iHeartRadio: Fostering feelings of closeness through Audio AI
I designed a personalized AI powered audio station experience that helps foster a feeling of closeness amongst listeners.
Client
iHeartRadio
Responsibilities
End-to-End Concept Design
Skills
UX Design
UI Design
Content Design
Wireframing
Information Architecture
Conversation Design
Timeline
16 Weeks
iHeart tasked our team with helping them understand how people would feel about different kinds of AI-generated radio content and if they were to move forward with an AI integration, what type of content formats should be offered.
I utilized multiple research methods to inform the development and iteration of two primary artifacts:
A set of Responsible Audio AI principles designed to help designers and researchers be aware of the considerations that need to be kept in mind while designing audio AI solutions.
A concept to prototype solution detailing the integration of a potential AI solution to the iHeartRadio experience.
Fig - Community: A concept solution detailing of an audio experience that's powered by GenAI to foster and strengthen social connections amongst listeners of the channel. Play the audio to listen to the experience.
Fig - A set of Responsible Audio AI principles designed to help designers and researchers be aware of the considerations that need to be kept in mind while designing audio AI solutions.
Research
My research focused on understanding perceptions and opportunities for integration of potential AI solutions. My goal was to understand:
What are user perceptions of AI Audio solutions?
What types of content formats should be offered?
What potential concerns and opportunities does iHeart need to consider?
To answer these questions, I undertook 4 methodologies:
Desk Research - Understanding the existing work regarding perceptions, concerns and challenges around AI technologies and recent industry news
Online Survey - Understanding the platform usage and gauging initial perceptions of AI technologies
User Interviews - Identifying drivers/rationales for platform usage and underlying reasons for reactions to AI technologies
Co-Design Sessions - Presenting and iterating on AI-integrated audio concepts generated from prior research
Fig - Desk research focused on identifying what the competition was doing and what recent developments were happening within the space.
Fig - Online surveys were conducted to understand user perceptions and how familiar they were current audio AI technologies.
Fig - User interviews were conducted with a diverse set of participants to understand their perceptions and thoughts around audio AI technology.
Fig - Co-design session were conducted with multiple participants to identify and gauge potential solutions and what aspects of the solution participants related with and why.
Fig - We presented two different spaces within the concept testing and usability phase. We focused on two different variations of personalization that users could experience
Research - Learnings
Within the space of artistic merit, spontaneity, and creativity, AI is NOT a proper substitute for the human elements.
AI should be used to enhance/support/supplement the experience.
The experience of using an AI is augmented only when the AI offers a certain form of personalization.
Transparency with regards to AI is quintessential to having a positive experience with the solution.
Users aren't concerned with how well an AI voice sounds but rather how relevant or useful is the content hat it is producing.
Pivot: Initially, our concentration lied in creating a unique AI solution. However, we redirected our attention towards recognising potentials presented through the use of Audio AI.
Community
Your Personalized AI powered radio station
Community aims to provide members with a shared experience, fostering a sense of belonging and strengthening bonds among listeners. It is a personalized audio solution that brings listeners with shared interests and connections to feel connected through a new audio experience.
Experience the Solution via our samples
Fig - Maplewood Middle School station: Designed to help parents be aware of what's happening in their child's school and helping them discover new kid friendly songs.
Fig - BobaBeats: A family radio channel where family members of a group can share updates with one another across the world.
Content Creation Flow
Based on our research, I ensured that the system provided a personalized audio experience by sharing relevant and updated information. This content would be generated by the listeners based on their inputs or pre-defined interests.
The system would then generate a script based on the information it decided to integrate into the script. This script would account for the tone and style preferences that the creator defined. The same considerations will be taken when the system converts the script into an audio format.
Fig - A flowchart explaining how the system creates an episode. The system takes input from various touchpoints, matches the script's tone and voice to the pre-defined parameters and then pushes out the episode.
Salient Features
Multi-input System
To facilitate sharing, two different input methods have been considered:
Paraphrase
A text-based input system where members share content updates that the AI Host can use and integrate into the episode’s script.
Shoutout
A voice-based input system where members can share voice notes that will be integrated into the episode.
Fig - Each channel has a dedicated input button that allows a listener to share updates in the form of voice or message that the system can generate.
Adaptive Voice Change
Depending on the time of use, the AI host for each Community changes its tone based on the time of day.
Day time Community Station Voice
Night time Community Station Voice
Fig - Adaptive voice was added to the system post usability to ensure that the voice felt personal and relatable to the users by contextualizing tone based on the time of the day.
Shared Playlist Algorithm
A key feature of Community lies in its ability to provide new a way to discover music and podcasts. The algorithm is built upon the shared interests of members in a Community channel.
Fig - Shared playlist allows listeners to discover new and common songs from each other's playlists allowing them to feel close to each other.
Personalized Content Updates
A Community station creator can define content recommendations which would be filtered and added to the station's mix based on the level of personalized information (Geography, Location. Age, Gender, and so on) they may have shared with the system.
Interest based recommendations are also integrated. Ex - The station creator or it's listeners can define the teams they support and receive updates about their team and its players.
Fig - An example of how interest based community channels allow people to get updates about their favorite team's performance.
Diverse AI Voice Library
To amplify this experience, we ensure that users get a wide variety of voices they can choose from during the setup process.
Additionally, the user can add their voice to their AI host library while setting up a station.
Fig - An example of how interest based community channels allow people to get updates about their favorite team's performance.
Accessibility & Customization
I also focused on ensuring that this system was design with accessibility and customization in mind. I focused on ensuring that the system was accessible for people hearing loss ranging from partial to complete, people belonging different age, and people who had particular language preferences.
A Settings section allows users to customize the AI voice across parameters like - Pitch, Tone, and Variability.
Next Steps: Designing for Scale
One of the key considerations is the size of a Community. While a Community can be of any size, two different access types have been considered:
Private Community Channels
Public Community Channels
Fig - Private channel content creation and approval flowchart and considerations
Fig - Public channel content creation and approval flowchart and considerations