Learnlets
Clark Quinn's Learnings about Learning
(The Official Quinnovation blog)

28 October 2014

Cognitive prostheses

Clark @ 8:05 am

While our cognitive architecture has incredible capabilities (how else could we come up with advances such as Mystery Science Theater 3000?), it also has limitations. The same adaptive capabilities that let us cope with information overload in both familiar and new ways also lead to some systematic flaws. And it led me to think about the ways in which we support these limitations, as they have implications for designing solutions for our organizations.

The first limit is at the sensory level. Our mind actually processes pretty much all the visual and auditory sensory data that arrives, but it disappears pretty quickly (within milliseconds) except for what we attend to. Basically, your brain fills in the rest (which leaves open the opportunity to make mistakes). What do we do? We’ve created tools that allow us to capture things accurately: cameras and microphones with audio recording. This allows us to capture the context exactly, not as our memory reconstructs it.

A second limitation is our ‘working’ memory. We can’t hold too much in mind at one time. We ‘chunk’ information together as we learn it, and can then hold more total information at one time. Also, the format of working memory largely is ‘verbal’. Consequently, using tools like diagramming, outlines, or mindmaps add structure to our knowledge and support our ability to work on it.

Another limitation to our working memory is that it doesn’t support complex calculations, with many intermediate steps. Consequently we need ways to deal with this. External representations (as above), such as recording intermediate steps, works, but we can also build tools that offload that process, such as calculators. Wizards, or interactive dialog tools, are another form of a calculator.

Processing information in short term memory can lead to it being retained in long term memory. Here the storage is almost unlimited in time and scope, but it is hard to get in there, and isn’t remembered exactly, but instead by meaning. Consequently, models are a better learning strategy than rote learning. But external sources like the ability to look up or search for information is far better than trying to get it in the head.

Similarly, external support for when we do have to do things by rote is a good idea. So, support for process is useful and the reason why checklists have been a ubiquitous and useful way to get more accurate execution.

In execution, we have a few flaws too. We’re heavily biased to solve new problems in the ways we’ve solved previous problems (even if that’s not the best approach. We’re also likely to use tools in familiar ways and miss new ways to use tools to solve problems. There are ways to prompt lateral thinking at appropriate times, and we can both make access to such support available, and even trigger same if we’ve contextual clues.

We’re also biased to prematurely converge on an answer (intuition) rather than seek to challenge our findings. Access to data and support for capturing and invoking alternative ways of thinking are more likely to prevent such mistakes.

Overall, our use of more formal logical thinking fatigues quickly. Scaffolding help like the above decreases the likelihood of a mistake and increases the likelihood of an optimal outcome.

When you look at performance gaps, you should look to such approaches first, and look to putting information in the head last. This more closely aligns our support efforts with how our brains really think, work, and learn. This isn’t a complete list, I’m sure, but it’s a useful beginning.

24 October 2014

#DevLearn Schedule

Clark @ 8:30 am

As usual, I will be at DevLearn (in Las Vegas) this next week, and welcome meeting up with you there.  There is a lot going on.  Here’re the things I’m involved in:

  • On Tuesday, I’m running an all day workshop on eLearning Strategy. (Hint: it’s really a Revolutionize L&D workshop ;).  I’m pleasantly surprised at how many folks will be there!
  • On Wednesday at 1:15 (right after lunch), I’ll be speaking on the design approach I’m leading at the Wadhwani Foundation, where we’re trying to integrate learning science with pragmatic execution.  It’s at least partly a Serious eLearning Manifesto session.
  • On Wednesday at 2:45, I’ll be part of a panel on mlearning with my fellow mLearnCon advisory board members Robert Gadd, Sarah Gilbert, and Chad Udell, chaired by conference program director David Kelly.

Of course, there’s much more. A few things I’m looking forward to:

  • The keynotes:
    •  Neil DeGrasse Tyson, a fave for his witty support of science
    • Beau Lotto talking about perception
    • Belinda Parmar talking about women in tech (a burning issue right now)
  • DemoFest, all the great examples people are bringing
  • and, of course, the networking opportunities

DevLearn is probably my favorite conference of the year: learning focused, technologically advanced, well organized, and with the right people.  If you can’t make it this year, you might want to put it on your calendar for another!

