Learnlets

Secondary

Clark Quinn’s Learnings about Learning

Clear about the concept

by Leave a Comment

I went to hear a talk the other day. It was about competency-based education (CBE) for organizations. Ostensibly. And, while I’m now affiliated with IBSTPI, it’s not like I’m a competency expert. And maybe I expect too much, but I really hope for people to be clear about the concept. Alas, that’s not what I found.

So, it started out reasonably well, talking about how competencies are valuable. There were a number of points, and many made sense, although some were redundant. Maybe I missed some nuance? I try to be open-minded. It’s about creating clear definitions of performance, and aligning those with assessments. Thus, you’re working on very clear descriptions of what people should be doing.

It got  interesting when the speaker decided to link CBE to Universal Design for Learning (UDL).  And it’s a good program.  UDL talks about using multiple representations to increase the likelihood for different learners to be able to comprehend and respond. This, in the talk, was mapped to three different segments: engaging the learners in multiple ways, communicating concepts in multiple ways, and allowing assessment in multiple ways. And this is good. For learning. Does it make sense for CBE?

To start, the argument was, you should make the rationale for the learning in multiple ways. While in general CBE inherently embodies meaningfulness in the nature of clear and needed skills, I don’t have a problem with this. I argue you should hook learners in emotionally  and cognitively, and those can be separate activities. There was a brief mention of something like ‘learning styles’, but while now wary, I was ready to let it go.

However, the talk went on to make a case for multiple representations of content. And here the slide  explicitly  said ‘learning styles’ and used VARK. And don’t get me wrong, multiple representations and media are good,  but not for learning styles! The current status is that there’s essentially no valid instrument to measure learning styles, and no evidence that even if you did, that it makes a difference. None. So, of course, I raised the issue. And we agreed that maybe not for learning styles, but multiple representations weren’t bad.

The final point was that there could be multiple forms of assessment. At this point, I wasn’t going to interrupt again, but at the end of the session raised the point that the critical element of CBE is aligning the assessment with the performance! You can’t have them do an interpretative dance about identifying fire hazards, for instance, you have to have them identify fire hazards! So, here the audience ultimately agreed that variability was acceptable  as long as it measured the actual performance. Again, I don’t think the speaker was clear about the concept.

There were two major flaws in this talk. One was casually mashing up a couple of essentially incommensurate ideas. CBE and UDL aren’t natural partners. There can be overlapping concepts, but… The second, of course, is using a popular but fundamentally flawed myth about learning. If you’re going to claim authority, don’t depend on broken concepts.

To put it another way, I think it’s fair to expect speakers to be clear about the concept. (Either that, or maybe the lesson is that Clark shouldn’t be allowed to listen to normal speakers. ;)  Please, please, know what you’re talking about before you talk about it. Is that too much to ask?

Sub-symbolic and Situated

by Leave a Comment

At the time that the connectionist folks were working on neural nets, another similar approach was genetic algorithms. Both were working in a different way than the previous formal approaches to AI. The distinction between the two became known as symbolic vs sub-symbolic. And it’s useful to review why, particularly in the current climate of increasing interest in AI and cognitive science. An interesting outcome is that the sub-symbolic work exposed the contextualized nature of our reasoning. So there’s a link between sub-symbolic and situated cognition.

The prevailing model, starting with the cognitive revolution which arguably began in 1956 (an auspicious year ;) was a formal logical one. Whether in ‘production’ rules of IF THEN, or other formal mechanisms, the notion was to operate on semantic objects like numbers and concepts. This reflected, at the time, the belief that we’re formal logical thinkers.

As cognitive research continued, there was a growing recognition that our behaviors didn’t match particularly well with formal logic (c.f. Kahnemann & Tversky’s work, summed up in  Thinking Fast and Slow). Several cognitive scientists separately came up with structures that more aptly described some of the properties we saw: Roger Schank called them scripts (he was focused on episodic thinking, not semantic), Marvin Minksy called them frames, and Dave Rumelhart called them schemas (after Bartlett).

