Design Driven AI
Entropic Designentropicadjective1. The definition of entropic is having a tendency to change from a state of order to a state of disorder.
Here is a quick glimpse into the design process as it is today.
Firstly, when starting out designing a site that adapts to content there is an obvious area lacking — the content. Traditionally you establish the content up front in order to direct the design and presentation. Without it the possibilities and permutations are endless.
We have deep dived into the potential permutations of content types; small text, large text, titles, subtitles, video, imagery, media and all the combinations that are possible by mixing a few or all the elements together. It is both laborious and seemingly endless.
This is what many of the design frameworks, themes, and UI kits provide, to an extent, today. A laborious list of workable templates that can accommodate content in a presentable fashion. Unique, unorthodox or unexpecting content then has to be retro-fitted which then quickly starts shaping sites that look ‘samey’.
The focus is in the wrong area. Currently we focus on containers for design. Not the forces that drive the design.
This shift in focus revealed that a few fundamental laws and adding more and more constraints can produce remarkable results. Yet, still accommodate more unforeseen permutations that is simply not available with current design kits or design frameworks.
When designing for a site like thegrid.io there is still the same job of generating a lot of mockups but there is a big difference in intention. They are not meant for replication or colour by number in developers
Firstly, the designs are purely for the purpose of research and development. A communication and investigation tool to extrapolate the core fundamentals and highlight the constraints implicitly being applied to the designs.
Secondly, there is a job of articulating those fundamentals, intentions and constraints into a format that the AI can understand. We translate design intention into an algorithmic format that can then be used to generate desired outputs based upon highlighted constraints.
We channel the designs using GSS (Grid Style Sheets) to produce thousands of variations of content presentations based on the combinations of our explicit constraints. Producing in minutes what a single designer could do in, at best, months.
Thirdly, for the designer, there is an art direction phase of looking through the output generated. There is an inevitable move from low to high entropy as we generate thousands of variations. However, we can add more constraints again to mark which outputs are required, which are strong and which are weak. Over time, this process becomes more and more refined letting the best presentations bubble to the surface.
Lastly, we take our findings and feed that back into the AI; adding more constraints, more refinement and more polish until we are left with a layout filter that can accommodate more content types and variation than any UI kit could offer.
The job of the designer and developer are not going away. They are evolving, adapting and changing to become even more relevant to the unique requirements and needs of every individual.
At the core, humans are needed to drive the AI, drive the direction and the vision for each person's unique slice of the internet. If you are interested in helping or wanting to know more, let me know.
In the meantime here are some key changes in approach that I have already noticed in the last twelve months.
The designer workload is not reduced. It is refocused. Time is saved from menial tasks and applied directly to efforts of significance and value to the project in hand.A healthy acceptance to giving up absolute control of the output. Not to be confused with compromise or total control. Nor pixel perfection.There is less focus on a traditional website “page” and more focus on the elements, components, and sections that come together to make up these pages. Ultimately, a purer focus on the content and how it leads design.Strengths in relationships between elements, components, and sections are much more important to be explicitly expressed. Are they required, strong or weak — just like in GSS. Think beyond media queries, breakpoints and viewports. It becomes more granular and focused on the content.The ability to articulate and communicate design intention in an algorithmic way is essential. This stems more from the ability to really understand why you design the way you do, not the ability to code. Those colors, typefaces, grids, white-space, layout — keep asking yourself “why?”.A deeper understanding of color, type, geometry; in fact all aspects of design and how they impact human emotion, mood and reactions are hugely beneficial.Focus and foresight are needed in the constraints we use, guide and provide. Understand the ripples you create with edits, additions, and changes. These ripples become the foundations on which we scale design exponentially.Perceived small tweaks i.e. change in typeface, can cause the most drastic change in tone and mood conveyed. Be purposeful and extremely selective.A mind shift in providing systems beyond sets (or absolute values) for colour, type, grids, shape, affordance, icons et al.
Fundamentally we are providing a solution, not a result. A framework, not a static site. A guideline, not a rulebook.
The main focus is to set a stage so you can compose an orchestra — rather than play each instrument individually. A way to deliver music that feels magical, to the masses.
Yes, the approach is different, the focus is different, it is a paradigm shift beyond the norm of today. To alter our approach and focus is no small feat yet the potential is once again exciting, intoxicating and full of possibilities.
At the heart of it though, it is still human-centered; by humans for humans. The potential with assistive AI is that together we can raise the bar for everyone online — not just for the select few that can.