AI-First: Exploring a new era in design

If we consider the growing number of venture-backed artificial intelligence startups, the significant government funding, the many AI-themed talks and forums popping up at business conferences, not to mention all the media coverage and pop-cultural obsession with post-apocalyptic AI stories, it’s fair to say that AI has entered the public consciousness as a “big deal”. While we’ve known this to be the case for a while, it’s exciting to see such widespread enthusiasm and support for our rapidly-expanding field. We believe AI is such a big deal in fact, that we see our role as builders of an AI-First world, where artificial intelligence will impact the creation of any business, product, or service. Within this bold and vast approach of framing the future (which we’ll examine more closely in an upcoming blog post), one aspect of “AI-First” that the design team at Element AI is particularly interested in exploring, is AI-First Design.

Why now? While certain large design studios and AI startups are venturing into designing with AI, it’s still a relatively new field, without clear guiding principles or practices for designers and data scientists to adopt. We’re motivated to figure this out not only by our own curiosity, but also by conversations within the AI community, and with our clients. People seem to be curious, and anxious, to figure this out sooner rather than later. As AI researcher Patrick Hebron states in his ebook Machine Learning for Designers, “machine learning is in the midst of a renaissance that will transform countless industries and provide designers with a wide assortment of new tools for better engaging with and understanding users”. Beyond visual and spatial design, designers now also have to consider how people will interact with AI assistants, connected wearables and home appliances, to mention a few. If we want our technology to be smarter, i.e. less intrusive and more useful, we have to consider how designing with AI can help us get there.

Our goals

With this in mind, today we’re excited to launch our investigation of AI-first design with the goal of defining the practice, and developing a methodology for design practitioners, data scientists, and those working at the intersection of AI and design. Beyond these tactics, we also want to combine our enthusiasm for a better future with a healthy dose of critical thought about the impact of our work, not only on enterprise, customers or users, but on humans.

Our hypothesis is that the impending ubiquity of AI will require a new relationship between designers and machines, and a new approach to design altogether. To us, “AI-first” doesn’t mean “humans-second”, but rather, a consideration of AI from the outset of a project — an open-minded approach to the possibility of gaining insight from machines, as they learn from us. Much like how mobile-first has revolutionized web design with a responsive approach to content creation, we expect that AI-first design will have an equally, if not greater impact on the future of design.

As designers and scientists begin to foster stronger relationships with algorithms and systems, as odd as this may sound today, we will see companies that take on this challenge earlier rather than later outpace their competitors. Yet we also believe that it’s important to think beyond economic objectives and consider the possibility of greater human well-being through smarter design. How can we develop a symbiotic relationship with AI, where not only do we improve systems, but the systems allow us to be better as well? Can AI help us improve our relationship with technology, with more meaningful interactions, built-in limits and timely notifications, rather than simply pushing for addictive algorithms that waste our time and limit our horizons with filter-bubbled news? In considering these broad challenges today, we’re laying the groundwork for an emerging, rapidly-evolving field.

How AI is already impacting design

When we speak of the impact of AI on design, at first glance, it may not be obvious what we are referring to. Therein lies our first challenge. In future blog posts we will look to define terms like “AI” and “design” more specifically, but today, for the purposes of this introductory post, let’s dig beyond the surface definition of design as a practice of visual and spatial creation and examine some widely-used examples of AI-influenced design.

Let’s consider the common example of voice recognition. While early voice applications usually produced frustrating or confusing results, with rapid advances in natural language processing, today’s voice technology is vastly improved. According to Shawn DuBravac of the Consumer Technology Association as mentioned in Forbes, error rates have dropped from 43% in 1995 to 6% this year. With a number of personal assistant bots on the market, voice recognition is starting to live up to the hype depicted in those early Siri ads with Zooey Deschanel dancing around in her pyjamas. As wearables and smart home technology become more widespread, the need for voice recognition that actually works will increase, as screen interfaces become smaller, or disappear altogether. The future of voice recognition will become more refined, with experts predicting the ability to recognize tonal inflections and mood, replicate individual voices, as chatting with sensors embedded in your household appliances becomes commonplace. How’s it going fridge, feeling empty?

Another highly anticipated example of AI design is the self-driving car. While we’re years away from widespread adoption of this technology, it does illustrate an interesting design quandary — namely, what would cars look like if they had always been driverless? Maybe more like your living room? It could be argued that car design is still stuck in a horse and buggy mindset. Design Thinking proponents would point out that this is an example of solution-based design, whereas we should be focusing on problems — however, how do we break out of simply trying to make a faster version of a horse? Can we instead address the problem that humans are typically unsafe drivers, and could be making much better use of all their time spent getting from A to B?

Our assertion is that AI will open up problem spaces that are unimaginable within our current solutions-based model. If, as AI practitioners, we’re only working with businesses to come up with solutions to existing problems, we’re missing the point. We don’t want to only solve incremental problems, but rather address paradigm shifting, disruptive ones. In order to do so, we need a new approach.

