Handling Large and Complex Subjects in Every Page is Page One

Not only is Every Page is Page One a viable way to tackle a large and complex subject, it is the only way to tackle the truly large and complex.

People regularly tell me that Every Page is Page One is fine for small simple stuff, but it won’t work for large and complex subjects. Actually, it is just the opposite. The book model will work okay for small and simple stuff, but it falls down completely when you are dealing with a large and complex information set. When things get really big and really complex, only an Every Page is Page One approach will work.

Confusion on this point is understandable. If you look at Every Page is Page One simply as a prescription for individual topics, then you will necessarily see it as dealing only with small subjects, since each Every Page is Page One topic by itself serves a specific and limited purpose.

But, just as a particular bus route in a particular city serves a specific and limited purpose in the world’s transportation network, so an EPPO topic can serve a specific and limited purpose in a very large and complex information set.

Exactly because it is so vast and so complex, the world’s transportation network could not function if it tried to provide unique direct transportation of any person or product from point A to point B on demand. It requires a system of vehicles each of which performs a specific and limited purpose in the network.

But it could not function either if it tried to run every vehicle in every fleet in every mode of transportation from one single system. It can only function through the cooperation of many independent authorities locally administering their local part and mode of the global system, and running individual vehicles to serve specific and limited purposes within their local networks.

As economist Tim Harford notes in his book Adapt,

Complexity means there are many different ways for things to go wrong. Tight coupling means the unintended consequences proliferate so quickly that it is impossible to adapt to the failure or to try something different.

Harford, Tim (2011-05-10). Adapt: Why Success Always Starts with Failure (pp. 185-186). Doubleday Canada. Kindle Edition.

The solution to such fragile systems is decoupling. Decoupling means that a system is designed so that the failure of one part does not cause the failure of other parts. If the world’s transportation networks were tightly coupled, a train derailment in Virginia would cause planes to be grounded in Moscow, ships confined to harbor in Singapore,  and road traffic to be halted in Madrid.

But because the parts of the grid are not closely tied to each other, shutting down one does not interrupt the flow of the others. The loose coupling of the system enables most of it to stay in operation even though several parts of it are broken down at any given time.

The Internet is one of the most highly decoupled systems in the world — it was designed to survive and continue operating in times of war when major parts of its infrastructure might be destroyed. The Web inherits this decoupled nature, which is why, while individual sites may go down, the Web itself never goes down. Hypertext provides a way to navigate the loosely coupled content of the Web.

Every Page is Page One is not about creating individual articles, but about creating loosely coupled information sets. A book is a tightly coupled information set. It lays out a single route for every reader to follow. Each part of it depends on what comes before and what follows after. But if any reader becomes blocked on that route, the book offers them no way to detour around the blockade.

Fortunately, while individual books are tightly coupled, the world’s information systems are not, and the reader who is blocked in one book can look for other information sources to help them get going again. Thus readers seldom follow the prescribed path laid down by authors, but hop about from one information source to another following the information scent that appeals to them in the moment. This is why Every Page is Page One is not simply a design choice, but first and foremost a fact about how people consume information, particularly on the Web, but in off-line content as well.

As John Carroll observed, in studies conducted long before the Web:

Learners … often skip over crucial material if it does not address their current task-oriented concern or skip around among several manuals, composing their own ersatz instructional procedure on the fly….

Many sequencing problems reside not in the material alone but in the learner’s use of it. When people refer to instruction opportunistically in support of their own goal-directed activities, it becomes difficult or impossible to predict what sequencing will be appropriate…

[M]aterials designed to be read in any order cannot be read in the wrong order. … The orderly accumulation of prerequisite skill and understanding that can be assumed when material is embedded in a sequenced curriculum cannot be assumed if learners use the material in any order they wish. But, of course, this is just what learners do anyway and is one of the key reasons that materials that depend on carefully sequenced prerequisites fail.

The problem of correctly anticipating the order in which people will want to read content grows in proportion to the size and complexity of the subject matter. For a small subject like a recipe, the established order of the recipe topic type — essentially: name of dish, description, ingredients, preparation steps — works very well. Even readers who don’t want to read in that order can easily read in the order they wish because the form is consistent and the model familiar. Thus the hierarchical approach works fine at the small scale.