What Rumelhart subsequently saw was that the properties he was trying to capture were very hard to represent in formal logic. He went on, with his colleague Jay McLelland and their collaborators) to develop what they called Parallel Distributed Processing (PDP). These are now known as neural nets (NNs) and are the basis for much of machine learning.

I was in the lab at the time Dave and Jay were working on neural nets, but detoured down a different path. Following work on analogical reasoning (my Ph.D. thesis topic), I became aware of the work Holland, Holyoak, Nisbett, & Thagard were doing with induction. Their framework was genetic algorithms (GAs). Both GAs and NNs use input strings and output strings to work, but internally they represent things differently.

After so much work on symbolic reasoning, here were mechanisms operating beneath the symbolic level. Yet they were attempting to create symbolic behavior. NNs obviously, more closely resemble our cognitive architecture (though GAs are still used in some areas like program generation). So, our conscious thinking  is symbolic, but our actual cognition is happening below our conscious thinking. Hence things like illusions, fallacies, myths, and more.

What emerged from this realization is that our cognition isn’t just sub-symbolic, but  situated.  That is, what is conscious is a combination of what comes in from our senses, and what we know. In fact, with the limited attention we have,  much of what we think we’re perceiving, we’re actually generating!

This it accounts for why we’re bad at doing things by rote; we’re liable to confound steps and contexts. This ends up being important because it means we have to work harder for any learning interventions to work effectively  across  contexts. The relationship between sub-symbolic and situated is, at least to me, and interesting story of the development of cognitive science.

Yet, it still means that our learning  works most effectively at the conscious level of symbols, because that can accelerate learning over having to deal with everything through practice and feedback.  (And explains why programs talking about neural really aren’t working there.) We still need those, but conscious models can provide a framework to become self-improving over time. So don’t forget to provide the models, and sufficient practice, and feedback.

Lucky on Foundations

by Leave a Comment

I was thinking about my next directions, and it led to me to think a bit about my foundations. And I realized I’ve been very lucky (and I’m grateful). I’ve had good parents, mentors, colleagues, and friends. But I’ve also had some fortunate timings, and it’s worth reflecting how I’ve been lucky on foundations upon which to build. (A personal reflection, not necessarily worth your time ;)

It started with college, really. I’d always been a typical lad, but with an extra serving of geek (I didn’t fit in with any clique so hung with a few similarly chaotic-good chaps :).  I started college interested in marine bio, but there was no formal link between undergrad study and Scripps. The bio program was all cut-throat med, and while I  could cut it, it was all rote memorization and deadly boring. So…

I took some comp sci classes, and was tutoring for extra money on the side.  Lucky chance: I got a job doing the computer support for the office that coordinated the tutoring. That sparked my awareness of the connections between computers and learning. Of course, back then, at my school, there was no such program. Luck 2: my school had a program where you could design your  own major. I found a couple of professors doing a project on using email for classroom discussion (circa ’78; we had the DARPAnet, otherwise there  was no email; more luck). They agreed to sponsor my project.

After graduating, I looked all over the country for an org that wanted someone interested in computers and learning. More luck, I finally came across Jim Schuyler, and as he was starting DesignWare, I got a job! And, importantly, it was designing and programming on the earliest personal computers. And I realized that there was real potential for learning in games! But I also realized that we didn’t know enough how to design them. And then I read about ‘cognitive engineering’ (applying what we know about cognition to the design of systems).

I was accepted into the cog program with Don Norman, who’d written the article. And this was another major stroke of luck. While Don’s students were researching how to build systems for how people think, my twist was about how people learn. I got to study behavioral, cognitive, social, even machine learning!  Also, Don’s lab partner Dave Rumelhart was conducting his research with Jay McClleland on what became neural nets. You can’t help but get exposed to related research through lab meetings, seminars, and more, even if you’re not active in the particular work. And Ed Hutchins was doing his work on distributed cognition.  This was a fundamental shift in perspective from formal to situated cognition.

