Balancing textual agency between humans and machines
Typography is based on a notion of encoded experiences, meaning that the typeface in which a message is set becomes part of the message itself. A font absolutely in tune with the text — comfortable legibility, as Jan Tschichold called it — is the benchmark for all typography. And the question of how you ensure a comfortable reading experience still fascinates me.
Jan Tschichold was one of the many who talked about the importance of working with typefaces that are in spiritual accordance with our times. But in a time defined by a constant stream of information, proclaiming one typeface or style over another proves to be a challenging pursuit. In an never-ending stream, universality seems to lose its rule. This is a time of circumstantial solutions, a time of contextual thinking. And this is where, I think, the concept of generative typography comes in.
With one foot firmly lodged in parametric design (or by its first name variational geometry) and the other suspended in potential futures, generative typography, at the moment, hangs in a limbo.
The idea is fairly simple: in the old days, designers had a set of punches and chisels, now they have a mouse and a keyboard. So instead of being carved into metal, the typeface is now being generated by a set of rules or algorithms which can churn out a variety of forms. And when you factor in that those algorithms can now tap into the network through APIs, the possibilities seem endless. And what amazing possibilities they must be. It used to be all about geometrical rules, now typefaces can be tied to environmental, social and individual activities feeding from networked information.
However, all potentially generative systems can quickly go into an exponential “runaway” if uncorrected. What keeps them in check, in a stable state, is a sort of governing loop — the parameter. And in the case of typography, what usually defines the relationships between parameters is the shape of the letter themselves.
Inherited from architecture, where parametric design has a richer history, this preoccupation for form is exactly what keeps generative typography planted within a niched area. At the moment, generative typography is all about futuristic form-making and experimenting: posters, book experiments, interactive installations, and conceptual artworks. It’s still very much a playground for new technologies.
But it doesn’t have to be.
“Every piece of typography which originates in a preconceived idea of form, of whatever kind, is wrong.”
The great misunderstanding of generative typography is to believe that it is only the content and the appearance that matters, while its true potential — comfortable legibility — remains unrealised. Its function — readability, its essence — clarity, is what merits further discussion. And what I propose is we bring systems thinking into generative typography.
Good typography catches the spirit of a text and stays loyal to it. It’s like a perfect speaking voice, it neither overshadows nor patronises the content. How the text speaks to us, how it is presented to us determines how we approach it. And unconsciously, we demand the shape to which we have been accustomed to. Simply put, when reading any text, the typeface is vital to the clear transmission of ideas.
That can have tremendous implications for — since this article is published on Medium, say — a publishing platform, where the wide range of topics and writing styles is bewildering. Beautiful typography is essential, but a single typeface is simply not enough. And certainly, designers can’t design every piece of writing on the platform.
So, to ground this discussion into more practical terms, if every text requires a certain typographical atmosphere, why not generate that atmosphere with the help of algorithms?
I think there’s a new kind of product emerging in that space and I wanted to see what it would take to put it together. So if you will, let your imagination run along these following lines as I will attempt to describe an integrated vision and a few speculative details.
Getting to that comfortable reading experience means starting with the thing that creates it in the first place: writing. When you write, words have value. Arguably, that value can be represented or encoded in a system. And it is that system that is at the heart of it all.
As you type, algorithms can analyse the text in real time, scanning, identifying and constantly matching keywords to sentences and paragraphs — in short, context — in order to find the best possible typographical match. They go scouring through every word to look for indicators that can be fed into that value system. Language processing algorithms, data detectors, intent understanding algorithms come together to align with the semantics of the text. A sort of semantic snap-to-grid.
Based on linguistic style interpretations, the system snaps to those typographical matches that make the most sense depending on the text you’re writing. The algorithms can learn to distinguish between news and personal pieces, between political and fiction, and ultimately between any two styles. They can then automatically set the type for you. The whole idea is about bringing a certain level of intelligence at the interface, using Tom Gruber’s words.
Basically, as you write, the typeface changes in real-time to match the style of your writing. No superfluous, visual interface. Your writing is the interface. You want a more ragged look, write more raggedly. If you’re looking for a soft look, tone down your voice. The system is there only if you need it, acting as your own aesthetic assistant, if you will.
In this scenario, generative typography becomes a real-time visualisation of context. The system reflects your writing style and adapts to it. You can see the changes happening and you can shape them. This idea of systems that show their context, proposed by Paul Dourish in 2004, is one I wholeheartedly agree with.
Under the hood, the architecture would be based on a declarative model of “which typeface is good at what?”, connected to different APIs, mapped to domain models and domain models mapped to language.
A lot of services are available on the web with structured data over APIs. For relevant services, designers can declare and model the capabilities.
This is an environment with new possibilities. Designers don’t just fiddle with fonts anymore. Nor do they design an overarching style or proclaim an universal solution. They can also design the system that intelligently delivers the best possible typography for any given piece of writing. It’s a fine balance of form and function.
Good typography — again, in Jan Tschichold’s words — acts as a “tactful servant to the written word”. And maybe, just maybe, the day in which we give new meaning to those words is closer than we thought.
Just imagine what a partnership between Medium and — for instance — Typekit would mean for the future of typography and writing. I can’t. But I would love to find out.