14 October 2014

Types of meaningful processing

Clark @ 8:21 am

In an previous post, I argued for different types and ratios for worthwhile learning activities. I’ve been thinking about this (and working on it) quite a bit lately. I know there are other resources that I should know about (pointers welcome), but I’m currently wrestling with several types of situations and wanted to share my thinking. This is aside from scenarios/simulations (e.g. games) that are the first, best, learning practice you can engage in, of course. What I’m looking for is ways to get learners to do processing in ways that will assist their ability to do.  This isn’t recitation, but application.

So one situation is where the learner has to execute the right procedure. This seems easy, but the problem is that they’re liable to get it right in practice.  The problem is that they still can get it wrong when in real situations. An idea I had heard of before, but was reiterated through Socratic Arts (Roger Schank & cohorts) was to have learners observe (e.g. video) of someone performing it and identifying whether it was right or not. This is a more challenging task than  just doing it right for many routine but important tasks (e.g. sanitation). It has learners monitor the process, and then they can turn that on themselves to become self-monitoring.  If the selection of mistakes is broad enough, they’ll have experience that will transfer to their whole performance.

Another task that I faced earlier was the situation where people had to interpret guidelines to make a decision. Typically, the extreme cases are obvious, and instructors argue that they all are, but in reality there are many ambiguous situations.  Here, as I’ve argued before, the thing to do is have folks work in groups and be presented with increasingly ambiguous situations. What emerges from the discussion is usually a rich unpacking of the elements.  This processing of the rules in context exposes the underlying issues in important ways.

Another type of task is helping people understand applying models to make decisions. Rather than present them with the models, I’m again looking for more meaningful processing.  Eventually I’ll expect learners to make decisions with them, but as a scaffolding step, I’m asking them to interpret the models in terms of their recommendations for use.  So before I have them engage in scenarios, I’ll ask them to use the models to create, say, a guide to how to use that information. To diagnose, to remedy, to put in place initial protections.  At other times, I’ll have them derive subsequent processes from the theoretical model.

One other example I recall came from a paper that Tom Reeves wrote (and I can’t find) where he had learners pick from a number of options that indicated problems or actions to take. The interesting difference was then there was a followup question about why. Every choice was two stages: decision and then rationale. This is a very clever way to see if they’re not just getting the right answer but can understand why it’s right.  I wonder if any of the authoring tools on the market right now include such a template!

I know there are more categories of learning and associated tasks that require useful processing (towards do, not know, mind you ;), but here are a couple that are ‘top of mind’ right now. Thoughts?

 

 

1 October 2014

Constructive vs instructive

Clark @ 8:11 am

A commenter on last week’s post asked an implicit question that caused me to think. The issue was whether the solutions I was proposing are having the learners be self directed or whether it was ‘push’ learning.  And I reckon there’s a bit of both, but I’m fighting for more of a constuctivist approach  than the instructivist model.

I’ve argued in the past for a more active learning, and I think the argument for pure instructivism sets up a straw man (Feuerzeig argued for guided discovery back in ’85!).  Obviously, I think that pure exploration is doomed to failure, as we know that learners can stay in one small corner of a search space without support (hence the coaching in Quest).  However, a completely guided experience doesn’t ‘stick’ as well, either.

Another factor is our target learners.  In my experience, more constructivist approaches can be disturbing to learners who have had more instructivist approaches.  And the learners we are dealing with haven’t been that successful in school, and typically need a lot of scaffolding.

Yet our goals are fairly pragmatic overall (and in general we should be looking for ways to pragmatic in more of our learning). We’re focused on meaningful skills, so we should leverage this.

In this case, I’m moving the design to more and more “here’s a goal, here’re some resources” type of approach where the goal is to generate a work-related integration (requiring relevant cognitive processing).  Even if it’s conceptual material, I want learners to be doing this, and of course the main focus is on real contextualized practice.

I’m pushing a very activity-based pedagogy (and curriculum). Yes, the tasks are designed, but they’re expected to take some responsibility for processing the information to produce outputs. The longer term goal is to increase the challenge and variety as we go through the curriculum, developing learner’s ability to  learn to learn and ability to adapt as well. Make sense?

