• Voyager
  • Posts
  • Designing in the age of AI

Designing in the age of AI

I recently read a 36-pages long research paper titled “Design in the age of Artificial Intelligence” and it has made so many things clear in my head. Let’s get right into it, on today’s menu we have —

Hello from the beyond! This is Voyager, the only designers-aboard spaceflight navigating you through the AI-UX universe.

We’re like a crispy Indian dosa, spicy on the inside, crunchy on the outside.

I recently read a 36-pages long research paper titled “Design in the age of Artificial Intelligence” and it has made so many things clear in my head. Let’s get right into it, on today’s menu we have —

  • 🧐 The current design scenario

  • 🔮 AIverse updates

  • 😲 The new age, designing with AI (How Netflix uses AIxDesign)

  • 💯 Impact of AI on practice of design

  • 😎 Speak the language

  • 🤣 Meme of the day

Before we talk about the future of Design, let’s quickly look at the past & the present

The current design scenario —

3 components: Practice, Principles, Theory.

1️⃣ The Practice of Design: The tangible stuff: objects, those nitty-gritty specifications, and the 'Voila!' moments. Think of this as a pizza – with its toppings, crust, and that occasional pineapple..(don’t kill me).

2️⃣ The Principles of Design: The golden rules: like being user-centered, so that Uncle Bob can use your app without calling you every 5 minutes. This is the secret sauce recipe – like making sure there's just enough garlic but not too much. There are 3 essential design principles -

  • User centered - the feeling, understanding and prioritising what our users crave for.

    • It’s like a sommelier (a wine expert) carefully sniffing, tasting, and pinpointing the subtle notes and flavours

    • That's our ethnographic research in action.

  • Abductive reasoning - a pinch of some daring spice!

    • While most designers stick to the recipe (deductive reasoning, how things are) or tweak based on past reviews (inductive reasoning, how things likely are), the saucy designers are all about the wild guesswork, “What if users could register by simply recording a short voice command?” Crazy, right?

    • That’s what our brainstorming sessions often feel like. We're creating designs that no one has even dreamt of yet!

  • Iterative - And finally, the taste-test ritual.

    • We don’t just whip up a design and call it a day. We taste, tweak, toss, and try again, until perfection is brewed.

    • We build initial mockups, trying out different layouts, seeing what sizzles, and what fizzles.

3️⃣ The Theory of Design: This is essentially grandma’s cookbook. It's not just about what works, but WHY it works. Those age-old tales and wisdom – or as we call it, the 'theoretical frameworks'.

*static noise*

Sorry to interrupt, but we just got a word from the HQ…

we have info that there have been major updates on the AIxDesign platform!

🔮 The library just got diverse!! New interactions have been added, your research time has been reduced;

screenshot of the new AI-UX interactions added

🔮 Mini AI-UX guides added!! Becoming an AIxDesigner, learning and upskilling just got easier.

screenshot of the mini-guides section on the AIverse website

Keep checking the website, updates coming in faster than I can click the publish button!

*static noise*

As we progress into the future of Design, AI has an influence on all 3 components. Today, we’re going to deep dive into the practice of design through a real-life case study on Netflix, one of pioneers in AIxDesign.

The new age, designing with AI

Netflix has been leveraging AI as early as 2010 to understand user behaviour and deliver a “designed in the moment” streaming experience. What does that mean, designed in the moment? Let me break it down, using 3 general approaches used in machine learning (no tech jargon, just a simple format - explanation, example).

1️⃣ Supervised learning

The first step of this approach is to create a labelled data set; like the classic example of analyzing a picture and having a human being label it as a cat or a dog. You then separate it into training and validation data set; former to train the machine on the pattern which you know is correct, and then validate by having the machine predict if it’s a cat or a dog. You do this multiple times to make the trained model satisfactory. 

Netflix uses a labeled data set made up of actions and results, movies liked and chosen by users through their direct selection/action. Using this, they fuel their recommendation engine to predict more movies suggestions for the user. 

Pretty simple and direct right? Even a toddler would get this - she likes bunnies, she likes the cute pink bunny, then obviously, she likes the cute yellow bunny. More the observation (user data and context of use like time & place of an action, type of device), better are the results.

2️⃣ Unsupervised learning

Unlike the previous approach, in unsupervised leaning, the algorithm (or the super genius machine) finds “natural” groupings, without known outcomes or labels. It finds observations and patterns that may not be obvious to human being; like in the previous example, indoor vs outdoor photos of dog/cat, day vs night etc.

