Bodystorming helped me to explore how my concept could prompt carers at the optimal time to capture a moment, to avoid prompts feeling like disturbances. I found that that I could infer many receptive moments with simple phone data - detecting that picking up the phone after a period of not using it often meant the carer was having a small micro-break where they could be receptive to a prompt.

Participants acted out care scenarios and I simulated prompts by playing a sound and reading out a notification text, to which the participant acting as the carer could respond to with "Yes", "No", or "Later". In the evening, I sent the captured photos back to the participants to simulate a review of the day.

Overall, the concept was received very positively, and I found that gentle prompt sounds and phrasings were consistently received the best by all participants.

I created key user flows, an initial sitemap, and then sketched wireframes on paper. After a carer participant cancelled an in-person user testing session, I took that as an opportunity to set up a unmoderated online wireframe test with Maze, enabling both the original participant and an additional 3 peers to complete the test.

I could validate the intuitiveness of my basic user flows, compared layout options, and developed a new review option (Photo Collection) in response to user feedback.

To explore how AI could be leveraged for Memento's review features, I conducted research around "AI-powered reminiscing" by reviewing a case study from Google Photos.

I also learned about and incorporated key rules and patterns for designing with AI into Memento - for example "Letting users supervise automation" by making AI features optional and by giving users the option to hide mementos from view based on faces, pets, or dates.

To explore if current AI tools could create engaging stories with just Mementos as input, I first analysed an existing app, Luna, that generates illustrated children's bedtime stories based on a few words of user input. This gave me confidence that AI-based story generation was feasible in general.

Then, I got to work with ChatGPT to create my own adventure stories with just some sample Mementos I had created from my own life as input. The results were surprisingly engaging stories and illustrations, allowing me to fully validate the feasibility of my adventure story review idea.

I conducted one more experience prototyping session with a carer, where I simulated app prompts by sending them text messages throughout a real day in their life. Before that, they told me the approximate times when they would be with their loved one, like they would also do when setting up the Memento app.

I sent the participant three prompts throughout the day and they responded positively to each of them, finally validating my core assumption.

I conducted the user test with a participant who had no prior knowledge of the concept. This allowed me to get very positive experiential feedback and I could validate that the design was easy to use, intuitive, and conveyed its intentions effectively to a novice user.

I realised that I needed to make some refinements of the design to increase contrasts and readability for better accessibility. This is very relevant for Memento, as its target users can be older and struggle with over the top designs - the proto-persona David being a good example of that.

My use of gradients in the app was such an over-the top design, so I changed it to rely more on simple drop shadows as well as increasing the opacity of elements like the nav bar. I also changed the text colour in orange buttons from white to dark purple, improving button readability.