30 September 2014

Types and proportions of learning activities?

Clark @ 8:49 am

I’ve been on quite the roll of late, calling out some bad practices and calling for learning science. And it occurs to me that there could be some pushback.  So let me be clear, I strongly suggest that the types of learning that are needed are not info dump and knowledge test, by and large.  What does that mean? Let’s break it down.

First, let me suggest that what’s going to make a difference to organizations is not better fact-remembering. There are times when fact remembering is needed, such as medical vocabulary (my go-to example). When that needs to happen, tarted up drill-and-kill (e.g .quiz show templates, etc) are the way to do it.   Getting people to remember rote facts or arbitrary things (like part names) is very difficult. And largely unnecessary if people can look it up, e.g. the information is in the world (or can be).  There are some things that need to be known cold, e.g. emergency procedures, hence the tremendous emphasis on drills in aviation and the military. Other than that, put it in the world, not the head.  Look up tables, info sheets, etc are the solution.  And I’ll argue that the need for this is less than 5-10% of the time.

So what is useful?  I’ll argue that what is useful is making better decisions.  That is, the ability to explain what’s happened and react, or predict what will happen and make the right choice as as consequence.  This comes from model-based reasoning.  What sort of learning helps model-based reasoning? Two types, in a simple framework. You need to process the models to help them be comprehended, and use them in context to make decisions with the consequences providing feedback.  Yes, there likely will be some content presentation, but it’s not everything, and instead is the core model with examples of how it plays out in context. That is, annotated diagrams or narrated animations for the models; comic books, cartoons, or videos for the examples.  Media, not bullet points.

The processing that helps make models stick includes having learners generate products: giving them data or outcomes and having them develop explanatory models. They can produce summary charts and tables that serve as decision aids. They can create syntheses and recommendations.  This really leads to internalization and ownership, but it may be more time-consuming than worthwhile. The other approach is to have learners make predictions using the models, explaining things.  Worst case, they can answer questions about what this model implies in particular contexts.  So this is a knowledge question, but not a “is this an X or a Y”, but rather “you have to achieve Z, would you use approach X, or approach Y”.

Most importantly, you need people to use the models to make decisions like they’ll be making in the workplace.  That means scenarios and simulations.  Yes, a mini-scenario of one question is essentially a multiple choice (though better written with a context and a decision), but really things tend to be bundled up, and you at least need branching scenarios. A series of these might be enough if the task isn’t too complex, but if it’s somewhat complex, it might be worth creating a model-based simulation and giving the learners lots of goals with it (read: serious game).

And, don’t forget, if it matters (and why are you bothering if it doesn’t), you need to practice until they can’t get it wrong.  And you need to be facilitating reflection.  The alternatives to the right answer should reflect ways learners often go wrong, and address them individually. “No, that’s not correct, try again” is a really rude way to respond to learner actions.  Connect their actions to the model!

What this also implies is that learning is much more practice than content presentation.  Presenting content and drilling knowledge (particularly in about an 80/20 ratio), is essentially a waste of time.  Meaningful practice should be more than half the time.  And you should consider putting the practice up front and driving them to the content, as opposed to presenting the content first.  Make the task make the content meaningful.

Yes, I’m making these numbers up, but they’re a framework for thinking. You should be having lots of meaningful practice.  There’s essentially no role for bullet points or prose and simplistic quizzes, very little role for tarted up quizzes, and lots of role for media on the content side and  branching scenarios and model-driven interactions on the interaction side.  This kind of is an inverse of the tools and outputs I see.  Hence my continuing campaign for better learning.  Make sense?

24 September 2014

Better Learning in the Real World

Clark @ 8:25 am

I tout the value of learning science and good design.  And yet, I also recognize that to do it to the full extent is beyond most people’s abilities.  In my own work, I’m not resourced to do it the way I would and should do it. So how can we strike a balance?  I believe that we need to use smart heuristics instead of the full process.

I have been talking to a few different people recently who basically are resourced to do it the right way.  They talk about getting the right SMEs (e.g. with sufficient depth to develop models), using a cognitive task analysis process to get the objectives, align the processing activities to the type of learning objective, developing appropriate materials and rich simulations, testing the learning and using  feedback to refine the product, all before final release.  That’s great, and I laud them.  Unfortunately, the cost to get a team capable of doing this, and the time schedule to do it right, doesn’t fit in the situation I’m usually in (nor most of  you).  To be fair, if it really matters (e.g. lives depend on it or you’re going to sell it), you really do need to do this (as medical, aviation, military training usually do).