But expand the scale, and add more complex relationships along different axes, and anticipating sequence becomes increasingly difficult. And, as Carroll points out, there is no one perfect sequence to be found, because each reader wants to read in their own sequence driven by their personal goals and the personal background.

This is why we find clear topic types emerging in many forms of short content. Wikipedia is the perfect place to study topic types as its articles on many different kinds of subjects, from cities, to movies, to countries, to diseases, to car models, to a thousand other things, each tend assume a common shape, presenting similar information fields in a similar order.

In books, on the other hand, no such native topic typing occurs. The scale of a book is too large to allow for simple logical types (except in reference works, which are actually collections of smaller units). The larger the scale, the more arbitrary the ordering of content becomes as the author tries to pick the best path through the content out of the many competing approaches that might fit the material in different ways.

But while the reader can tolerate the writer’s imposition of an arbitrary order in a book, particularly if there is a strong narrative thread or a strong thread of argument running through it, it does not scale up to larger subjects.

The early encyclopedia writers tried to construct their works according to a tree into which they tried to map all of human knowledge. Invariably, each of these tree schemas was different, and all of them proved inadequate. The attempt was eventually abandoned and encyclopedias were then simple organized in alphabetical order. In the 21st century, even this has been abandoned. Wikipedia is not in alphabetical order. It is not in any order. Wikipedia articles are richly connected, but they are not ordered relative to one another. There is no top down structure to Wikipedia; it is all bottom-up.

Their are a number of reasons why top down categorization fails for content. The most obvious is that categories and the terms within them are often unfamiliar to the reader. The less obvious, but ultimately far more important reason is that content by its very nature does not fit into neat categories. This is because the real world does not fit neatly into these categories either.

The Manicouagan crater in Quebec is both a geological phenomena and an astronomical one. It is a recreation zone, tourist attraction, and a reservoir. It lies within multiple Quebec municipalities. It is an annular lake, and has been studied by scholars from multiple universities. It may be associated with a major extinction event and may be part of a multiple-impact event, the remnants of which span continents, and have shifted locations relative to one another thanks to continental drift.

People in half a dozen academic fields and with a variety of political or recreational interests may want information on the Manicouagan crater. Each will seek that information through the lens of their own goal-driven interest, and the categories and vocabulary of their own field. And yet, each will also have some degree of interest in the other aspects of the crater and its life and existence, since they will influence the use they want to make of it in some way or another.

Content about the Manicouagan crater, like the Wikipedia article on it, will therefore lie at the intersection of may different categories and vocabularies. Indeed, it is the function of content to deal with these intersections of categories, and all the complexities that arise from the fact that they intersect in a particular way at a particular point at a particular time.

Where categories line up neatly and intersect in neat and orderly ways, we don’t need content to describe their interaction. A database will suffice. The relational database model, indeed, is designed specifically to handle the orderly and predictable intersection of categories. It is when the intersections become tangled and messy that you need to create narrative content to describe what is going on.

This is precisely what constitutes complexity in content: the tangled intersection of many different categories and vocabularies. The thing that makes content complex is precisely the thing that make top-down organization, particularly hierarchical organization, ineffective.

Look at the Wikipedia article on the Manicouagan crater and you will see that it places the crater on all of these axes of categorization and vocabulary. In itself this is unremarkable. That is just the ordinary work that narrative content does. But in Wikipedia, those tangled connections are linked, so that you can follow any of these lines of categorization (lines of subject affinity) that interest you, based on your particular goal-driven reason for being there. You will also find associated metadata that further locates the article in its many intersecting categories.

There is categorization and metadata at work here, but rather than seeing one whole category at a time, as you do when metadata is used for top-down organization, what you see in the Wikipedia entry is the particular points of intersection between the category and the subject matter of the article. The article lies at the intersection point of individual items in multiple taxonomies. Thus they place the article, and thus that reader, at the particular point of intersection of all these categories, and by linking on them, make them navigable.

This is what an Every Page is Page One information design and an bottom-up information architecture can do well, and what top down and hierarchical approaches cannot handle: to assemble and make navigable large quantities of complex information.

As Tim Harford notes:

complexity is a problem only in tightly coupled systems.

Books are fine for the small stuff. For large and complex information systems, you need to avoid tight coupling; you need Every Page is Page One.