Can taste be taught?

And can an artificially-intelligent system be the one to teach it? This is a question we ask ourselves every day as we continue building an AI-powered online logo maker and graphic design platform from the ground up.

But the future of AI is in exactly this sort of subjectivity.

AI is already moving toward more qualitative decision-making, as it continues to make strides in the worlds of art and design. Algorithm-generated art even made headlines when a work of art “painted” by AI sold for over $400,000 at a Christie’s auction. While this painting is just one instance, it raises the question of whether AI deserves a place in the domain of artists and designers.

The data scientists and engineers working on this question might just have an answer. If AI is going to work alongside designers, it needs to understand the principles of design.

This solution raises a few issues itself. Namely, how exactly do you go about teaching these design principles to AI? Teaching an AI-powered design platform to make better design decisions is showing us how these principles might one day form the basis of an artificially intelligent designer, one with its very own artificially intelligent sense of taste. 

What are the principles of design?

Before we can discuss the process of teaching design principles to an AI, we need to discuss exactly what those principles are.

Contrary to popular belief, designers don’t operate entirely on instinct. Behind all the split-second decisions about what looks good and what doesn’t, most designers can explain those feelings logically, using the principles of design.

Not to be confused with the elements of design, like color, shape, texture, and line, the principles of design are concerned with the arrangement of these elements. When designers evaluate whether something “works,” these principles provide the framework for judging the quality of visual design. 

They include balance, emphasis, movement, pattern, repetition, rhythm, variety, and unity. In the same way that designers use these principles to guide their decisions about what looks good, AI could learn to do the same.

Obviously, getting AI to understand the principles of design isn’t as easy as telling it “Look for contrast, emphasis, and unity in this work.” You need to teach it what those concepts mean. But by breaking down the principles of design into a series of smaller decisions, AI could learn them from the ground up. 

How an AI algorithm makes design decisions

While we’ve mostly looked at AI and design from a theoretical level so far, this idea is far from pure theory. Today, AI is already learning to make the kind of complex design decisions that form the basis of design principles.

Let’s take a look at a few design algorithms in action.

In all of these examples, the AI in question is leveraging data from hundreds of popular logos across different industries and finding commonalities for what makes these logos look good. By isolating different trends in logo designs, the AI can learn how to apply these rules on new logos, as they’re being created.

In short, we’re watching AI create its own set of design rules to work by.

In this first design, we can see the algorithm choosing to pair the style of the symbol with the style of the font. Across popular logos, this kind of continuity between weight, rounding, and color for both symbols and fonts are a popular way to make a logo more cohesive – something the AI learned to apply.

This kind of coordination is also one aspect of the design principle of repetition. Repeating certain elements makes a piece feel more like a unified whole, since its parts seem to “go together” visually. As the AI reviews more designs, it’ll be able to understand and apply more kinds of repetition, including complex designs like patterns.

For this logo design, the algorithm has picked up on something that’s equal parts practicality and visual appeal, and that’s the tendency for logos with long names to stack words. But beyond learning that this particular company name should be stacked to suit design tastes, it also has to understand how to properly arrange the words and space the letters. 

Together, these choices uphold the design principle of emphasis, since stacking the company name contributes to the visual hierarchy of the logo. By placing focus on the company name over the slogan, the AI is learning how and what to emphasize within a design.

When clients ask designers to make things “pop,” they’re usually referring to the design principle of contrast.

Here, the algorithm is starting to generate contrast in designs by selecting background and font colors that meet a minimum contrast value. By giving itself limits based on what it sees in popular logo design, the AI can decide which color combination will most likely catch the viewer’s eye.

While the AI here has learned how to make a diverse set of design decisions, the rules it has created based on popular logo designs only touch on small areas of much larger design principles. But as AI begins to learn from the sum of all these self-taught rules, the principles of design could start guiding the work of AI-powered designers so that they can go beyond suggesting to creating.

What AI-generated symbols tell us about design

Making a few design decisions is one thing. Creating a design from scratch is something else entirely.

To create cohesive designs, these decisions (and the principles they relate to) need to work together, with some taking priority and others taking a back seat. This is where things get interesting.

AI-generated designs will show us exactly what all those decisions look like when working with and competing against each other. 

Take a look at these four hearts:

What do they all have in common?

If you guessed that they were all created by artificial intelligence, you’re correct.

While it might seem simple, creating this kind of variation isn’t easy. As AI’s design sensibilities continue to improve, symbols like these are only going to get better. But that’s not to say they aren’t already doing well. 

In fact, people who use a logo generator choose AI-generated symbols like these ones about 50 percent of the time when they have the option, making them just as popular as human-created symbols. Not too bad for a system still learning the principles of design. 

What can we expect for the future of AI and design?

While we might still be a long way off from artificial intelligence that can create complex, human-like designs, teaching AI the principles of design is how we’re going to get there. As machine learning algorithms incorporate more and more information on what makes for good design, they’ll eventually begin to understand the principles of design.

It’s through those larger principles that AI and designers will eventually become indistinguishable. Because whether those principles were learned in a classroom or through code, the same rules still apply. 

Want to take a dive into all things graphic design? Browse our content hub with over 50 unmatched graphic design resources.