VOICE

PERSONA

TOY

NLU

Overview

Hello Barbie launched Christmas 2015 as the first internet connected fashion doll to offer sustained, interactive conversation. Using a push-to-talk button in her belt buckle, kids could talk to Barbie about school, family, games, fashion, and careers. The doll featured over 10,000 lines of recorded dialogue and was voted one of the top Toys of the Year by TIME.

My role

I worked in a three-person team tasked with designing and implementing all conversational content for Hello Barbie with our in-house conversation authoring tool. I created individual conversational modules, devised the overall logic between modules, and coordinated with Mattel's brand and innovation teams.

Hello Barbie

2015 · PullString · Mattel

1.

DISCOVERY

How do users know what's available?

Discoverability is a persistent problem in voice design. For Barbie I developed a set of thematic return greetings that cued users to the breadth of content available in the doll. She might cite a concern with a friend and then segue to friendship-based content. Or signal she's in a playful mood and segue to more game-like conversational modules. This helped introduced new topics without feeling like a menu.

Robust return greetings

Flexible navigation

2.

NAVIGATION

How do you balance user freedom with guided content?

Kids can and do say anything. The system needed to be flexible enough to handle unexpected inputs gracefully, but also steer users to content the doll could actually respond to. I designed a navigation system that leveraged global commands to redirect users to content randomly, by subject, or name. Unique and memorable naming of particular modules, as well as support for "stop", "help", and "quit" commands gave users a wide range of navigational control.

3.

OPTIMIZATION

How do you maintain variety?

I devised the overarching content delivery logic to ensure the doll didn't surface the same modules repeatedly. Barbie's "neural net" acted as a conversational switchboard, routing users from one module to another. It took into account the prior activity and how it concluded. It managed segues and prioritized unsurfaced content. This kept things fresh for users and provided graceful transitions from one conversation to the next.

Module refresh logic

Add a Title