The Incompleteness of Connectivism

Stephen has written a terrific post on connectivism as a learning theory. This is one of the briefest – and consequently, best – statements I’ve read on the subject.

Let me begin by saying that I’m a fan of connectivism. Personally, I’m inclined to be persuaded by the connectivist account as Stephen, George, and others have articulated it. But – while I haven’t read every piece written on the topic – those I have read contain a gaping hole which I feel must be addressed before the theory can be considered complete and, therefore, a legitimate alternative to longer established learning theories.

Stephen explains:

When I say of connectivism that ‘learning is the formation of connections in a network’ I mean this quite literally. The sort of connections I refer to are between entities (or, more formally, ‘nodes’)… In particular, I define a connection as follows (other accounts may vary): “A connection exists between two entities when a change of state in one entity can cause or result in a change of state in the second entity.”

In Stephen’s account, connections are defined as a kind of relationship between entities. However, I have never read a connectivist account of where entities come from, or a connectivist description of their nature. And defining an undefined word exclusively in terms of a second undefined word kicks the semantic can down the road. And building a learning theory on a term with such a definition seems “risky.”

But as I said above, this is not a critique of what has been written about connectivism – I’ve found that writing to be quite persuasive. This is simply a statement about what remains to be considered and written about before connectivism can be considered sufficiently complete.

7 thoughts on “The Incompleteness of Connectivism”

  1. I like connectivism, too. The main problem I’ve had with it as a theory is that they seem to look at the learning of an individual as forming connections with other individuals. However, learning is based on the system in which nodes are connected, so, what they’re really talking about is the learning of a network of individuals while the learning of a single individual is the connecting of brain “nodes.” Connectivists seem to conflate these two different systems.

  2. Interesting comment David. I don’t specify where the entities come from because any entity can be a part of a network. So what constitutes an entity is definied implicitly in the definition of a connection – an entity is anything that can have a changeable internal state.

    I know that’s really vague. In practice entities have many more properties, and these properties have all sorts of bearing upon how they interact with other entities for form a network. For example, the capacity of one entity to cause a change of state in another entity depends on the communications medium and the properties of each of the two entities.This supposes that there is a causal relationship between the entities. But the actual instantiation of connectivism is independent of its physical instantiation. If we ever find non-causal ways one entity can change the state of another entity, then it also is a network, and can learn.

    This addresses the point raised by Charles, above. Had he read the background literature on connectivism (ahem) he would know that we do not conflate the two types of network, but are very clear about identifying them as separate types of networks with separate types of (physical) properties. But the principles of connectivisty that apply to them apply in (eg.) the way the principles of mathematics apply to them. We don’t say that mathematics ‘conflates’ between people and neurons because it is used to count both. It is rather the same logic applied to two physical systems.

    Why would I make this claim? because these are principles of networks that have been observed in independent types of networks – from the cricket networks identified by Watts to the self-synchronizing nature of metronomes on a plank of wood. They are described, formally, by graph theory, and computationally by connectionist logic (aka ‘neural nets’). If there is anything controversial to the idea of connectivism, it is the principle that non-human entities can learn (or, conversely, the way people learn is not inherently human). But that’s an empirical question, and IMHO, the answer is becoming increasingly clear: people learn, animals learn, societies learn, computers learn, and so on and on.

  3. I hadn’t read “the literature” on connectivism in quite a while, so, thanks to Stephen, I went back and re-read some seminal articles. Stephen in his Introduction to Connectivism is much clearer about different levels of networks and learning. What troubled me at one time and still does is George’s “A Learning Theory for the Digital Age.” Although it’s mentioned that there are different levels of networks and learning, the focus seemed to be on the individual but the description of the learning was social. In the conclusion he writes,

    “The pipe is more important than the content within the pipe. Our ability to learn what we need for tomorrow is more important than what we know today. A real challenge for any learning theory is to actuate known knowledge at the point of application. When knowledge, however, is needed, but not known, the ability to plug into sources to meet the requirements becomes a vital skill. As knowledge continues to grow and evolve, access to what is needed is more important than what the learner currently possesses.”

    The “pipe” is a connector between individuals and thus allows learning at the social level. Yet this paragraph seems to be talking about what the individual needs to do to learn. Of course, the more pipes, or connections, the more experiences and patterns an “individual” can perceive. But that perception is grounded inside the individual. And it’s statements (paragraphs) like these that lead me to say that the systems are being conflated.

    And it’s not just that one paragraph. Even so, after reviewing a few of Stephen’s articles, I can see that he doesn’t conflate the systems.

  4. Great post. To pick up on something Stephen said, for me the most controversial idea of connectivisim is that it comes out of an educational perspective. While I don’t really disagree with the very interesting descriptions here, I do think that connectivism does need to specify the entities involved, if it’s going to pertain to concepts that involve Choice and Intention (ie: Education, Design, and even Learning…vs Change or Adapt). If not, it seems better suited for wider applications like complex systems, network theory, or even certain branches of philosophy, if there’s a relevant gap in these areas that connectivism can address.

  5. All we would need then is a convention whereby ‘knowledge’ is any (designated) material or intangible entity that exists for at least one individual or institution. Stephen’s sentence then reads, fused : “A connection exists between two entities when a change of state in (the knowledge of) one entity can cause or result in a change of state in the (knowledge of the) second entity.”

  6. David, I can’t help thinking: is it “completeness” that Connectivism requires or further articulation? I’m not sure that theories necessarily become complete, even though they may purport to be. In Stephen’s view, theories explain. I agree, although would go further with a qualification: explanations just need to be plausible or express some kind of coherence to function as theories. And because plausibility is contextual & subjective many theories provide useful prespective — the growing database of learning theories first developed by Greg Kearsley back in 1994 demonstrates this: though I’ve not properly understood why Connectivism does not yet appear in Kearsley’s list.

    I agree the growing discourse on Connectivism & its core insight is also persuasive — basically because of its emphasis upon networks as ubiquitous in the context of our contemporary learning. Stephen’s concise account of Connectivism is also compelling, although I think there’s a few points worth contesting or getting clarification on.

    I’m not sure that theories of learning are necessarily mutually exclusive. Stephen’s account doesn’t explicitly say that although the comparison of the “isms” is that somehow Connectivism offers a better, or at least a distinct, explanation. For me, there’s plenty of evidence that supports the plausibility of most learning theories, often depending upon context. Rote learning, for example, might be classified as instructivist (although to monks immersed in solitary rote learning there’s little instruction, there’s just reading & repeated verbalization) — but it is also demonstrably effective.

    A statement from Stephen I do like is that “A connectivist account of literacy reinterprets both syntax and semantics, looking well beyond rules and meaning”. One reason why this has resonance for me is that the emphasis upon “meaning-making” in Constructivist theories misses aspects of sense-making & seems to mask the importance of the connections.

    One aspect of Connectivism that I think warrants further clarification is how it is distinct from the Connectionist (neural network) theories that came out of AI in the 1980s & early 1990s. Perhaps that’s alrready been answered somewhere?

  7. Thanks for this contribution. I agree that connectivism is not yet sufficiently complete. In fact, after reading about it, I came up with some doubts about its “innovative ideas”, especially regarding the development of certain skills, the prioritization of some specific personal characteristics and the evaluation methods of the connectivism… Here goes the link to my text… I’m looking forward to hearing opinions…

Comments are closed.