The lab ran a Unix system, so I was getting steeped in computing systems to complement my personal computer work, along with the cognition focus. I subsequently did a post-doc at LRDC, getting deeper steeped into cognitive learning, and then joined a school of Computer Science, getting further background in computation. I was on the internet before there was a web (and foolishly was rather complacent about it)! And it’s enabled me to keep an eye on new developments like mobile and content and more, and understand their core affordance.

I also got steeped in design, having a chance to look at graphic, industrial, software, architecture, and other approaches (more luck). I combined that with a study of the academic literature, of course. These three foundations have been the basis of my work: applying cognitive and learning sciences to the design of technology to create learning and performance systems.

There’s much more to the story, of course. Serendipity continued in jobs and people to guide me, I’m happy to say.  Mentors being shy, you can’t really thank folks enough, so if I’ve been lucky in foundations, it’s my job to pass it on. I hope that this blog helps in  some way!

Theory or Research?

by Leave a Comment

There’s a lot of call for evidence-based methods (as mentioned yesterday): L&D, learning design, and more. And this is a good thing. But…do you want to be basing your steps on a particular empirical study, or the framework within which that study emerged? Let me make the case for one approach. My answer to theory or research is theory. Here’s why.

Most research experiments are done in the context of a theoretical framework. For instance, the work on worked examples comes from John Sweller’s Cognitive Load theory. Ann Brown & Ann-Marie Palincsar’s experiments on reading were framed within Reciprocal Teaching, etc. Theory generates experiments which refine theory.

The individual experiments illuminate aspects of the broader perspective. Researchers tend to run experiments driven by a theory. The theory leads to a hypothesis, and then that hypothesis is testable. There  are some exploratory studies done, but typically a theoretical explanation is generated to explain the results. That explanation is then subject to further testing.

Some theories are even meta-theories! Collins & Brown’s Cognitive Apprenticeship  (a favorite) is based upon integrating several different theories, including the Reciprocal Teaching, Alan Schoenfeld’s work on examples in math, and the work of Scardemalia & Bereiter on scaffolding writing. And, of course, most theories have to account for others’ results from other frameworks if they’re empirically sound.

The approach I discuss in things like my Learning Experience Design workshops is a synthesis of theories as well. It’s an eclectic mix including the above mentioned, Cognitive Flexibility, Elaboration, ARCS, and more. If I were in a research setting, I’d be conducting experiments on engagement (pushing beyond ARCS) to test my own theories of what makes experiences as engaging and effective. Which, not coincidentally, was the research I was doing when I  was  an academic (and led to  Engaging Learning). (As well as integration of systems for a ubiquitous coaching environment, which generates many related topics.)

While individual results, such as the benefits of relearning, are valuable and easy to point to, it’s the extended body of work on topics that provides for longevity and applicability. Any one study may or may not be directly applicable to your work, but the theoretical implications give you a basis to make decisions even in situations that don’t directly map. There’s the possibility to extend to far, but it’s better than having no guidance at all.

Having theories to hand that complement each other is a principled way to design individual solutions  and design processes. Similarly for strategic work as well (Revolutionize L&D) is a similar integration of diverse elements to make a coherent whole. Knowing, and mastering, the valid and useful theories is a good basis for making organizational learning decisions. And avoiding myths!  Being able to apply them, of course, is also critical ;).

So, while they’re complementary, in the choice between theory or research I’ll point to one having more utility. Here’s to theories and those who develop and advance them!

Direct Instruction or Guided Discovery

by Leave a Comment

Recently, colleague Jos Arets of the 70:20:10 institute wrote a post promoting evidence-based work. And I’m a big fan, both of his work and the post. In the post, however, he wrote one thing that bugs me. And I realize I’m flying in the face of many august folks on whether to promote direct instruction or guided discovery. So let me explain myself ;).

