Disney Streaming Assist

A Tailored Movie Selection Process, Maximizing the Potential of AI

Summary
Disney Streaming Assist is a personalized virtual assistant for Disney+ streamers that helps users learn how to create more efficient directions for AI models to generate more relevant media recommendation outputs.

Viewers will create a personal environment with characters and stories they love to make the movie selection process more engaging and tailor-made.
Duration
13 Week Independent Study
Role
Product Designer, UX Researcher
Tools
Figma, Adobe Creative Suite, Blender
TL;DR
Problem: Disney Streaming’s artificial intelligence algorithms are not human-centered.
Opportunity: How might we help users efficiently communicate with AI models to generate personalized outputs?
Solution: A personalized virtual assistant that familiarizes users with AI interactions for generating better content recommendations through simplifying prompting.

Discover the content that’s important to you

Streaming services can feel overwhelming with the sheer amount of content immediately presented to you without any direction or guidance. I aimed to develop a feature that helps users communicate with AI to generate personalized outputs, assisting in the movie selection.

Personalize Your Virtual Assistant

Users can use familiar characters, themes, and stories to build an environment they feel comfortable interacting with. Characters will retain their personality from films and interact with the themes to help the screen feel alive.

Simplyfiying Prompting

Creating the prompt for the AI models is broken into several segments, with questions adapting to previous answers, external environment, time of day and viewing history patterns. 

Manually Input Preferences

The formerly mentioned survey function is optional as users can directly input and edit their preferences into the AI model to generate a pool of options to watch instantly.

Browse Options with Assistance

The virtual assistant exists beyond the Assist feature. It accompanies users through the entirety of the movie selection process - indicating when something users are hovering over aligns with the preferences they currently have selected.

Background
Disney+ launched as a digital streaming service and brought the magic of Disney into peoples' homes in one central hub, making its media more accessible than ever before. With about 164.2 million users, it quickly became one of the most popular streaming services in the few years it has been around, thanks to the impressive content library.

I began by looking for ways to make Disney+ (and streaming as a whole) more interactive

When I started looking into the problems that I and other streamers were facing, my initial idea focused on mirroring little magic moments from the parks and movies through micro-interactions to breathe life into movie selection.

But then I realized that there was a missing link

Current AI models weren't helping people find relevant content at that moment but instead giving suggestions based on watching history, which may not always be accurate.

Problem

Disney Streaming’s artificial intelligence algorithms are not human-centered.

In Disney+ and other streaming services, Artificial Intelligence (AI) algorithms were introduced to simplify content discovery by presenting suggestions based on previous viewing history. However, the ways that AI algorithms operate are not human-centered since they solely base recommendations on past data rather than the meaningful content preferences of individuals at that moment.

With AI models for Disney+ being a one-way channel to provide users with content recommendations, many are left unsure of other channels for content discovery.
Goal

Improve the movie selection experience by making content discovery more intuitive + personalized.

I sought to design a feature that adds a new perspective to content discovery by allowing consumers to communicate directly with AI models, allowing users to convey their current preferences for valuable results. I created three goals to satisfy experience, technology, and business needs while keeping users at the forefront.
User Research
I interviewed various viewers regarding their streaming subscriptions to gain a more comprehensive understanding of streamers and their challenges. I included casual viewers, hardcore fans, and everyone in between. In these interviews, I took note of recurring patterns in behaviors and challenges, such as users often experiencing decision fatigue and users having difficulty finding relevant content organically in-app.

I then created a journey map of the movie selection process for a casual streamer.
User Interviews
Affinity Map
Journey Map
Secondary Research
Following the interviews, I conducted secondary research on navigational patterns in the context of user flow and how individual components play a role. This research served as the basis for how I structured communication styles with the AI models.

While search/text input was the standard for AI prompting and pushed for in most websites, this was only beneficial for people who knew what they were looking for. This style did not solve the problem that we had at hand - content discovery through AI.

