< Home 

Exploring AI empathy as a new way for inner reflection and good mental health

For many people, there is a great deal of purpose and self-worth tied to our work. We often face new challenges, which can turn into learning opportunities that foster growth and change. As we commit ourselves to meaningful endeavours in the workplace, how might we make sure that we're actively tracking our thoughts and moods to help promote and maintain good mental health? This was a brief presented to me by one of the world’s leading digital agencies on behalf of one of their client - one of the most popular mindfulness and meditation apps.

1. Process

I started my design process by analysing the Brief document to understand the current thinking about the problem, as well as the in-built assumptions and design constraints. Then, pinning the Brief to a brand-new workspace in Figma, I proceeded to the first step of the design process - Contextual Research.

1.1 Contextual Research

As a regular meditator and a very occasional journal writer, I knew something about the benefits of mindfulness practices (such as journaling). Yet, checking my assumptions at the door, I had to get a deeper (and more varied) understanding of the context of this design problem. So I tried to open my mind as much as possible, researching the web for any interesting bit of information about this topic (divergent thinking phase).

Reading articles from a range of respected popular press sources (Forbes, Wired, TechCrunch), academic publications (Ivey Business Review), market research reports (Market Watch, Compare Camp, Statista), it became quickly evident that the lines between all the different mindfulness practices were so blurred, that one had to approach this problem holistically, thinking about the end outcome, rather than any specific tool, method or practice. ‘Whether you’re keeping a journal or writing as a meditation, it’s the same thing. What’s important is you’re having a relationship with your mind’ - writes Natalie Goldberg in Positive Psychology.

Missing a bit of the first-person perspective, I then watched a series of TED talks and YouTube videos about people’s experiences with journaling and mindfulness at work, noting their motivations, hopes, challenges and emotions. To understand our potential target user demographic, I then looked at demographic data for proxy mindfulness apps.

Noting down every new interesting bit of information on a separate digital equivalent of a post-it note in Figma, I soon had my screen full of new interesting facts, records, quotes and insights, waiting to be grouped and organised into an affinity map (Figma link).

 Desk research information, grouped and organised by insights and topics

 An example of one such insights cluster

This exercise gave me numerous new insights about the topic, most notably:

Mental health is a serious issue, as well as a serious business.
  • Mental health costs are predicted to hit $6 trillion by 2030, greater than the cost of diabetes, respiratory disorders, cardiovascular disease or even cancer.
  • While millions of people (200-500m) around the world practice meditation regularly, a vast majority of people are yet to discover its benefits. This represents a huge untapped market.
  • Growing at 8.5% CAGR, the mindfulness meditation industry is poised to reach $2 billion by 2022, with leading mindfulness apps valued over $350 million.

Due to strong competition new mindfulness apps would have to find ways to differentiate.
  • With over 2500 mindfulness apps out there, just two - Headspace and Calm - own over 70% of the market. This suggests a strong need for differentiation for any new player.
  • The future of the business seems to be in data, and providing predictive, personalised recommendations to users (e.g. flight anxiety meditation as the user is heading to the airport).

Leading businesses are embracing mindfulness.
  • Studies have shown how mindfulness increased productivity (in some cases by 120%), decreased absenteeism (up to 85%) and increased profits.
  • There is a strong interest in mindfulness among leading businesses. 600+ companies, including Google, LinkedIn, Starbucks, Adobe, General Electric and Turner, partner with Headspace.
  • Meditation rooms are the hottest work perk offered by the likes of Apple, Salesforce, and Nike.
  • Airlines are now partnering with mindfulness apps as a part of in-flight entertainment.

While journaling can be powerful, many find it to be tedious, suggesting a design opportunity.
  • Clinical studies have shown significant positive benefits of journaling.
  • Many people reported journaling only when dealing with difficult decisions or tough life situations, and not when things are going well.
  • Many people find journaling consistently tedious, difficult and time-consuming.

All the findings about user data converged its way into two proto Persona artefacts. As demographics of potential target users were split evenly by age and gender (as suggested by the demographics of a proxy app, Headspace, as well as first-person reports by people on TED and YouTube), I decided to go with two proto personas on the extreme ends of demographics and JTBDs. My assumption is that if our offering would suit them, it would probably also suit all the other users ‘in the middle’ (Figma link). Believing in the power of Jobs To Be Done (JTBD), I tried to ensure that the end ‘job’ would also be included in the artefacts.
 Proto Persona 1 - Frank

 Proto Persona 2 - Astrid (on purpose, I gave her rather special reflecting needs)

To make sure that my thinking would not yet be anchored for the initial ideation, I made a decision not to include review of existing solutions at this stage yet. Moreover, with more time, I would have also preferred to interview people to gain a deeper level of empathy.