It starts with a famous article by noted educational researchers Paul Kirschner, John Sweller, and Richard Clark. In it, they argue against “constructivist, discovery, problem-based, experiential, and inquiry-based teaching”. That’s a pretty comprehensive list. Yet these are respected authors; I’ve seen Richard Clark talk, have talked with John Sweller personally, and have interacted with Paul Kirschner online. They’re smart and good folks committed to excellent work. So how can I quibble?

First, it comes from their characterization of the opposition as ‘minimally guided’.Way back in 1985, Wallace Feurzig was talking about ‘guided discovery’, not pure exploration. To me, that’s a bit of a ‘straw man’ argument. Not minimally guided, but appropriately guided, would seem to me to be the appropriate approach.

Further, work by David Jonassen for one, and a meta-analysis conducted by Stroebel & Van Barneveld for another, suggested different outcomes. The general outcome is problem-based (as one instance being argued against) doesn’t yield  quite as good performance on a subsequent test, but is retained longer  and transfers better. And those, I suggest, are the goals we  should care about.  Similarly, research supports attempting to solve problems even if you can’t before you learn.

And I worry about the phrase “direct instruction”. That easy to interpret as ‘information dump and knowledge test’; it sounds like the old ‘error-free learning’! I’m definitely  not accusing those esteemed researchers of implying that, but I am afraid that under informed instructors could take that implication. It’s all too easy to see too much of that in classrooms. Teacher strategies tend to ignore results like spaced, varied, and deliberate practice. Similarly, the support for students to learn effective study skills is woeful.

Is there a reconciliation? I suggest there is. Professors Kirschner, Sweller, & Clark would, I suggest, expect sufficient practice to a criteria, and that the practice should match the desired performance. I suspect they want learners solving meaningful problems in context, which to me  is problem-based learning. And their direct instruction would be targeted feedback, along with models and examples. Which is what I strongly suggest. The more transfer you need, however, the broader contexts you need. Similarly, the more flexible application required would suggest the gradual removal of scaffolding.

So I really think that guided exploration, and meaningful direct instruction, will converge in what eventuates in practice. Look,  insufficiently guided practice isn’t effective, and I suspect that they wouldn’t suggest that bullet points are effective instruction. I just want to ensure that we focus on the important elements, e.g. what we highlighted in the Serious eLearning Manifesto. There  is a reason to think that direct instruction or guided discovery isn’t the dichotomy proposed, I’ll suggest. FWIW.

Reconciling Cognitions and Contexts

by Leave a Comment

In my past two posts, I first looked at cognitions (situated, distributed, social) by contexts (think, work, and learn), and then the reverse. And, having filled out the matrixes anew, they weren’t quite the same. And that, I think, is the benefit of the exercise, a chance to think anew. So what emerged? Here’s the result of reconciling cognitions and contexts.

Situated/Distributed/Social by Think/Work/LearnSo, taking each cell back in the original pass of cognitions by contexts, what results? I took the Think row to, indeed, be Harold Jarche’s Seek > Sense > Share model (ok, my interpretation). We have in Situated, the feeds you’ve set up to see, and then the particular searches you need in the current context. Then, of course, you experiment  and  represent as ways to externalize thinking for Distributed. Finally, you share Socially.

For Work, not practices but principles (and the associated practices therefrom) as well as facilitation to support Situated Work. Performance support is, indeed, the Distributed support for Work. And Socially, you need to collaborate on specific tasks and cooperate in general.

Finally, for Learning, for a Situated world you need (spread) contextualized practice to support appropriate abstraction of the principles. You want models and examples to support performance  in the practice, as Distributed resources. And, finally, for Social Learning, you need to communicate (e.g. discussions) and collaborate (group projects).

What’s changed is that I added search and feeds, and moved experiment, in the Think row. I went to principles from practices to support performance in ambiguity, left performance support untouched, and stayed with collaborations and cooperation instead of just shared representations (they’re part of collaborate). And, finally, I made practice about contexts, went from blended learning to support materials for learning, and interpreted social assignments as communicating and collaborating.

