The Humanity Of Inclusion

AI. Artificial intelligence. Neural networks. Large language models. We find ourselves in exciting and uncertain times, no? Uncertainty is to be embraced. It is, after all, a catalyst for positive change. But is this AI a revolution for the field of design, or a veneer over centuries-old exploitative practices?
Figma, last week, announced a handful of new AI-driven features. One in particular, Make Designs, has inspired existential dread from the design community at large.
The premise is simple. Instead of drawing interface designs ourselves, Figma wants us to let AI do it. At face value, this sounds exciting. After all, who likes manually pushing pixels around a screen? Have we not suffered collective anxiety for decades over the misconception of that being our only value?
For many, the prospect of the industry's premier design tool—used primarily by product designers—beta testing a way to forgo (in the eyes of many business owners) the very need to hire those designers in the first place, broaches the root of their anxieties. It can be difficult not to view such an advancement as another in the long line of technologies resulting in higher performance quotas without corresponding to income security or quality of life for workers.
Certainly, Make Designs has real potential to cure the tedium of drawing mockups. Yet it also threatens to cure designers of their careers entirely. Not because the tool itself is meant to replace designers, but because of the preconcieved notion by employers that it can. An alarming prospect in a world that routinely commoditizes professionals as cost centers, to be reduced in the name of shareholder value.
But this post won't dwell for any length on that inevitable impact to many people's livelihoods. What follows instead are meditations on how this feature might impact product quality from the perspective of underrepresented groups. By the end, you'll have a clear understanding of my rationale for why AI isn't ready to replace designers, as viewed through the lens of accessibility.
Everyone's Responsibility Top
AI's output is, as of the time of this writing, a direct reflection of its training. It knows only what it is shown, and what it is shown is work that has already been created. It cannot empathize. It cannot rationalize. Like an infant, it more or less regurgitates what it ingests. A gross over-simplification, but illustrative nonetheless.
An argument can be made that this is also how human designers create. We are each the byproduct of our own environments and experiences. There is nothing new under the sun, after all. We are, like AI, blissfully unaware of what we have not seen. Everything we create is informed by that which was created before us.
But, we can empathize and rationalize, and we often redefine that which informs our work, through our work. This empathy is how we improve experiences for more people over time; how we break the mold to include underrepresented groups. As designers, we are each responsible for making things that work for everyone, not just the people that things have worked for in the past. In our field, edge cases may not be prioritized, but they should also never be ignored.
That is unfortunately something current state AI is neither concerned with nor capable of solving for. The output from a given prompt relies entirely on the LLM's training. What results is different enough on the surface, but the underlying principles are always the same. While this modeling will produce compelling designs that measure up against existing work, it's unlikely to intentionally reveal solutions for excluded groups of people.
So, realistically, the would-be disruptor of the status quo is instead its systematic perpetuator. Philosophies that routinely leave populations behind, inherently baked into the technology. About 13% of people alive today have some form of disability. According to UserWay, only 3% of the web is currently accessible to them.
What result should we then expect from an LLM trained on the web as it currently exists?
AI improves only through training. Training occures when a model is bombarded with new input. If that input reinforces current design philosophies, for example, then the output will also reinforce those philosophies without intent. And intent, in design, is everything.
In this way, AI is a mimic. Less like a human child, and more like a parrot. Surely some form of intelligence exists, though only peripherally human-like. Any hope of introducing real change requires human intervention; even as the corporate appetite for such interventions has already begun to wane.
Thus, the actual result of tools like Figma's Make Designs is likely to be a staggering loss of design quality—particularly for people who rely on inclusive and accessible design practices. These are, unsurprisingly, mostly members of marginalized communities.
Poorly Performing Prompt Products Top
Folks in positions like product management are primed for the role of "prompt engineer." Much of their day already centers requirements. Prompt engineering is a logical next step, enabling them to produce instantly compelling output by feeding requirements into an LLM. As they do, the process of designing products will become less and less expensive, and corporate shareholders will be rewarded. The resulting race to the bottom will have a fundamental and lasting impact on product experiences.
As designers disappear, so too will their advocacy for inclusive and accessible product designs. Accessibility will become stagnant and out dated, or vanish altogether. It may still get prioritized as a line item within the requirements dictated by prompt engineers. Without truly new concepts to train models on, though, output will rapidly homogenize around lowest common denomenators. Like a virus, the same results will be replicated over and over. Algorithms will weed out edge cases in favor of what works for the majority. But what works for the majority is demonstrably not accessible (see UserWay research quoted above).
Because of this unspoken yet well understood reality, senior designers with experience will keep their jobs in the short term, in so far as one can today. Their bargaining power may even temporarily increase as their skillset becomes more valuable. But their numbers will dwindle as they vest, retire, and move on.
Junior designers and aspiring fresh graduates stand to pay the heaviest price. Why invest in training a new designer when your Product Manager can write a prompt in a fraction of the time and achieve a comparable short-term result? Or when you, the business owner, can do the same; and without dealing with all of the unpleasantries of other people challenging your vision?
If junior designers never grow, never learn to think outside of boxes, and never question the status quo, the result will be a talent wasteland overrun by designers whose only skill is prompt engineering.
The web has matured, but it is still the most rapidly evolving design medium today. We would all suffer the consequences of a world full of prompt products, where accessibility is not just an afterthought, but no longer feasible at all.
The Light At The End Top
I'll end on a positive note. There's plenty that LLMs can help improve within and automate out of the design process. For example, prompts like these could make for incredibly useful Figma automations:
- Using Figma best practice, optimize this component to include the least number of variants and layers possible, without altering its functionality or appearance.
- Translate these bulleted user shadowing notes into a user journey diagram. Include visualizations for how the user felt during each stage of the journey, as outlined in the notes.
- Remove all of the unnecessary groups from this frame, without altering its layout.
- Name all of the layers of this frame consistently with other layers seen throughout other files in this project.
- Replace all hex colors in this frame with the closest matching Figma variable colors from our published design tokens library.
- Replace non-component content in this frame with the closest matching components available in our published Figma component library. Ignore content for which there is no suitable component available.
- Create a simple table that lists all of the properties of this component, with a brief description for what each property does and how it works.
Notice that none of these prompts take the human out of the loop. They're not asking AI to design actual components, or full product compositions. Instead, they're automating out the necessary but tedius processes that are otherwise not uniquely valuable for humans to do.
There are a lot more opportunities following that same basic logic. I believe this to be the right focus for the current technology, one much closer to realizing the true promise of technology upon which society seems so ready to jump the gun.
—
The idilic techno-future that we were promised is one where we get to spend more time with our families, friends, hobbies, and passions. More time spent in exploration of life; watching our children grow up; supporting our aging parents.
The day job—the concept of a singular career purpose—should ultimately be rendered obsolete by technology, freeing us from the bondage of wage labor. Maybe AI can eventually become a powerful tool for that liberation, but the technology of today is far from capable of replacing designers, and our society is far from prepared for that replacement. Without care, we risk undoing through technology the very thing it was created for; giving people a better and more fulfilling life.
At present rate, many people will suffer the consequences of ill considered AI tools that strive for short term creative automation at the expense of long term product quality. Designers will lose career opportunities. Consumers will lose trust. People who have been excluded will stay that way. But the challenge for us to overcome is not AI itself. Technology is a tool. The true challenge is a uniquely human one, rooted in the perverse incentives of a profit driven economy. As designers, we should embrace tools that empower us to solve problems faster and for more people, while tactfully advocating to retain those (still very necessary) inclusive human practices.