1.2 Initial Ideation

Now, to capture my early ideas, I wanted to ideate on the possible solution concepts (after the initial research, but before reviewing existing solutions), using the Crazy 8s technique. To help guide the process, I initially reviewed (and somewhat shortened) our How Might We problem statement to ‘How might we make the experience of inner reflecting smooth and compelling’?  
 Adjusted ‘How Might We’ Statement

Ideation Round 1

With all my initial ideas expressed without being anchored to existing solutions, I then dove back into the divergent thinking phase, looking for what is already out there, what trends and innovations are shaping this (and the broader) market, and diving back into research.

1.3 Solution and Trends Research.

From the first research phase, I knew that there are numerous way to reflect - dream journaling, diet journaling, thoughts journaling and many others. So now I wanted to take a look at what these could look like - initially using what I hypothesise to be the most popular journaling tools out there - a talk with a friend, pen and paper, or a simple note apps (such as Apple Notes or Evernote).

Probably the oldest and the most common form of reflection - a talk with a friend

 The diversity of styles and decorations in the paper format indicated that this is a highly personal experience.

Next, I looked at the leading existing mindfulness and reflection apps (there was some overlap between them, making me feel glad that I kept an open mind about the problem space until then).

 Today’s leading mindfulness apps

Today’s leading journaling apps

Thinking about the technology trends shaping our culture and innovation, I could not stop thinking about the movie ‘Her’ where the main character falls in love with his operating system because of ‘her’ ability to understand him and to empathise. To my mind also came some of the recent articles I read about screen fatigue (Wired) and how voice is the new medium of social (Wired, Andreesen Horowitz). As an avid user of the Clubhouse app and Podcast listener, I also registered feelings of excitement that I had about this new medium.

Taking into account some sci-fi and technological innovations…

1.4 Solution Ideation

After this additional research, inspired by the trends in the market and the existing solutions, I have used the Crazy 8s method again (a somewhat lazy variation, as I could only come up with 2.5 extra ideas) to add a few more ideas to my original list.
Now that the writer has written the lines, it was the editor’s job to look at them critically. Typically, I would do this exercise together with other experts and stakeholders using a Dot Voting technique, but now it was just me alone with my own thoughts and feelings (I was definitely missing having a creative team around at this point). Looking at the list, I started to notice how some ideas could be parts of the same puzzle, while some others had clear data privacy issues (and privacy was an important factor as I learned doing the desk research). Finally, one cluster of ideas was similar to the format used by most of the existing journaling apps, which went like this (and felt somewhat boring):

  1. Record your feelings on a Likert scale
  2. Add some notes by typing text
  3. See a graph with your mood over time

Keeping in mind one of the learnings of the research phase (that there is strong competition, so new players would have to differentiate), I picked a cluster of ideas that intrigued me the most…

Contemplating on a few shortlisted ideas...

1.5 Concept Sketching

Running out of time, I used the four post-it notes in front of me as both - the solution sketch and the storyboard of the experience. For the user, the experience would feel like this:

  1. Anonymously, users simply speak their mind into a microphone on their mobile device. No tapping on any Likert scales or entering text by hand.
  2. As the user speaks, the AI inside the app seems to understand the user, offering helpful insights connecting the content of the message with their mood and feelings.
  3. Over a period of time, using metadata across many users, the app provides mindblowingly deep insights.
  4. Users have an engaging way to browse voice messages of other people, with a filter functionality that helps users to hear voices of people most like them, or strugging with the same issues. The experience would feel like walking through a crowd, where the voices would play automatically, and you could hear the person next to you the best, but at no point you had to stay around them.

Behind this experience, and the core of the app would be an immensely powerful AI/ML/Big Data ‘brain’ that would transcribe the user’s voice input into words, moods and emotions (based on content, sentiment, tonality and intensity) finding correlations between the inputs. It would then use Machine Learning to improve the accuracy of its analysis over time, taking in inputs from a wide range of users. Eventually, it would be able to offer immediate data-powered insights, such as ‘Ah yes, working from home is a source of anxiety for many people these days.’ Taking this approach even further, it would be able to predict your moods and help you design your day in a way that would make you feel happier (for example, by talking to your Philips Hue lights and Spotify to create a perfect morning for you on a day when you would otherwise be a little cranky).

