Trees vs Networks

A new metaphor for organising knowledge is emerging.

Trees have inspired how people understand the world around them for millennia. The worship of trees, known as dendrolatry, is present in practically all cultures and their ability to represent spreading connections has inspired many to adopt their form for a wide variety of knowledge representations. The tree metaphor can be seen in everything from the religious illustrations created by Joachim of Fiore, the biological classifications of Aristotle and Darwin, and your own lineage represented as family trees.

(L-R) Tree of Life by Joachim of Fiore C12th; Tree of Virtues by Lambert of Saint-Omer, 1120; Genealogical tree by Paul Veronese, 1568–73; Tree of Life by Ernst Haeckel, 1879

The root of it all

There is no doubt this versatility has allowed the metaphor to thrive. It’s also nice and easy to understand — trunk, branches, leaves — we get it. Let’s look at Melvil Dewey’s Decimal Classification system which has been in persistent use since 1876. It works on a tree structure of ten subject classes that are then split into ten divisions, each containing ten sections. Further granularity comes after the decimal point. The Decimal Classification allowed libraries to organise new books by subject rather than when they were acquired.

A treemap demonstrating pigeon-holing in the Dewey Library Classification system

Although immediately effective, this is extreme pigeon-holing, and does not recognise how materials can easily be relevant to multiple topics. So whilst the simplicity of trees and their linear categorisation of the world allows us to comprehend complex information through order, tree structures can be confining and are not a fair reflection of reality. That does not stop us using them though. It is a clear and convenient method to map direct connections or flows and will always have a place in visualising information.

It’s not as easy as that

Upon closer inspection however, we are now realising that the systems we once thought were simple and sequential are in fact much more intricate. The world has always been complex and as our understanding of it becomes more nuanced, these underlying structures are gradually revealing themselves to us in ways that challenge the pristine order of trees.

The North Atlantic cod illustrates this perfectly. We once believed the food chain for this fish was pretty clear cut. We knew what it ate and what ate it. As it turns out the life of a humble cod involves interaction with hundreds of other species in a broad ecosystem. The classic food chain, a tree metaphor, falls significantly short and fails to take into account the many connections needed to represent and sustain cod in their habitat. The world, it turns out, is often better understood as a complex network.

North Atlantic cod’s complex ecosystem. Still taken from an RSA Animate of Manuel Lima’s The Power of Networks lecture.
“When everything is connected to everything else, for better or for worse, everything matters.” ◦◦◦ Bruce Mau, designer

Everything’s connected

As we increasingly spot more and more of these networks around us, we also recognise that they appear to run within an organised complexity — a notion first coined by scientist Warren Weaver. He argues that disparate parts of a system do, on many levels, still interact with each other. These relationships create interdependent systems, that can in turn interact with other systems.

For example, Charles Darwin’s Tree of Life, a hierarchy of evolution, isn’t as successive as he might have thought. Evolution probably began as the Net of Life, a concept promoted by biochemist Ford Doolittle, who argued that because the same bacteria existed within different species, they were directly connected. The map loses its arboreal shape, becoming instead a dense mesh of interconnections.

a. Darwin’s first sketch of the Tree of Life, 1837 — b. Stylisation of Doolittle’s Net of Life, 2000

The network works

The more that systems are considered networks, the greater opportunity there is to effectively map and depict them. Manuel Lima, a designer of information visualisation who often writes and gives talks on this particular subject, has popularised the network’s role in illustrating relationships, reminding us that complexity needs at least three dimensions to be mapped.

Consider visualising Wikipedia in a tree structure. There are so many direct and lateral references connecting one article to another that a tree hierarchy would not be fit for purpose. It needs the vertices (nodes that represent articles) and edges (lines that represent relationships) to identify the multiplicity of the system. The complexity is immense. The diagram below merely shows art history Wikipedia entries and their interconnections.

Relationships between art history articles on Wikipedia, by Arends, Froschauer, Goldfarb & Merkl

Networks! They’re everywhere!

Not only is the network a powerful visual identity, it is also a universal structure that comprises many elements of life. Networks are seen at a molecular level, in brain patterns, when mapping social connections, plotting transport infrastructure or power grids and of course in the links of the internet. Manuel Lima even saw strong similarities between the neural pathways of a mouse and the historic growth of cosmic structure across 20 million galaxies.

Neuronal network in the brain of a mouse, by Mark Miller (left) and a simulation of the growth of cosmic structure, by the Virgo Consortium (right)

At Bibblio we have been investigating networks, with a particular emphasis on ontologies, and how they can influence and improve discovery in online learning and information seeking.

Here are five features of networks that we think are worth considering:

Reliance on other networks, or between the data within a network, is common in both nature and technology. Adjustments to one element can have knock-on effects on the performance or relationships of other elements.

Although a piece of content is primarily a standalone object, it is very likely to be connected on some level to other objects with similar properties. It is these connections that create a discovery dependency between each data node, which is crucial in achieving interactions through content relationships. When content changes, the connections to other content will, to varying degrees, change too.

Ever thought about an educational curriculum in which all subjects were connected dynamically? Networks allow learners to explore the surprising connectedness and multiple degrees of separation between disparate topics. Jumping from plant biology to music, or algebra to sociology, reinforces the holistic nature of existence on both a technical and philosophical level and offers cross-cutting opportunities.