But what if you’ve a team that’s not composed of PhDs in the learning sciences, your development resources are tied to the usual tools, your budgets far more stringent, and schedules are likewise constrained? Do you have to abandon hope?  My claim is no.

Law of diminishing returns curveI believe that a smart, heuristic approach is plausible.  Using the typical ‘law of diminishing returns’ curve (and the shape of this curve is open to debate), I  suggest that it’s plausible that there is a sweet spot of design processes that gives you an high amount of value for a pragmatic investment of time and resources.  Conceptually, I believe you can get good outcomes with some steps that tap into the core of learning science without following the letter.  Learning is a probabilistic game, overall, so we’re taking a small tradeoff in probability to meet real world constraints.

What are these steps? Instead of doing a full cognitive task analysis, we’ll do our best guess of meaningful activities before getting feedback from the SME.  We’ll switch the emphasis from knowledge test to mini- and branching-scenarios for practice tasks, or we’ll have them take information resources and use them to generate work products (charts, tables, analyses) as processing.  We’ll try to anticipate the models,  and ask for misconceptions & stories to build in.  And we’ll align pre-, in-, and post-class activities in a pragmatic way.  Finally, we’ll do a learning equivalent of heuristic evaluation, not do a full scientifically valid test, but we’ll run it by the SMEs and fix their (legitimate) complaints, then run it with some students and fix the observed flaws.

In short, what we’re doing here are  approximations to the full process that includes some smart guesses instead of full validation.  There’s not the expectation that the outcome will be as good as we’d like, but it’s going to be a lot better than throwing quizzes on content. And we can do it with a smart team that aren’t learning scientists but are informed, in a longer but still reasonable schedule.

I believe we can create transformative learning under real world constraints.  At least, I’ll claim this approach is far more justifiable than the too oft-seen approach of info dump and knowledge test. What say you?

23 September 2014

Design like a pro

Clark @ 8:20 am

In other fields of endeavors, there is a science behind the approaches.  In civil engineering, it’s the properties of materials.  In aviation, it’s aeronautical engineering.  In medicine, it’s medical science.  If you’re going to be a professional in your field, you have to know the science.  So, two questions: is there a science of learning, and is it used.  The answers appear to be yes and no.  And yet, if you’re going to be a learning designer or engineer, you should know the science and be using it.

There is a science of learning, and it’s increasingly easy to find.  That’s the premise behind the Serious eLearning Manifesto, for instance (read it, sign it, use it!).  You could read Julie Dirksen’s Design for How People Learn as a very good interpretation of the science.  The Pittsburgh Science of Learning Center is compiling research to provide guidance about learning if you want a fuller scientific treatment.  Or read Bransford, et al’s summary of the science of How People Learna very rich overview.  And Hess & Saxberg’s recent Breakthrough Leadership in the Digital Age: Using Learning Science to Reboot Schooling is both a call for why and some guidance on how.

Among the things we know are that rote and abstract information isn’t retained, knowledge test doesn’t mean ability to do, getting it right once doesn’t mean it’s known, the list goes on.  Yet, somehow, we see elearning tools like ‘click to learn more’ (er, less), tarted up quiz show templates to drill knowledge, easy ways to take content and add quizzes to them, and more.  We see elearning that’s arbitrary info dump and simplistic knowledge test.  Which will have a negligible impact on anything meaningful.

We’re focused on speed and cost efficiencies, not on learning outcomes, and that’s not professional.  Look, if you’re going to do design, do it right.   Anything less is really malpractice!

