For a few minutes the other day I thought that I had invented a new kind of fallacy or, at least, a great term to describe it. Disappointingly, a quick search revealed that it was not only an old idea but one that has been independently invented at least twice before (Berry & Martin, 1974; Weinstock, 1981). Here is its definition from Weinstock (1981):
“a synecdochic fallacy is a deceptive, misleading, erroneous, or false notion, belief, idea, or statement where a part is substituted for a whole, a whole for a part, cause for effect, effect for cause, and so on.”
Most synecdoches (syn-NEK-doh-kees in case you were wondering – I have been getting it totally wrong for decades) are positively useful. Synecdoches make aspects of a whole more salient by focusing on the parts. No one, for instance, thinks “all hands on deck” actually means the crew should put their hands on the deck let alone that disembodied hands should crew the ship, but it does focus on an aspect of the whole that is of great interest: that there is an expectation that those hands will be used to do what hands do. Equally, synecdoches can make the parts more salient by focusing on the whole. When we say “Canada beat the USA in the finals” no one thinks that one literal country got up and thrashed the other, but it draws attention to a symbolic aspect of a hockey game that reveals one of its richer social roles. It becomes a fallacy only when we take it literally. Unfortunately, doing so is surprisingly common in research about education and educational technologies.
Technologies as synecdoches
The labels we use for technologies are very liable to be synecdochic (syn-nek-DOH-kik if you were wondering): it is almost a defining characteristic. Technologies are assemblies, and parts of assemblies, often contained by other technologies, often containing an indeterminate number of technologies that themselves consist of indeterminate numbers of technologies, that participate in richly recursive webs of further technologies with dynamic boundaries, where the interplay of process, product, structure, and use constantly shifts and shimmers. The labels we give to technologies are as much descriptions of sets of dynamic relationships as they are of objects (cognitive, physical, virtual, organizational, etc) in the world, and the boundaries we use to distinguish one from another are very, very fluid.
There is no technology that cannot be combined with different others or in different ways in order to create a different whole. Without changing or adding anything to the physical assembly a screwdriver, say, can be a paint stirrer, a pointer, a weapon, or unprestatably many other technologies, far from all of which are so easily labelled. Virtually every use of a technology is itself a technology, and it is often one that has never occurred in exactly the same way in the entire history of the universe. This sentence is one such technology: though there may be lots of sentences that are similar, the chances that anyone has ever used exactly this combination of words and punctuation before now are close to zero. Same for this post. This post has a title: that is the name of this technology, though it is a synecdoche for… what? The words it contains? Not quite, because now (literally as I write) it contains more of them but it is still this post. Is it still this post when it is syndicated? If the URL changes? Or the title? Or if I read it and turn it into podcast? I don’t know. This sentence does not have a name, but it is no less a technology. So is your reading of it. So is much of what is involved in the sense you are making of it, and that is the technology that probably matters most right now. No one has ever made sense of anything in exactly this way, right now, the way you are doing it, and no one ever will. The technosphere is almost as awesomely complex as the biosphere and, in education, the technosphere extends deep into every learner, not just as an object of learning but as part of learning itself.
Synecdoches and educational/edtech research
Let’s say you wanted to investigate the effects of putting computers in classrooms. It seems reasonable enough: after all, it’s a big investment so you’d want to know whether it was worth it. But what do you actually learn from doing so apart from that, in this particular instance, with this particular set of orchestrations and uses, something happened? Yes, computers might have been prerequisites for it happening but so what? An infinite number of different things could have happened if you had done something else even slightly different with them, there are infinitely many other things you could have done that might have been better, and all bets would be off if the computers themselves had been different. The same is equally true for what happens in classrooms without computers. What can you predict as a result? Even if you were to find that, 100% of the time until now, computers in classrooms led to better/worse learning (whatever that might mean to you) I guarantee that I could find plenty of ways of using them to do the precise opposite. This is functionally similar to taking “all hands on deck” literally: the hands may be very salient but, without taking into account the people they are attached to and exactly what they are doing with those hands, there is little or no value in making comparisons. Averages, maybe; patterns, perhaps, as long as you can keep everything else more or less similar (e.g. a traditional formal school setting); but reliable predictions of cause and effect? No. Or anything that can usefully transfer to a different setting (e.g. unschooling or – ha – online learning)? Not at all.
