reliability (e.g., multiple choice tests) have, in general, a relatively low external validity, and, vice-versa, assessment methods with a high external validity (e.g., on the job performance) have a relatively low reliability.  Therefore, the Ten Steps recommend using a rich mix of assessment methods….A mix of assessors should also be used.

vM&K say that the most important assessor in self-directed learning is the learner herself.  Next come peer assessments from fellow learners or colleagues.  Finally, assessors can include teachers, instructors, experts, customers, and others who engage with the learner.

Vertical and horizontal assessment

This chapter ends with a concept that confused me for quite a while.  The authors talk about “protocol portfolio scoring,” a complicated tool for standards, assessment results, and “vertical and horizontal assessment.”  Here’s what I think they’re talking about:

The Ten Steps is all about learning whole tasks.  It’s also about learning through task classes: clusters of learning tasks with similar overall difficulty but variations that reflect the real world.

Imagine that there were eight aspects (or facets) that could apply to the tasks in a class, even if each task didn’t have each aspect.  A vertical assessment looks at one aspect for all the tasks, while a horizontal assessment looks at all aspects for one task.

You can see an example in this Google Books search of Ten Steps (opens in a new window).  Look at the chart on page 236.  The center columns show “vertical standards” for eight aspects that apply to the six tasks that make up class 1.  Not every aspect applies–for example, only five aspects apply to task 1.1.

If you’re training hairstylists, class 1 might involve simple kinds of haircuts.   You could have several basic skills in class 1, with some tasks involving curly hair and some straight.

As I understand the explanation that follows, vertical scoring is cumulative–so I could miss one or two curly-hair aspects but achieve competence in this aspect over the entire class.  The horizontal scoring for each task deals with all its aspects.

I thought this was wildly and needlessly complicated until I read that the horizontal scores reflect adaptation.  Without going into too much detail, because the learner failed the horizontal score for task 1.1 (scored 3.0 when the standard was 3.74), task 1.2 provided more support.  The learner missed that one, too, so task 1.3 provided even more support.

At the same time, the vertical scores reflect emphasis.  If overall I don’t do well on, say, doing a basic trim, the vertical score can tell me that I’m falling short on scissor use and on working with curly hair.   I’m doing fine on asking the customer what he wants and on choosing where to cut.

Support and guidance for self-directed learning

More learner control isn’t always a good idea.  Low-ability learners can’t always make wise choices.  If you’ve never used HTML, you can’t make good decisions about whether to use CSS.  How do you decide the number of tasks to choose, and how do you select the specific tasks?

One approach is what vM&K call the intelligent agent–a person or a smart tool.  I’ve recently had experience with the latter–built-in advice from the Wii Fit.  The program has built-in assessments, offers a selection of activities (exercises), and monitors progress.  Probably not at the same level as a personal trainer, but suited to my not-quite-Olympic level of fitness.

“Support” in Ten Steps terms refers to procedural skills–so an intelligent agent is a bit like a Consumer Reports article, discussing various facts that you apply to your own situation.  “Guidance” refers to cognitive strategies and rules of thumb.

An unanswered question is who puts all this stuff together?  I have some thoughts I’m still mulling over.  The short-term answer might be, “What do you mean by who?”  For some complex skills, and for some contexts for those skills, you might have a fairly permanent body, like a medical school or a pilot-training program.  For others, “who” and “this stuff” may vary according to the skills and the performers.

CC-licensed images:
Graffiti report card by bbaunach.
“Loading” (from actual software) by me.