A new relationship between humans and machines

As AI-first design becomes more prevalent, we hypothesize that as designers, we’ll need to foster a new kind of relationship with technology, based increasingly on collaboration. If we consider that up until recently, machines have primarily learned from human-generated data, we now need to look at the reverse: what insights humans can gain from algorithms. Josh Clark of Big Medium describes the initial tension that this situation may produce: “Typically digital designers have been creating interfaces for flows and for content for which we have complete control, and now we are gradually and slightly uncomfortably feeding some of the control to algorithms and digital models that we don’t quite understand — and that don’t quite understand us.” As with all massive technological change, if we’re open to diving into this discomfort head on, we’ll likely come out ahead of those running away from it.

One area where we have already seen insights from machine learning is in exposing some of the darker aspects of humanity, revealing the inherent subjectivity and bias in data. While it’s tempting to imagine that machines aren’t biased, Princeton computer scientist Aylin Caliskan points out simply in a recent Vox article: “…machines are trained on human data. And humans are biased.” When Google noticed that it’s type-ahead suggestions were auto-completing the query “are Jews…” with the word “evil”, they moved quickly to correct it. In an equally troubling case, for a while, asking Google Home “Are women evil?” would produce an emphatic “Yes!” followed by a 30-second explanation of why. Yikes.

Rather than dismiss these troubling revelations or shirk away from the ugliness they’re revealing, we need to take responsibility for our data and our designs. Machine learning models, after all, are simply mathematical encodings of our perceptions — the AI will reflect our biases and behaviours, good and bad. So it’s not humans vs. machines, but rather, humans and machines, attempting to learn from each other. When the darker aspects of human bias are revealed by AI, we shouldn’t blame the algorithm, but rather, design with the goal of becoming better humans.

Our approach

So how are we going to achieve this? We’re the first to admit that it’s no easy feat. It’s bigger than us, and bigger than the companies where we work. It’s a paradigm shift in Design, and we need all hands on deck. That’s why we’re going to be working iteratively and in the open, incorporating your feedback along the way. While we have a general sense of what AI-first design will look like, in order to come up with a more precise and credible definition, we believe that a deeper investigation and open consultation is required. With this in mind, over the coming months, we’ll begin by aiming to define the term by breaking it down into its most fundamental principles:

  • What is AI?
  • What is design?
  • What is the impact of AI on design?
  • What is AI-first?
  • What is AI-first design?

We have already begun to survey our internal teams and will continue our explorations at public events, in online forums, and in discussions with AI and design experts on our soon to launch AI-First Design podcast, publishing our findings in our ‘Fundamentals Series’ on this blog as we go.

By tackling this question out in the open, we aim to develop practical tactics that you can adopt as a designer, all the while exploring our industry’s challenging ethical concerns, thereby benefiting not only our clients, but the industry at large. As anthropologist Genevieve Bell stated in 2016 in an article on Engadget: “People from different interdisciplinary backgrounds are hugely important for the next wave of technology. You need to have people who are historians and philosophers and even poets. There has to be this capacity to think differently about data, time, history and logic. It requires as many different ways of thinking as you can possibly tolerate.” While multidisciplinary research teams are by no means easier to manage, we do believe that having a wider range of viewpoints and opinions will generate healthier debates and outcomes.

Inherent in this process is a desire for improved collaboration between designers and data scientists. We want to explore how developers’ and designers’ objectives may differ, and where they are the same. Where are the areas for greatest potential collaboration and how can we learn from each other? In the creation of a new vocabulary, we want to generate greater understanding and increased interaction between AI practitioners.

Getting involved and looking ahead

As mentioned above, we need your help in developing a detailed portrait of the AI-First Design landscape — what is today, and how do we see this growing into a broader movement in the coming years? What are your challenges and concerns as a designer, machine learning scientist, or business? Here are a number of ways you can get involved:

  • Say hi at these upcoming events:
  • StartupFest in Montreal, July 12–15, 2017
  • UX Week in San Francisco, August 29th — September 1st, 2017
  • CanUX in Ottawa, November 2–5, 2017
  • Look out for our upcoming AI-First Design podcast
  • Sign up to our email list for future updates
  • Comment below or get in touch

We want to emphasize that this will be an iterative, experimental, likely messy process. While we have a lot of talent and expertise at Element AI, we can’t define this important new design phase alone. We hope you will join us to create richer, more diverse outcomes that will better reflect the needs and dreams of our industry as a whole, rather than those of an optimistic, yet narrow subset of entrepreneurs and scientists based in Montreal.

Thanks for reading, and look out for our next blog post exploring the first in our Fundamentals Series: “What is artificial intelligence?”

. . .

Written in collaboration with Rebecca West, a Copywriter and Editor with a focus on projects at the intersection of tech, design, and creativity.