1.6 Wireframe Prototyping

After the initial sketches, I would typically create a semi-interactive wireframe prototype of the solution. First of all, going through this process helps me to think deeper about some of the features, but also - once you have the prototype in hands - you can show it around and get some early learnings about its value, usability, viability and feasibility.

Early wireframe prototypes

Early wireframe prototypes

Eventually, after wiring the screens together in Figma, two more design artefacts came to life - the initial User Flow Diagram (helps to keep a birds-eye view of the concept), as well as a clickable prototype.

Early wireframe prototypes (could be used for validation with engineers)

A disclaimer: since I was working alone and running out of time, the quality of the prototype suffered. It’s a rough working version, and if I had more time - I would polish it a bit more.

Also, if this were a real project, next, to determine how much our solution is valuable, usable, feasible and viable, I would conduct a number of user tests (concept tests + light usability tests) with potential users, stakeholders and the team. Particularly, I would focus on testing the following (riskiest) assumptions:

  • Users would intuitively immediately understand what this app is for.
  • Users would have no problem sharing their recordings anonymously with the community
  • Users would love the idea of exploring voices of other people who worry about similar issues.
  • Users would love to input information with voice.
  • Users would love to get AI-generated insights to their problems, and would find them helpful.
  • The solution is technologically feasible and we have the required talent for it.
  • Client is onboard with this approach (client buy-in).

1.7 Brand Personality

This is now the beginning of the second part of the test assignment where we’re going to take the idea from a wireframe prototype to a polished-looking UI. Since we currently have no brand assets to work with, we would have to start the process by designing them ourselves.

I usually start the branding process with identifying the brand personality - how would we describe our brand if it were a real person? To do that, I would usually set up a workshop with key stakeholders and team members, asking each participant to fill out a brand personality questionnaire. Containing some tough questions like ‘why does our brand exist?’ it helps to dig deep into the DNA of the future brand. We would then compare notes, and do some Dot Voting to measure how the team feels about certain aspects of the brand personality. For this project, I had to do the exercise by myself.

In this case, I felt that our desired brand personality should be a combination of the following:

1.8 Brand Identity

Once the brand personality was in place, keeping an open mind, I looked at numerous inspiring brands (such as Stripe, Spotify, Tesla, SpaceX), Pinterest boards, the creative process behind the design of interfaces for’Her’ (the movie), Mark Rothko paintings and James Turrell light installations. Since the envisioned product would be essentially a take on humanisation of AI, I also looked at various previous attempts to do that with products such as Bot Care, Roobo and Pudding  as well as the artworks from 'AI More than Human' exhibition at the Barbican and Wall-E (the animated film). Eeeva! This topic alone would require deep research, which I would feel excited to dive into, but not for now.

Brand Identity inspiration and explorations

Keeping an open mind

Exploring a suitable color palette, I first looked at what colors were used by the existing brands. Apart from the happier modern Headspace and the tender, tactile Sayana, I noticed lots of purple-violent gradients. Thinking how we all have different emotional states (sometimes all existing all at the same time), I felt like our color choice has to reflect that variety.

Like beatifully captured in ‘Inside out’ there is a variety of mood beings within each of us

Playing with diverse color palettes

A neon-pastel combination of primary colors

Knowing that sometimes our brand would need to be bold, and at other times - gentle and sweet, I went with a palette that reflected both - neon and pastel hues of the primary colors  in a way that felt fresh, contemporary and different from the competition.

For typeface discovery, I used a program called Fontstand that allows to rent premium fonts for free for a limited time. Since the proxy demography of apps
indicated an even distribution of ages, I wanted to go with something that felt fresh, but not too experimental to raise too many eye brows.

Eventually, I settled with a wonderful free variable rounded sans-serif called Manrope (also in part because its focus on geometric digits and legibility). Moreover, I’ve had a lot of fun with variable fonts before (example) giving us more options down the road.

Manrope typefaces

I felt like the brand’s identity was shaping up to feel modern, yet not too opinionated - allowing us to change it later once we understood our own brand better (something I like to do for young brands).

