Teen Courtney
This Project was the second Project of my Designing Digital Play course at the ITU Copenhagen, after the Snack Attack project. For this project we had to, we had to create a functioning prototype for a disobedient technology or electronic. And I worked with the same 5 teammates from Snack Attack, who were: Sophie Rinnerthaler, Lian Feng, Alexander Dacre, Daniel Boel and Frederik Betting.
Concept
“Teen Courtney” is a 16-year-old teenage voice assistant that shows what it would be like to deal with a voice assistant that has not quite come of age yet. Users can speak with her and she will respond accordingly, based on how her personality was developed.
She portrays cliches of a teenage girl, but she has a deeper side to her personality. By asking the right kinds of questions, the user can discover more of these personality quirks. She has several topics that she’s very enthusiastic about like the environment, Avril Lavigne, and horoscopes, so she repeatedly steers the conversation in these directions until the user actively engages her with one of these topics.
This project creates a disobedient voice assistant that challenges the way Social AI generally, and voice assistants specifically, reflect and reinforce power dynamics between genders and age groups. It builds on top of existing research into Social AI, gender, and the internet. Even though the user usually takes on the authoritative and respected role in the relationship with their voice assistants, Courtney challenges this dynamic and questions the user’s relationship with Social AI.
Process
Courtney was built on top of Google technologies. Her personality is created with the help of the online tool Dialogflow for conversation scripting, where responses for different requests and fallback scenario responses are modeled to make it possible for Courtney to hold a conversation. We would write out Courtney’s responses based on keywords that she could pick up. For example, if the machine recognized “Hi, Courtney.” Courtney would recognize “Hi” and respond with something like “What’s Up?” Thus, we basically hard-coded every response she had. Google Assistant was chosen to implement the voice assistant.
During the prototyping process, tests were done with the Google Assistant App on the smartphones of the development team to improve the conversation by making it more natural and fluent. After the first working version of the voice assistant was created, the work on the AIY Projects Voice Kit started to set up the Google Assistant and Courtney’s physical form because the goal of this project was more to have a home voice assistant then to have a mobile voice assistant.
To voice the assistant, an audition script was posted to a voice acting forum to give professional speakers the chance to send an audition file to us. The actor whose voice fit Courtney the most, was then sent the whole script. Later, the decision was made to not integrate a human voice and let the voice assistant have a higher, faster, more robotic voice. The clash of sarcasm and sass with the robotic voice was amusing for test users and reminded them that she was still meant to be a voice assistant rather than just a rebellious companion.
During the process of simultaneous developing and testing, the biggest problem that occurred was the creation of fluent conversation. For us, it was difficult testing the voice assistant because we knew what she was able to say because we developed her, and therefore knew exactly how to converse with her to create fluent conversation. When the voice assistant got tested with users, however, who had not been involved in the development process, those conversations got less fluent and often they did not know how to keep up a conversation with her. Besides the limited skill set of possible answers and questions that Courtney was provided with, Google’s voice recognition software often misunderstood the phrases and sentences users said to her.
User Testing
The user tests with Courtney went very well and she was greatly enjoyable for everyone interacting with her. We introduced her as Courtney and what she could do, and let the test users have free reign on what they wanted to say or request from her. However, after the first few user tests it became obvious that there were certain situations Courtney was still unprepared for.
Initially, there was no introduction for the user at the start of the interaction, which proved to be an issue for the user because they did not know how to interact with Courtney and what to ask her. After realizing this was an issue, a few introductory lines to lead conversation were introduced that would be chosen at random, but still people were sometimes clueless as to how to interact with Courtney properly. The lines in the introduction phrase were hints to Courtney’s passions, but were often not enough for people to realize that they should use these hints as a starting point for a conversation with the assistant.
Final Prototype
For our final prototype, we presented Courtney in the Google AIY box to give her the feel of a home assistant. However, she is also available here on this site. Just start chatting with teen Courtney, but remember, if you want to access any of her personality quirks, talk to her about Avril Lavigne, the environment, and/or horoscopes!