17 September 2014

Learning in 2024 #LRN2024

Clark @ 8:14 am

The eLearning Guild is celebrating it’s 10th year, and is using the opportunity to reflect on what learning will look like 10 years from now.  While I couldn’t participate in the twitter chat they held, I optimistically weighed in: “learning in 2024 will look like individualized personal mentoring via augmented reality, AI, and the network”.  However, I thought I would elaborate in line with a series of followup posts leveraging the #lrn2024 hashtag.  The twitter chat had a series of questions, so I’ll address them here (with a caveat that our learning really hasn’t changed, our wetware hasn’t evolved in the past decade and won’t again in the next; our support of learning is what I’m referring to here):

1. How has learning changed in the last 10 years (from the perspective of the learner)?

I reckon the learner has seen a significant move to more elearning instead of an almost complete dependence on face-to-face events.  And I reckon most learners have begun to use technology in their own ways to get answers, whether via the Google, or social networks like FaceBook and LinkedIn.  And I expect they’re seeing more media such as videos and animations, and may even be creating their own. I also expect that the elearning they’re seeing is not particularly good, nor improving, if not actually decreasing in quality.  I expect they’re seeing more info dump/knowledge test, more and more ‘click to learn more‘, more tarted-up drill-and-kill.  For which we should apologize!

2. What is the most significant change technology has made to organizational learning in the past decade?

I reckon there are two significant changes that have happened. One is rather subtle as yet, but will be profound, and that is the ability to track more activity, mine more data, and gain more insights. The ExperienceAPI coupled with analytics is a huge opportunity.  The other is the rise of social networks.  The ability to stay more tightly coupled with colleagues, sharing information and collaborating, has really become mainstream in our lives, and is going to have a big impact on our organizations.  Working ‘out loud’, showing our work, and working together is a critical inflection point in bringing learning back into the workflow in a natural way and away from the ‘event’ model.

3. What are the most significant challenges facing organizational learning today?

The most significant change is the status quo: the belief that an information oriented event model has any relationship to meaningful outcomes.  This plays out in so many ways: order-taking for courses, equating information with skills, being concerned with speed and quantity instead of quality of outcomes, not measuring the impact, the list goes on.   We’ve become self-deluded that an LMS and a rapid elearning tool means you’re doing something worthwhile, when it’s profoundly wrong.  L&D needs a revolution.

4. What technologies will have the greatest impact on learning in the next decade? Why?

The short answer is mobile.  Mobile is the catalyst for change. So many other technologies go through the hype cycle: initial over-excitement, crash, and then a gradual resurgence (c.f. virtual worlds), but mobile has been resistant for the simple reason that there’s so much value proposition.  The cognitive augmentation that digital technology provides, available whenever and wherever you are clearly has benefits, and it’s not courses!  It will naturally incorporate augmented reality with the variety of new devices we’re seeing, and be contextualized as well.  We’re seeing a richer picture of how technology can support us in being effective, and L&D can facilitate these other activities as a way to move to a more strategic and valuable role in the organization.  As above, also new tracking and analysis tools, and social networks.  I’ll add that simulations/serious games are an opportunity that is yet to really be capitalized on.  (There are reasons I wrote those books :)

5. What new skills will professionals need to develop to support learning in the future?

As I wrote (PDF), the new skills that are necessary fall into two major categories: performance consulting and interaction facilitation.  We need to not design courses until we’ve ascertained that no other approach will work, so we need to get down to the real problems. We should hope that the answer comes from the network when it can, and we should want to design performance support solutions if it can’t, and reserve courses for only when it absolutely has to be in the head. To get good outcomes from the network, it takes facilitation, and I think facilitation is a good model for promoting innovation, supporting coaching and mentoring, and helping individuals develop self-learning skills.  So the ability to get those root causes of problems, choose between solutions, and measure the impact are key for the first part, and understanding what skills are needed by the individuals (whether performers or mentors/coaches/leaders) and how to develop them are the key new additions.

6. What will learning look like in the year 2024?

Ideally, it would look like an ‘always on’ mentoring solution, so the experience is that of someone always with you to watch your performance and provide just the right guidance to help you perform in the moment and develop you over time. Learning will be layered on to your activities, and only occasionally will require some special events but mostly will be wrapped around your life in a supportive way.  Some of this will be system-delivered, and some will come from the network, but it should feel like you’re being cared for in the most efficacious way.

