We frequently hear that ‘perfection is the enemy of the good’. And that may well be true. However, I want to suggest that there’s another enemy that plagues us as learning experience designers. We may be trying to do good, but there are barriers. These are worthy of explicit discussion.

You also hear about the holy trinity of engineering: cheap, fast, or good; pick two. We have real world pressures that want us to do things efficiently. For instance, we have lots of claims that generative AI will allow us to generate more learning faster. Thus, we can do more with less. Which isn’t a bad thing…if what we produce is good enough. If we’re doing good, I’ll suggest, then we can worry about fast and cheap. But doing bad faster and cheaper isn’t a good thing! Which brings us to the second issue.

What is our definition of ‘good’? It appears that, too often, good is if people ‘like’ it. Which isn’t a bad thing, it’s even the first level in the Kirkpatrick-Katzell model: asking what people think of the experience. One small problem: the correlation between what people think of an experience, and it’s actual impact, is .09 (Salas, et al, 2012). That’s zero with a rounding error! What it means is that people’s evaluation of what they think of it, and the actual impact, isn’t correlated at all. It could be highly rated and not be effective, or highly rated and be effective. Etc. At core, you can’t tell by the rating.

What should be ‘good’? The general intent of a learning intervention (or any intervention, really) is to have an impact! If we’re providing learning, it should yield a new ability to ‘do’. There are a multitude of problems here. For one, we don’t evaluate performance, so how would we know if our intervention is having an impact? Have learners acquired new abilities that are persisting in the workplace and leading to the necessary organizational change? Who knows? For another, folks don’t have realistic expectations about what it takes to have an impact. We’ve devolved to a state where if we build it, it must be good. Which isn’t a sound basis for determining outcomes.

There is, of course, a perfectly good reason to evaluate people’s affective experience of the learning. If we’re designing experiences, having it be ‘hard fun’ means we’ve optimized the engagement. This is fine, but only after, we’ve established efficacy. If we’re not having a learning impact in terms of new abilities to perform, what people think about it isn’t of use.

Look, I’d prefer us to be in the situation where perfection to be the enemy of the good! That’d mean we’re actually doing good. Yet, in our industry, too often we don’t have any idea whether we are or not. We’re not measuring ‘good’, so we’re not designing for it. If we measured impact first, then experience, we could get overly focused on perfection. That’d be a good problem to have, I reckon. Right now, however, we’re only focused on fast and cheap. We won’t get ‘good’ until we insist upon it from and for ourselves. So, let’s shall we?

I’m a fan of models. Good models that are causal or explanatory can provide guidance for making the right decisions. However, there are some approaches that are, I suggest, less than helpful. What makes a good or bad model? My problem is about distinguishing when to talk about each: simple models and complex problems.

A colleague of ours sent me an issue of a newsletter (it included the phrase ‘make it meaningful‘ ;). In it, the author was touting a four letter acronym-based model. And, to be fair, there was nothing wrong with what the model stipulated. Chunking, maintaining attention, elaboration, and emotion are all good things. What bothered me was that these elements weren’t sufficient! They covered important elements, but only some. If you just took this model’s advice, you’d have somewhat more memorable learning, but you’d fall short on the real potential impact. For instance, there wasn’t anything there about the importance of contextualized practice nor feedback. Nor models, for that matter!

I’m not allergic to n letter acronym models. For instance, I keep the coaster I was given for Michael Allen’s CCAF on my desk. (It’s a nice memento.) His Context-Challenge-Activity-Feedback model is pretty comprehensive for the elements that a practice has to have (not surprisingly). However, learning experiences need more than just practice, they need introductions, and models, and examples and closings as well as practice. And while the aforementioned elements are necessary, they’re not sufficient. Heck, Gagné talked about nine elements.

