Martin Danty
Anticipatory Design and Future of Human Experiences
In the era of Netflix suggestions, Spotify and Apple Music playlists and Amazon’s ‘you might also like’ sections, the anticipatory design has grown into our daily lives. We asked Martin Danty, a Senior designer at Danske Bank who previously worked in Microsoft’s Cloud + AI Studios, to share his views on the impact of anticipatory design on users and what to expect from it in the next 10 years. Check your seatbelts, and let’s go.
Martin Danty, Design Consultant (ex-Microsoft)
Martin is a product designer, mentor and co-founder of Juicebox, a platform for music industry, who specializes in anticipatory design and building products that scale. He has worked on Microsoft’s Cloud and AI team, defining the next iteration of products for the supply chain sector. He is a board member of Nordic Blockchain Association and is now back in the fintech industry.
Let's start from the beginning, Martin, what is anticipatory design?
The shows Netflix recommend you to watch, the songs Spotify recommend you to play and the products Amazon recommend you to buy, are all examples of anticipatory design.
Tiktok is another example of how much the right implementation of anticipatory design can affect the user experience. How exactly their algorithm works is not yet public, but it clearly has a lot of perimeters, like how long you watch a clip, whether you interact with it (share, comment, like) or rewatch it — all of this data enables the algorithm to produce a personalized feed for each user.
From Spotify
How does anticipatory design work?
On the technical side, most recommendation engines are based on a “nearest-neighbor” principle. This is a very effective algorithm that can provide accurate recommendations even if you don’t have a ton of labels and metadata on all of your content.
Nearest Neighbour Algorithm
For example, if I listen to a lot of The Weeknd on Spotify, my “nearest neighbors” will be those users who also listen to the same songs of The Weeknd. If a new song by The Weeknd is released, the algorithm can recommend it to me, not because it is the same artist, but because some of the other The Weeknd fans on Spotify will probably search for it and listen to it. This is also how Spotify recommends “similar artists”. They don’t have to be similar at all. If all The Weeknd fans also listen to Metallica, the algorithm will expect these artists to be similar in sound.
Why is anticipatory design so powerful in entertainment and other industries?
It became a cornerstone of web2.0 that we didn’t just consume content, we wanted to interact with it and create our own. Labeling and indexing all of this content manually has simply become impossible.
Anticipatory design is so powerful because it is a way to automate processes in real time, while increasing user engagement. But this is also what led to now a global problem with filter bubbles. As you engage with a specific type of content, the system anticipates that you want more of it, and this feeds a cycle that ends up showing you only content that you like. If you don’t like to engage with contrary opinions, the algorithm will pick up on this, and only show you content from authors whose views are in alignment with yours. This is not really an issue for entertainment such as music and films, but it became a problem for information-based platforms (news, blogs, tweets, streams, etc.) when SoMe began to apply the same algorithms on them.
Illustration from the Medium post by Martin Danty “How do you get Users to Trust your Recommendations?”
Avoiding filter bubbles in news has never been more important. Facebook and Instagram in particular are now being questioned about their influence over elections, extremism, depression and bullying, to name a few. That’s why I would argue that anticipatory design has transitioned from being mainly a developer challenge to a UX challenge. Not just in terms of avoiding filter bubbles, but also considering the wellbeing of the users and telling them how this works behind the scenes through the UI. This comes at the cost of the company’s KPIs such as user engagement, so it is important to strike the right balance so as not to lose the benefits of anticipatory design entirely.
What are the main principles/features of anticipatory design?
Let’s use a clothing analogy. Imagine your product is a T-shirt. Without anticipatory design you need a one-size-fits-all in a color everyone likes. But with anticipatory design you get to see what the rest of their wardrobe looks like, so you can send them a T-shirt that fits them perfectly and matches their personal style.
At first, this feels like the company really understands you. But then, your friend with a completely different style also praises this company, and you begin to wonder. How can they know exactly what I want every single time? Imagine the company now starts purchasing T-shirts for you automatically, and you receive them by mail. You love them, and you’d buy them anyway, but you feel violated because they made this choice for you.