The question is, what does this mean? Does it give us any traction? I’m thinking it does, as it shifts the focus in what we’re doing to support folks. So I think it  was interesting and valuable (to my thinking, at least ;) to consider reconciling cognitions and contexts.

Contexts By Cognitions

by Leave a Comment

So, in my last post, I talked about exploring the links between cognitions on the one hand (situated, distributed, social), and contexts (aligning with how we think, work, & learn).  I did it one way, but then I thought to do it another, to instead consider Contexts by Cognitions, to see if I came to the same elements. And they weren’t quite identical!  So I thought I should share that thinking, and then come to a reconciliation. Thinking out loud, as it were.

Considering thinking, working, and learning by situated, distributed, and social.So in this one, I swapped the headings, emptied the matrix, and took a second stab at filling them out, with a relatively clear mind. (I generated the first diagram several days ago and had been iterating on it, but not today. Today I was writing it up and was early in the process, so I came to it  relatively  free of contamination. And of course, not completely, but this is ‘business significance’, not ‘statistical significance’ ;).  The resulting diagram appears similar, but also some differences.

When we consider Thinking by Situated, we’re talking about coping with emergent situations. I thought being guided by best principles would be the way to cope, abstracted models. I thought representation was key for distributing one’s thinking, and sharing of course for social.

Working Situatedly suggested having in-house practices and facilitation. Of course, Distributed support for Work is performance support. And working socially suggests  shared representations.

Finally, learning situated suggests the need for much practice (across contexts, I now think). Distributed support for learning are models and examples. And social learning suggests communicating (e.g. discussions) and collaboration (group projects).

Interestingly, these results differ from my previous post. So, I think I’ll have to reconcile them. The fact that I  did get different results,  and it sparked some additional thinking, is good. The outcome of considering contexts by cognitions improved the outcomes, I think. And that’s worth thinking about!

Cognitions By Contexts

by Leave a Comment

I have, in the past, talked about the three cognitions: situated, distributed, and social. Similarly, I talk about aligning with the contexts: how we think, work, and learn. I then wondered about how they interacted. Naturally, I diagrammed it (surprise, right?). I created the 3 x 3 matrix, and then tried to fill the boxes.  So here’s some preliminary thoughts (ok, they’ve already been processed a few times) on considering cognitions by contexts.

The intersections do point to some implications.  Cutting through the contexts by cognitions, we can make some prescriptions. When we think of Situated by Think, I suggested experimentation as a mechanism to help resolve unclear outcomes. Situated by Work suggested the ambiguity inherent in new situations, and suggested supporting addressing that. Finally, Situated for Learning suggests the need for meaningful practice.

Similarly, when we look at Distributed by Thinking, I considered the need to represent understanding concretely. For Work, it’s about using external tools to support effective performance, e.g. performance support. For Learning, it’s about blending learning  across a variety of elements: technologies, interaction methodologies, etc, to support successful outcomes.

Social is a bit of a conflict, because I often mean that as a reflection of ‘work’. Here, however, I’m considering Work as ‘getting stuff done’. (Note to self: reconcile this!). So Social and Think is the notion of sharing the results (hmm, pondering in next paragraph). Social and Work is collaboration & cooperation, working together specifically on projects and also more broadly a willingness to contribute when/where/ever. Finally, Social for learning is social assignments.

Which makes me think that the whole ‘Think’ line could be Harold Jarche’s Seek > Sense > Share model, and then we’re talking about the Situated Thinking would be continually seeking new information to help settle ambiguity. Which is a nice idea I might put in, but then I have to consider where I put experiment. That may have to go in with ‘represent’  in Distributed and Think.