Conversely but following the same synecdochic logic we might ask questions about the effectiveness of online and distance learning (the whole), comparing it with in-person learning. Both encompass immense numbers of wildly diverse technologies, including not just course and class technologies but things like pedagogical techniques, institutional structures, and national standards, instantiated with wildly varying degrees of skill and talent, all of which matter at least as much as the fact that it is online and at a distance. Many may matter more. This is functionally similar to taking “Canada beat the US” literally. It did not. It remains a fallacy even if, on average, Canada (the hockey team) does win more often, or if online and distance learning is generally more effective than in-person learning, whatever that means. The problem is that it does not distinguish which of the many millions of parts of the distance or the in-person orchestration of phenomena matter and, for aforementioned and soon-to-be-mentioned reasons, it cannot.
Beyond causing physical harm – and even then with caveats – there is virtually nothing you could do or use to teach someone that, if you modified some other part of the assembly or organized the parts a little differently, could not have exactly the opposite effect the next time you do or use it. This sentence, say, will have quite different effects from the next despite using almost the exact same components. Almost components effects next the despite using different quite will sentence, say, this have the from exact. It’s a silly example and it is not difficult to argue that further components (rules of grammar, say) are sufficiently different that the comparison is flawed, but that’s exactly the point: all instantiations of educational technologies are different, in countless significant ways, each of which impacts lots of others which in turn impact others, in a complex adaptive system filled with positive and negative feedback loops, emergence, evolution, and random impacts from the systems that surround it. I didn’t actually even have to mix up the words. Had I repeated the exact same statement, its impact would have been different from the first because something else in the system had changed as a result of it: you and the sentence after. And this is just one sentence, and you are just one reader. Things get much more complex really fast.
In a nutshell, the synecdochic fallacy is why reductive research methods that serve us so well in the natural sciences are often completely inappropriate in the field of technology in general and education in particular. Natural science seeks and studies invariant phenomena but, because every use (at least in education) is a unique orchestration, technologies as they are actually enacted (i.e. the whole, including the current use) are never invariant and, even on those odd occasions that they do remain sufficiently similar for long enough to make study worthwhile, it just takes one small tweak to render useless everything we have learned about them.
All is not lost
There are lots of useful and effective kinds of research that we can do about educational technologies. Reductive science is great for identifying phenomena and what we can do with them in a technological assembly, and that can include other technologies that are parts of assemblies. It is really useful, say, to know about the properties of nuts and bolts used to build desks or computers, the performance characteristics of a database, or that students have persistent difficulties answering a particular quiz question. We can use this information to make good creative choices when changing or creating designs. Notice, though, that this is not a science of teaching or education. This is a science of parts and, if we do it with caution, their interactions with other parts. It is never going to tell us anything useful about, say, whether teaching to learning styles has any positive effect, that direct instruction is better than problem based learning, or that blended learning is better than in-person or online learning, but it might help us build a better LMS or design a lesson or two more effectively, if (and only if) we used the information creatively and wisely.
Other effective methods involve the telling of rich stories that reveal phenomena of interest and reasons for or effects of decisions we made about putting them together: these can help others faced with similar situations, providing inspirations and warnings that might be very useful. If we find new ways of assembling or orchestrating the parts (we do something no one has done before) then it is really helpful to share what we have done: this helps others to invent because it expands the adjacent possible. Similarly we can look for patterns in the assembly that seem to work and that we can re-use (as parts) in other assemblies. We can sometimes come up with rules of thumb that might help us to (though never to predict that we will) build better new ones. We can share plans. We can describe reasons.
What this all boils down to is that we can and we should learn a great deal that is useful about the component technologies and we can and should seek broad patterns in ways that they intertwingle. What we cannot do, neither in principle nor in practice, is to use what we have learned to accurately predict anything specific about what happens when we put them together to support learning. It’s about improving the palette, not improving the painting. As Longo & Kauffman (2012) put it, in a complex system of this nature – and this applies as much to the biosphere, culture, and economics as it does to education and technology – there are no laws of entailment, just of enablement. We are firmly in the land of emergence, evolution, craft, design, and bricolage, not engineering, manufacture and mass-production. I find this quite liberating.