More and more users experience this when they interact with companies that have access to a disturbing amount of user data.
Illustration from the Medium post by Martin Danty “How do you get Users to Trust your Recommendations?”
3 main principles of anticipatory design
- Trust. For users to appreciate anticipatory design, you first need to earn their trust. I have written a Medium post with more details on how to do that, but in short, you should do the opposite of Facebook.
- Control. Users also like to feel in control, and the option to disable the algorithm or adjust it makes them feel like they choose to see these recommendations, rather than being forced to see them.
- Transparency. Explain what data sources you use, and how they affect what users are recommended.
Our studies show that these are the main factors that’ll make users trust your recommendations. Usually, they won’t even impact the conversion rate or user engagement, as most users won’t actually disable or change anything — they just want to have the choice.
How powerful are the algorithms that predict users' next steps?
This is difficult to answer. It depends on how much content you have, and how picky your audience is. Imagine you have a streaming platform with only one popular streamer, which all users have rated positively, and two mediocre streamers. How many users will you need before you can recommend the popular streamer to new users? Probably not that many. In order to distinguish good content from bad content, you don’t need a lot of data.
People are quite predictable within certain settings. Users’ taste in music or film genres usually changes very slowly. If you are into comedies and heavy metal, this won’t suddenly change the next day. But platforms such as Facebook will need a huge amount of data to figure out what content I’m in the mood for, as this can significantly vary based on what I’ve experienced today, who I’ve talked to and what’s going on in my life in general.
How do you know if you are predictable? If you find yourself opening Instagram, Facebook or Youtube with a particular purpose, but get absorbed in the recommended content and forget about your initial purpose, then the answer is yes. The algorithm has figured you out :)
Top-3 tools for work
I could write a blog post about how great it is compared to previous design tools such as Sketch, and especially the move to working entirely cloud-based is fantastic
While at Microsoft, I also used ProtoPie a lot because I had some very complex needs that the built-in prototype tool in Figma couldn’t satisfy
Lastly, I guess the grandfather Adobe deserves a spot too. They have such a vast array of fantastic apps that I use for everything else
What are the current limitations of anticipatory design?
I actually tried to answer this question in my master’s thesis, and I have another blog post that goes into depth on the limitations and provides a few tips on how you can get started with anticipatory design. But in short, I would say the main limitations are how much data you have available, whether you have an imbalance in your users vs content ratio and how sophisticated your algorithm/AI is.
How can anticipatory design evolve in the near future?
There is a convergence of many current trends that point towards an increased need for anticipatory design. The amount of data continues to increase exponentially, along with the content that competes for our attention. As the process of creating products and entertainment gets easier, we will see an increase in niche content. To find the relevant niche content, we will use anticipatory design as guidance.
Lastly, as our attention span also decreases, especially in Gen Z. Companies won’t be able to afford bad recommendations, as only a few irrelevant suggestions can be a “make it or break it” experience for the user.
What changes do you think we will see in the design industry in the next 10 years?
I think that automation and AI tools like Uizard will increasingly impact our workflows, so we can focus on solving problems instead of pushing pixels. The whole process — from wireframes to working code — will get significantly shorter, so we can test our prototypes much faster. On a broader scale, I think that Mixed Realities will play a bigger part of both entertainment and workspaces, so designers will need to explore how they can utilize the affordances of the new 3D worlds to create even more amazing experiences.
Any final tips for people who want to become designers?
If you are not sure what industry you thrive in, apply for a position at an agency. It is a great way to get a feel for how different companies function, both in terms of size, culture and products. And you should work on your portfolio! Remember, it’s always quality over quantity :)
Oh, and don’t be afraid to get your hands dirty with the code. I don’t think designers should code, but I’ve seen first-hand that if you at least try to understand it, it will make a big difference. The developers will respect you for trying.
If you want to tell your design career story or share your expert opinion, send us a pitch at serafima@flowmapp.com