1.9 Screen Design

It was now time to turn those wireframes into screen designs. The first screen of the app - the Discover screen - would essentially set the tone for the rest of the app. Here the user would be scrolling the interface into any direction, and the face bubbles in the center of the screen would grow larger. The voices of those people would automatically start playing at 90% of the volume, while the voices of smaller character around would share the remaining 10%. The experience would feel as if the user is walking through a crowd.
The illustrations for that screen would be crucial. Not having the time to design anything custom (or to work with someone who would be able to do that), I started to look at Creative Market, Behance and other design resources for available toolkits.

Some of the available public toolkits for face illustrations

Although there were several new ideas that came out from this exploration (e.g. showing a person’s head from behind), eventually I settled on the stunning 3D head illustrations by Amrit Pal Singh, called Toy Faces.

Toy Faces by Amrit Pal Singh

Downloading the pack, and replacing the backgrounds with our own color palette has been one of the most wonderful visual moments of the project. Not only the aesthetic felt contemporary and aesthetically exciting, it was also incredibly modular - anybody could find a part of themselves in this crowd of faces and colors - even Astrid and Frank, regardless of their emotional state.

With such a visually rich and enticing Discovery screen, I wanted the Listening screen to be just as fun. The current wireframes suggested a basic messaging interface with voice recording, which no longer sat right with me because of:

  • Copy. I didn’t like the word ‘recording’ because it felt impersonal and invited data questions
  • Messaging interface. I don’t remember one in Wall-E.
  • Technical constraint. If we wanted our AI to correctly capture the sentiment, the message would probably need to be of a certain minimum length.

Trying to reach consistency in the UI

Looking at the aesthetic of conversational interfaces such as Siri, I found use of a lot of generative art and gradients. Although they looked aesthetically pleasing, there was a dissonance between the tactile feel of the Discovery screen, and the smooth, elegant look of the gradient. I knew I had to look for more a more tactile version of Siri. So I got some plasticine, and colorful post-it notes, a pair of scissors…

Experiments with “Tactile Siri”

Although this direction held some promise, it would have required more time to develop further. So instead I went back to the obvious that I missed - our avatars.

  Sometimes it’s good to miss the obvious and play with some plasticine instead

This approach also had some tough questions - would users understand the difference between an avatar for themselves and their ‘councelor’ avatar? To make the distinction more clear, I even tried adding a bare-bones robot avatar, but it didn’t feel right somehow.

A robot councellor?

A new idea also came - as the user speaks, the avatar’s face can try to mirror the user’s mental state, not empathically and sometimes offer helpful prompts and tips.

Mirroring is an important part of active listening

For insights, I had quite a complex design problem: to show users correlations between their emotions and keywords over time. Data visualisation can be very complex and a proper exploration would have required much more time than I had.

Designing for a complex data query can be exciting and very time-consuming

So instead I went with a more simple approach of showing one correlation at a time in the form of ‘Insight cards’ that the user could scroll through in the Insights section.

Due to time constraints, I kept the exploration for the Insights screen short

When it was time to put all the screens together, I was surprised by how small the prototype looked. Yet, that also made me quite happy. It felt like a sign of a humane solution where the immense complexity required to make it work was hidden in the code, behind the interface.

So small, so mighty!

Prototype video is here.
You can play with the prototype here.
And you can access the Figma file for this project here.

Code Experiments (Desktop only )

Poly CatThree.js / Planet WilsonThree.js / Shape Vortex / Hyperhi Festival / BreathsKodeLife / ZestyReact Carolyne Wade / Book WavePixi.js / Brutalist / Canvas MagazineBarba.js / Words of Women / Brux / Jiffy / Super Shapesanime.js

McKinsey & Company

Developing a lifestyle adjustment program for new diabetes type II patients

McKinsey & Company

Creating a vision for citizen experience of the future for a government in the Middle East

McKinsey & Company

Reimagining core customer journeys for a large Russian bank

McKinsey & Company

Designing a content-driven eCommerce experience to help women find their perfectly-fitting bra online


Designing world’s first wave of Augmented Reality (AR) experiences

L’Oreàl Brandstorm

Developing a new packaging strategy for Redken for Men

Personal project

UXKit - world’s most comprehensive kit of UX methods

Personal project

Inspiring millennials to reconnect with nature

Personal project

Building GoPro for music

Photo stories:

 Nazare
 The Race of Gentlemen
 Environment
 U-Hauling Ass
 California


 Instagram
 Soundcloud
 LinkedIn
 Resume (CV)


 Send me a message