Research suggested that most people began navigation and discovery on an existing page. With smart TVs being a standard device for streaming, text input for long prompts was less than ideal.
Research Results
While conducting secondary research, I wanted to ground the Disney Streaming Assist feature in human emotions to gain a different perspective and consider alternative solutions. The basis of the AI model centered on Paul Elkman’s 6 Basic Emotions with two additional ones in pleasure and neutrality to round out typical feelings from movies.

Research suggested that the emotions that we want to feel can help shape us into who we want to be. The AI model should not show what it thinks is best for us based on viewing history but what users want to see to help them grow.
6 types of basic emotions
Market Research

Why are subscribers driven to certain streaming services, and what causes them to stay?

In the beginning phases of my project, I reached out on social media to see who would be willing to talk about streaming services with me to understand how they chose their services and how they discover new content. There was a wide variety of services that users subscribed to based on a variety of factors.

To better understand this issue, I conducted a competitive analysis of Disney+ and some of its top competitors.
Competitive Analysis of Streaming Services

How do people currently interact with popular AI tools on the market?

I needed to look at more than just streaming services, so I looked at popular AI tools such as ChatGPT, Midjourney, and Grammarly, which are hitting the mainstream market. However, AI has existed for long before this - internet browsing, social media, and streaming services.

I discussed this topic with most of my interviewees, but it only seemed like those who kept up to date in the tech/design/creative sectors had a position, with most others feeling unsure about its rise.
Competitive Analysis of Popular AI Services

How might we help users efficiently communicate with Disney Streaming’s AI models to generate personalized outputs?

AI models are intimidating for most casual viewers. By simplifying and providing structure for AI prompting, we aim to make it easier to navigate. Customization helps users focus on the learning journey by making it feel interactive and personalized for casual viewers.

Success Metrics
The success metrics below were built on my previous user interviews while keeping business, technology, and user concerns at the forefront.
Development

Develop a feature for Disney+ that allows users to directly communicate with AI models to improve content discovery.

After identifying the main concerns with the Disney+ application and streaming services, I focused on developing my insights gathered into wireframes. The ideas focus on segmenting and simplifying how users could prompt AI models to discover content that is relevant to them through an engaged process that is tailor-made for individual users.
In one of my courses, we learned motion design and 3D design skills. I wanted to push myself creatively in this project and incorporated 3D design work - sculpting one scene and one figure from Disney. I chose Funko Pop as the base figure to streamline and standardize character rigging no matter what character it is. The stylized aesthetic of the environment was the in-between 3D animated, 2D animated, and live-action styles.
Blender Workspace
User Testing
Through conducting interviews, I was able to determine areas of frustration users have with content discovery and AI models of Disney+. After designing version 1.1 of my feature, I wanted to garner feedback and work from that point. Through presenting my work to my interviewees and peers, I determined points of possible improvement and patterns that users would prefer when it comes to information presentation and improving content discovery.
Updates
After analyzing the feedback I received, I honed in on specific areas to iteratively improve upon over the following weeks. Through two rounds, I updated the user interface and the application’s content discovery process to reflect better the feedback gathered.
Final Solution

Personalize content discovery for casual streamers and lead users to content that matches their unique preferences.

Customize the content discovery process from the visuals users interact with to the preferences that they give to the AI to generate better outputs.
Final Screens
Reflection

My love of movies and technology 🎬 🤝 🚀

Growing up in my immigrant household, communicating with my grandparents wasn’t always easy. Their English wasn’t excellent, and my Vietnamese/French/Cantonese were even worse. Whether it was a children’s movie like Finding Nemo or a time-period drama, movies were the universal language that brought us together.

Future state

During my research, I found data suggesting users respected what their families, friends, and social media offered. In my next iteration, I would love to look at ways accounts could interact with one another, offering suggestions from the AI model and peers - creating a social element in content discovery.

The importance of iteration

With this being a semester-long project, I had plenty of time to go back and look at my work and make continuous improvements over some time. As I looked at my research, I had new moments of clarity, leading me to a final product that I am proud of.
Some of my favorite Disney+ titles.

✉ If you’re on the Disney Streaming team with feedback on anything, please let me know what I can improve

☝ Back to The Top

Next Project: Exponential Pathways Admin ☞

Francis Phan

francistphan@gmail.com