In closing, I note that, unfortunately,my Revolution book and the Manifesto were both driven by a sense of frustration around the lack of meaningful change in L&D. Hopefully, they’re riding or catalyzing the needed change, but in a cynical mood I might believe that things won’t change near as much as I’d hope. I also remember a talk (cleverly titled: Predict Anything but the Future :) that said that the future does tend to come as an informed basis would predict with an unexpected twist, so it’ll be interesting to discover what that twist will be.

16 September 2014

On the Road Fall 2014

Clark @ 8:05 am

Fall always seems to be a busy time, and I reckon it’s worthwhile to let you know where I’ll be in case you might be there too! Coming up are a couple of different events that you might be interested in:

September 28-30 I’ll be at the Future of Talent retreat  at the Marconi Center up the coast from San Francisco. It’s a lovely spot with a limited number of participants who will go deep on what’s coming in the Talent world. I’ll be talking up the Revolution, of course.

October 28-31 I’ll be at the eLearning Guild’s DevLearn in Las Vegas (always a great event; if you’re into elearning you should be there).  I’ll be running a Revolution workshop (I believe there are still a few spots), part of  a mobile panel, and talking about how we are going about addressing the challenges of learning design at the Wadhwani Foundation.

November 12-13 I’ll be part of the mLearnNow event in New Orleans (well, that’s what I call it, they call it LearnNow mobile blah blah blah ;).  Again, there are some slots still available.  I’m honored to be co-presenting with Sarah Gilbert and Nick Floro (with Justin Brusino pulling strings in the background), and we’re working hard to make sure it should be a really great deep dive into mlearning.  (And, New Orleans!)

There may be one more opportunity, so if anyone in Sydney wants to talk, consider Nov 21.

Hope to cross paths with you at one or more of these places!

10 September 2014

Learning Engineering

Clark @ 8:37 am

Last week I had the opportunity to attend the inaugural meeting of the Global Learning Council.  While not really global in either sense (little representation from overseas nor from segments other than higher ed), it was a chance to refresh myself in some rigor around learning sciences. And one thing that struck me was folks talking about learning engineering.

If we take the analogy from regular science and engineering, we are talking about taking the research from the learning sciences, and applying it to the design of solutions.  And this sounds like a good thing, with some caveats.  When talking about the Serious eLearning Manifesto, for example, we’re talking about principles that should be embedded in your learning design approach.

While the intention was not to provide coverage of learning science, several points emerged at one point or another as research-based outcomes to be desired. For one, the value of models in learning.  Another was, of course, the value of spacing practice. The list goes on.  The focus of the engineering, however, is different.

While it wasn’t an explicit topic of the talk, it emerged in several side conversations, but the focus is on design processes and tools that increase the likelihood of creating effective learning practices.  This includes doing a suitable job of creating aligned outcomes through processes of working with SMEs, identifying misconceptions to be addressed, ensuring activities are designed that have learners appropriately processing and applying information, appropriate spread of examples, and more.

Of course, developing an accurate course for any topic is a thorough exercise.  Which is desirable, but not always pragmatic.  While the full rigor of science would go as far as adaptive intelligent tutoring systems, the amount of work to do so can be prohibitive under pragmatic constraints.  It takes a high importance and large potential audience to do this for other than research purposes.

In other cases, we use heuristics.  Sometimes we go too far; so just dumping information and adding a quiz is often seen, though that’s got little likelihood of having any impact.  Even if we do create an appropriate practice, we might only have learners practice until they get it right, not until they can’t get it wrong.

Finding the balance point is an ongoing effort. I reckon that the elements of good design is a starting point, but you need processes that are manageable, repeatable, and scalable.  You need structures to help, including representations that have support for identifying key elements and make it difficult to ignore the important elements.  You ideally have aligned tools that make it easy to do the right things.

And if this is what Learning Engineering can be, systematically applying learning science to design, I reckon there’s also a study of learning science engineering, aligning not just the learning, but the design process, with how we think, work, and learn.  And maybe then there’s a learning architecture as well – where just as an architect designs the basic look and feel of the halls & rooms and the engineers build them – that designs the curriculum approach and the pedagogy, but the learning engineers follow through on those principles for developing courses.

Is learning engineering an alternative to instructional design?  I’m wondering if the focus on engineering rather than design (applied science, rather than art) and learning rather than instruction (outcomes, not process), is a better characterization.  What do you think?

Next Page »

Powered by WordPress