What I realize as I reflect is that I like models that have the appropriate amount of complexity for the level of description they’re talking about. Yet I’ve seen far too many models that are cute (some actually spell words) and include some important ideas but they’re not comprehensive for what they cover. The problem, of course, is that you need to understand enough to be able to separate the wheat from the chaff. I’ll suggest to look to vetted models, that are supported by folks who know, and there are criticisms and accolades to accompany them. Read the criticisms, and see if they’re valid. Otherwise, the model may be useful.

Ok, one other thing bothered me. This model supposedly has support from neuroscience. However, as I’ve expressed before, there have yet to be results that aren’t already made from cognitive science research. This, to me, is just marketing, with no real reason to include it except to try to make it more trendy and appealing. A warning sign, to me at least.

Look, designing for learners is complex. Good models help us handle this complexity well. Bad ones, however, can mislead us into only paying attention to particular bits and create insufficient solutions. When you’re looking at simple models and complex problems, you need to keep an eye out for help, but maybe it needs to be a jaundiced eye.

There’s been a lot of talk, of late, about workflow learning. To be fair, Jay Cross was talking about learning in the flow of work way back in the late 1990s, but the idea has been recently suborned and become current. Yet, the question remains whether it’s real or a mislabeling (something I’m kind of  anal about, see microlearning). So, I think it’s worth unpacking the concept to see what’s there (and what may not be). Is workflow learning a myth?

To start, the notion is that it’s learning at the moment of need. Which sounds good. Yet, do we really need learning? The idea Jay pointed to in his book Informal Learning, was talking about Gloria Gery’s work on helping people in the moment. Which is good! But is it learning? Gloria was really talking about performance support, where we’re looking to overcome our cognitive limitations. In particular, memory, and putting the information into the world instead of in the head. Which isn’t learning! It’s valuable, and we don’t do it enough, but it’s not learning.

Why? Well, because learning requires action and reflection. The latter can just be thinking about the implications, or in Harold Jarche’s Personal Knowledge Mastery model, it’s about experimenting and representing. In formal learning, of course, it’s feedback. I’ve argued we could do that, by providing just a thin layer on top of our performance support. However, I’ve never seen same!  So,  you’re going to do, and then not learn. Okay, if it’s biologically primary (something we’re wired to learn, like speaking), you’re liable to pick it up over time, but if it’s biologically secondary (something we’ve created and aren’t tuned for, e.g. reading) I’d suggest it’s less likely. Again, performance is the goal. Though learning can be useful to support comprehending context and  making complex decisions, what we’re good at.

What is problematic is the notion of workflow and reflection in conjunction. Simply, if you’re reflecting, you’re by definition out of the workflow! You’re not performing, you’re stopping and thinking. Which is valuable, but not ‘flow’. Sure, I may be overly focused on workflow being in the ‘zone’, acting instead of thinking, but that, to me, is really the notion. Learning happens when you stop and contemplate and/or collaborate.

So, if you want to define workflow to include the reflection and thoughtful work, then there is such a thing. But I wonder if it’s more useful to separate out the reflection as things to value, facilitate, and develop. It’s not like we’re born with good reflection practices, or we wouldn’t need to do research on the value of concept mapping and sketch noting and how it’s better than highlighting. So being clear about the phases of work and how to do them best seems to me to be worthwhile.

Look, we should use performance support where we can. It’s typically cheaper and more effective than trying to put information into the head. We should also consider adding some learning content on top of performance support in times where people knowing why we’re doing it as much as what we should do is helpful. Learning should be used when it’s the best solution, of course. But we should be clear about what we’re doing.

I can see arguments why talking about workflow learning is good. It may be a way to get those not in our field to think about performance support. I can also see why it’s bad, leading us into the mistaken belief that we can learn while we do without breaking up our actions. I don’t have a definitive answer to “is workflow learning a myth” (so this would be an addition to the ‘misconceptions’ section of my myths book ;). What I think is important, however, is to unpack the concepts, so at least we’re clear about what learning is, about what workflow is, and when we should do either. Thoughts?