I also, as an experiment, decided to swap the labels (horizontal for vertical), and see if I came up with the same inputs. And, no, I didn’t. That’s my  next  post, the swapped version. It won’t be ’til the beginning of July, because next week I’m speaking at the Realities 360 conference, and will be posting mindmaps of the keynotes, if all things go per usual. And there’ll be a reconciliation after that, as the above paragraph suggests. Stay tuned! But here you see me ‘think out loud’ as I try to consider Cognitions By Contexts. I welcome any thoughts of yours!

Cognition external

by Leave a Comment

reading outsideI was thinking a bit about distributed cognition, and recognized that there as a potentially important way to tease that apart. And I’ll talk it out first here, and maybe a diagram will emerge. Or not. The point is to think about how external tools can augment our thinking. Or, really, a way that at least partly, we have cognition external.

The evidence says that our thinking isn’t completely in our head. And I’ve suggested that that makes a good case for performance support. But I realize it goes further in ways I’ve thought about it elsewhere. So I want to pull those together.

The alternative to performance support, a sort of cognitive scaffolding, is to think about representation. Here we’re not necessarily supporting any particular performance, but instead supporting developing thinking. I shared Jane Hart’s diagram yesterday, and I know that it’s a revision of a prior one. And that’s important!

The diagram is capturing her framework, and such externalizations are a way to share; they’re a social as well as artifactual sharing. It’s part of a ‘show your work‘ approach to continuing to think. Of course, it doesn’t have to be social, it can be personal.

So both of these forms of distributed cognition are externalizing our thinking in ways that our minds have trouble comprehending. We can play around with relationships by spatially representing them. We can augment our cognitive gaps both formally through performance support, and informally by supporting externalizing our thinking.  Spreadsheets are another tool to externalize our thinking. So, too, for that matter, is text.

So we can augment our performance, and scaffold our thinking. Both can be social or solitary, but they both qualify as forms of distributed cognition (beyond social). And, importantly, both then should be consciously considered in thinking about revolutionizing L&D. We should be designing for cognition external.  The tools should be there, and the facilitation, to use either when appropriate. So, think distributed, as well as situated, and social. It’s how our brains work, we ought to use that as a guide. You think?

A very insightful framework

by 1 Comment

Jane Hart has just come up with something new and, to me, intriguing. Ok, so she’s a colleague from the Internet Time Alliance, and I’ve been a fan of her work for a while, but I think this is particularly good.  If you’ve read here before, you’ll know I love a good model (Harold Jarche’s Seek>Sense>Share comes to mind). So when I parsed her “from training to modern workplace learning”, it resonated in many ways.  So here’s her framework with some comments.

First, some context. If you’ve known my work at all, you know that I’ve been pushing a L&D revolution. And that’s about rethinking training to be about transformative experience design, performance support to be included, and informal learning to be also addressed. That’s  intellectricity! And it’s sometimes hard to tie them together coherently.

Jane’s always had a talent for drilling down into the practicalities in sensible ways. Her books, continually updated, have great specifics about things to do. This is a framework that ties it together nicely.

The thing I like is the way she’s characterized different activities. The categories of Discovery (informal learning), Discourse (social learning), and Doing (experiential learning) provides a nice handle around which to talk about elements, roles, and tasks. And, importantly, prescriptions.  And I really like the ‘meta’ layer, where she suggests skills for each vertical.

I’m not without quibbles, however small. For instance, with her use of microlearning, because of my concerns about the label rather than her specific intention. She told me personally that she means “short daily learning”, and I think that’s great. I just think of that as spaced learning ;). And I might label ‘discovery’ to be ‘develop’, because it’s about the individual’s continual learning. And I’m not sure there’s what I call ‘slow’ innovation there, creating a culture and practices about experimentation and exposure to the ‘adjacent possible’. But it’s hard for one diagram to capture everything, and this does a great job.

I admit that I haven’t parsed all the nuances yet. But as an advocate of diagrams  and frameworks, I think this is truly insightful  and  useful. (And she’s updated it so I’ve grabbed this copy which appears to have lost microlearning.)   I’m sure she, as well as I, welcome your thoughts!