2022 Fiscal Year Research-status Report
Voice Against Bias: Exploring Voice Assistants as a Method of Countering Implicit Bias
| Project/Area Number |
21K18005
|
|---|
| Research Institution | Tokyo Institute of Technology |
|---|
Principal Investigator |
シーボーン ケイティー 東京工業大学, 工学院, 准教授 (70831262)
|
|---|
| Project Period (FY) |
2021-04-01 – 2024-03-31
|
| Keywords | HCI / voice assistant / implicit bias / ageism / older adults / voice ux / user experience / usability |
|---|
| Outline of Annual Research Achievements |
The ongoing COVID-19 pandemic created severe delays for this project. Even so, we developed three preliminary older adult text-to-speech (TTS) systems using the three recorded older adult voices using machine learning. We then created a preliminary voice assistant (VA) with a storytelling activity.
All design studies were completed. An online survey was conducted with older adults to gather ideas for a story-based activity with the VA. We used the top ideas for the system design. Three systematic reviews on data collection were conducted; two were published. An online study was conducted to assess the qualities of the three TTS's in comparison to other TTS's. This led to one side project (kawaii vocalics) and confirmed the "agedness" of the TTS's. A face-to-face drawing study was conducted with younger and older adults. This also confirmed the "agedness" of the TTS's. Both studies raised the problem of the "mechanical" nature of the TTS's. A usability study of the storytelling system with the best TTS (old woman) was conducted; however, no results for ageism were found, and while the younger, professional TTS's were preferred, the older adult TTS's were deemed most suitable for storytelling. Finally, a face-to-face study on an alternative activity, mindfulness meditation, was conducted; however, the storytelling activity was preferred.
One master's thesis, two bachelor's theses, and two student exchange projects were completed. Three short papers were published at the top international peer-reviewed conferences in the field of human-computer interaction (ACM CHI and DIS).
|
| Current Status of Research Progress |
Current Status of Research Progress
4: Progress in research has been delayed.
Reason
We are still severely delayed due to COVID-19 in FY21, which continued into FY22. For the most part, we had to rely on virtual participation. However, after COVID-19 restrictions were eased at the university, we were able to recruit older and younger adults for face-to-face studies. Even so, we are delayed. Additionally, the results of the design studies made it clear that the current TTS's are too machinelike. Unfortunately, the original voice actors could not participate again. We therefore recruited new voice actors. We completed recording in early May and are processing the voice recordings for machine learning, which will take about two months. We are currently developing the system for in-home use; however, we may need to scale back on the activity to enable this field study.
|
| Strategy for Future Research Activity |
We will adjust the plan to accommodate the delays due to COVID-19 and the need to improve the TTS's and system. Experiments 1 and 2 will be combined, i.e., we will recruit young and older people at the same time, and is planned to start in the summer.
|
| Causes of Carryover |
COVID-19 has severely delayed this project. Now that we can conduct face-to-face studies, we can run the experiments. We plan to do so this year and next year (requesting an extension to FY24).
|
| Remarks |
We will endeavour to make all of our work open access, if possible. We will use Open Science Framework (OSF) to do so.
|