The internet is undoubtedly the network which has had the biggest impact on society and it is a perfect example of a decentralised network. Data is distributed across the network without a reliance on any particular hub or central hierarchy.

However, in recent years the internet has become more centralised as the majority of data flows through a small group of corporations. With it, control has also become increasingly concentrated, to the detriment of genuine long-tail distribution and discovery online.

Nothing in nature is as linear as it seems. Just take the weather as an example. It is, like other systems around us, inherently nonlinear. Learning is another perfect illustration of a nonlinear system. Our brain connects the dots spatially, conceptually and temporally, rather than just following a straight line of progress. We cannot simply measure the educational value of a book or lesson to an individual purely based on the volume of information it contains.

By increasingly emulating how the brain works, creating and expressing connections, networks can aid discovery and connect people with new concepts by allowing similar nonlinear movement through subjects.

Networks are collections of data that can exponentially yet comfortably increase in quantity and complexity. Some elements have more connections than others and greater depth of detail, whilst others might be abstract or reliant on only a few key associations.

Whether content links to many other items or just a few, based on its complexity or popularity, an ever-expanding network of knowledge can always accommodate it, affording learners a greater range of discovery and exploration. The world of networks is round, not flat.

A word of warning

Although the network is a better symbol for a complicated system, it is still a symbol — an abstraction of reality. “The map is not the territory” as Alfred Korzybski famously remarked, meaning the model is only ever a representation of something, not actually the object itself. And with this model there is no guarantee that the object can be explicitly understood. The network metaphor does not make the complexity of the system any simpler. If anything, it’s more difficult to reason about it or anticipate consequences, since its intricacy is manifold.

This Gordian Knot cannot be easily cut.

The clouds are gathering

The problem remains that an estimated 80–90% of online materials exist in an unstructured form, meaning they are not connected, tagged or curated in any meaningful way. Literally, they have yet to find a place in an organised network.

As a result, there is a significant focus on machine learning and artificial intelligence, or cognitive computing, as IBM likes to call it. Could it be that we are entering a new ecology of knowledge, powered by machines that can understand meaning and represent it in vast connected networks and help us find the right answers quicker? Given the exponentially growing quantities of information saturating our digital lives, this new cognitive era couldn’t come soon enough. Countless attempts have already been made to bring this information together in open networks to improve accessibility.

Freebase was one such collaborative effort that created and maintained a community-curated database of “people, places and things”. Google bought the operating company in 2010 and used the technology as part of their search engine’s Knowledge Graph, a semantic-search store that was recently succeeded by their own next generation, Knowledge Vault.

Google’s Knowledge Vault: The end of open collaboration in learning material networks?

In a worrying trend, Google are busy hiring all of the talent in this space to ensure the Knowledge Vault becomes a supreme yet hermetic network solely for their own use. The original spirit of open collaboration is disintegrating as knowledge base production is becoming an increasingly insular arms race. Giants such as Facebook, Yahoo! and Baidu are busily putting up their own walled gardens too. This does not bode well for content publishers, who stand to lose further control of their value chain as the web companies vigorously assert their new-found power.

A new hope

We believe that open collaboration is crucial to innovation in the education technology space, particularly when structuring big networks of content. Non fiction publishers should have the opportunity to enhance and deliver their catalogues of learning materials effectively and at scale. Likewise, learning platforms should have access to quality, structured content, powered by clever search and recommendations.

This last point is where networks come into their own. To ensure people discover highly relevant and contextually similar content, a network of data must know how every item is aligned to one another. At Bibblio we’re building this within our Smart Graph, an ontology that maps content based on semantic fingerprints of each object.

Bibblio’s Smart Graph, an ever-learning knowledge graph.

Making your own luck

Manufacturing serendipitous discovery with this new powerful data encourages people to actively guide their own learning journeys. This helps them enter the realm of exploratory learning, which can improve their development in three ways:


Exploratory learning allows opportunities for multimodal learning. People can engage with multiple representations of meaning through various media, transferring patterns and behaviours from one context to another.


The challenges set through exploratory learning pose a “desirable difficulty” or “productive failure” that encourages people to rethink their previous conceptions and develop a deeper understanding. This helps boost the instruction they might later receive from teachers or colleagues.


Individuals can specialise and pursue their interests, using innate curiosity to carry them and ultimately feel a sense of pride and ownership from the knowledge they have acquired. Read much more about self-determined learning and its benefits in our previous post, Education vs Learning.

Although research has demonstrated that exploratory activities and self-determined learning indeed works, everyone is different and no learning situation is ever quite the same. In some cases exploring a tree structure and the intuitive scaffolding it provides may be preferable, whereas in others a more complex network and its many possible connections is the best solution for cultivating wisdom.

Rich Simmonds, Co-founder
Mads Holmen, Co-founder
Robbert van der Pluijm, Learning Coordinator

Making knowledge more discoverable

Bibblio is the intelligence between the content and the user, delivering smarter search, discovery and recommendations. Visit us on Twitter and Facebook.

More juicy posts by Bibblio:
Clicks vs Satisfaction
Popularity vs Diversity

Education vs Learning
Search vs Discovery
60 YouTube Channels that will make you smarter

If you enjoyed this post please press the recommend button so more people can as well!

Published in Higher Education Revolution