How does Netflix use this? It blew my mind —

The AI pioneer uses this approach to discover related groups of customers in analyzed viewed data, to create new customer segments that even human beings didn’t know were possible!

Possibilities?

  • Targeted marketing campaigns

  • Predict which content to create in the first place

    • In 2013, Netflix decided to produce an American adaptation of the British hit show House of Cards.

    • They dug deep into their data and found that viewers were more likely to binge-watch shows with complex storylines and high production value.

    • Knowing this, they released all 13 episodes of the first season of House of Cards at once, breaking away from traditional weekly release schedules.

    • The series was an immediate hit.

  • In design, create user personas to then create specific user interface.

*mind blown, again* 🤯

3️⃣ Reinforcement learning (RL)

Expecting the last to be the best? Well, you’re right… it’s the most impactful.

In this you don’t start with a data set or try to recognize patterns, you just have a starting point and a performance metric; start somewhere, probe around, check if the performance improved or worsened, then continue or go back to try again.

Makes no sense at all. Let’s take an analogy.

Let’s say you’ve been abandoned on a tall mountain and you have to walk down. It’s really foggy and there are no marked paths. Since you can’t see, you’d walk around slowly and explore different options. But there’s a tradeoff of time. You’re spending time getting the feel of the mountain and once you think you’ve found the best path, you’d spend time actually walking down that path. Exploiting a path is as important as exploring to know if you’re getting closer to the base, otherwise go back and try a different path. The key is maintaining the balance; you don’t want to spend all your time exploring.

That’s how RL works.

Netflix uses RL to personalize the movie recommendations and their visuals. Yes, the movies AND their visuals (cover images). We all have diverse taste but the covers that we see on Netflix are designed specifically for us, taken from the frames of the movie.

  • Just to blow your mind further, a single season of an average TV show, (about 10 episodes) has nearly 9 million total frames.

One of them is chosen so that we end up clicking on that TV show. The algorithm cycles between which movie to show and then which frame to use as the artwork cover, with a right balance between exploration and exploitation.

I’m still processing…wow! 🤯 

Personalized Artworks - source Netflix blog

In the age of AI, the work of designers is not to ideate products, but to conceive a new offering and then design the problem solving loops that will develop and deliver the specific solutions for specific users.

Design in the age of Artificial Intelligence ~ Roberto Verganti et al.

So, how does AI impact the Practice of Design?

  • To design = take a number of decisions.

  • & decision = problem solving skill

  • but decision imagination or creativity (during development)

As a designer, you don’t need to design a specific solution. No detailed development, that can be done by an AI engine. A designer needs to design “problem solving loops”. Loops which combine user collected data with other data, recognize patterns, and then curate user interfaces in the moment, personalized to the user.

Design practice in the new AIxDesign process

In the case of Netflix, it was the small problem solving loops that decided which movies to show, how to display them (the order of the sections “Trending Now” “Critically Acclaimed Movies” etc.) and even which cover images to use for the movies. Through various small loops, you can create a user centered design for every single user. Netflix’s then-chief of communications says “there are 33 million versions of Netflix”

AIxDesign isn’t just about automating the design tasks (text prompt to a wireframe or app screen) but also designing problem solving loops to create specific solutions for a specific user.

😎 Speak the language

Things are changing again. AI is bringing in a new era, and yes, it’s got its own set of tricky words. We, as designers, are not just about choosing the right colours anymore; it’s time to get smart with tech words! We can’t just be the 'can we have it green and on the right?' squad.

We don’t want the tech folks to decide the future, do we?

Some words added to our design-mind-vocabulary-list today were:

  • Data set

  • Labelled vs Unlabelled date set

  • Training data

  • Supervised learning

  • Unsupervised learning

  • Reinforcement learning

  • Exploration vs Exploitation

🤣 Meme of the day

Klean humor

ummmm.. my personality isn’t ChatGPT..

Found this article with a “created by humans” tag haha! haha?

- a designer in the aiverse in zen mode on a Monday evening 😌 

Wait wait, time for quick feedback? What did you think of this piece? Reply to this email to let me know, did you like it this rich? Would you prefer less content? More/Less examples? Hit me up!

P.S. As always, don’t forget to invite your other designer friends onboard the Voyager - we have a few empty window seats on this spaceship 😉

Reply

or to participate.