Providing foreknowledge to improve the efficiency of assembling with the voice assistant.
M12 Personal Design Research Project.
Coach: Yaliang Chuang. Feb. 2019 – Jun. 2019. Seamless Squad, TU/e
Novice assembly line workers nowadays are facing a long and hard learning process. To reduce this learning burden, giving instructions through the voice user interface (VUI) is considered to be a potential direction for further exploration. However, past research shows that interaction with VUI may cause a decline in production efficiency. In this project, in order to improve this efficiency, the countdown light pattern that shows the duration of the next reply was evaluated as a potential solution of providing foreknowledge of the upcoming interruption. The light pattern was embedded in a VUI system named Pre-Echo and was tested in a user test. The user test showed that this countdown pattern had no negative impact but the potential to improve the efficiency of assembly with the VUI. However, as the foreknowledge was only effective when the instruction was relatively long and lacked some important information as participants reported, it was indicated that maybe some functional light patterns can be explored in future work. This project provided an empirical work exploring how light patterns could be used to assist the learning process of assembly workers and thus increase production efficiency.
To improve users’ production efficiency in the VUI assisted assembly tasks, we first observed the reason for the decrease in efficiency. As we learned, users’ assembly tasks were sometimes interrupted by voice information. Therefore, based on the literature review of interruption theories, we believed that it might be possible to avoid the occurrence of interruption behavior in the middle of the tasks by pre-informing it in advance, thereby weakening the interruption’s negative impact on efficiency. As a result, we designed Pre-Echo, which contained an Amazon Echo Dot to provide voice interactions and a WIFI-enabled light strip that connected to the backend of Echo Dot. When applied in use, SlumberBot could collect the time length data of the upcoming voice reply synchronously when the user requested a VUI reply, and showed this time length through a countdown light pattern as an ambient notification before the voice reply began.
The designed architecture of the Pre-Echo system
The prototype of Pre-Echo in use (countdown light pattern as an ambient notification)
User Study and Findings
We used the with-in subject approach to design the experiment, in which every participant attended three sessions to complete the same assembly tasks in the same space. The first session was designed to help participants learn and understand how to assemble parts with the VUI system. Like pre-employment training for teaching the specific skills needed to work in the factory, the researcher gave them instructions and guided them to practice the tasks. In the following two sessions, these trained novice workers got new jobs for them. Each participant was asked to assemble the same part of a robot with one of the two lighting feedback designs, i.e., countdown or breathing light. The settings were randomized and counter-balanced to eliminate the order effect. Meanwhile, we also allocated 3-5 days between the sessions to minimize participants’ learning effects.
The setup of the workstation
1. Cognitive Load Questionnaire, 2. Pre-Echo, 3. Toys to be assembled
4. Image instructions, 5. A consent form, 6. Workflow
A shot of the user study
Some questions could be answered based on the results of this experiment. Firstly, from both the recorded videos and the quotes of interviews, it could be seen that participants’ interruption behavior was influenced by the countdown light pattern. Secondly, as a step further, it is also shown in the interview that the countdown light pattern can be helpful for workers to better manage the workflow, and thus being overall beneficial to the VUI-assisted assembly task. However, can the countdown light finally improve the efficiency of learning with a VUI system? From the experiment data, it is found that the countdown light might not be beneficial in all settings (e.g., decrease the rate of making mistakes), but still useful in some conditions. For example, it can be observed that the countdown light pattern might value more when the voice instructions are relatively long. Other conclusions and detailed data could be found in the paperwork of Pre-Echo.
Participants’ interruption behavior was influenced by the countdown light pattern
Generally speaking, I’m satisfied with my performance, as I can always keep a proactive attitude in learning, reflect on the reasons behind every choice, and complete the research paper in time with fairly good quality. As the most remarkable professional profit, I had the chance to gain a deep understanding of research topics such as attention theory and peripheral interaction, which I didn’t know before. Based on my personal experience, I believed that these theories could be applied to many aspects of my futural design projects. Also, I got trained a lot on how to conduct a proper lab study and how to prepare the questions for the stimulated recall interview and semi-structured interview. I went through the whole process from being clueless at the beginning, then slowly working out the script and finally verifying the functionality of this script through pilot tests. This process benefited me a lot. Moreover, this project raised my interest in doing projects related to the voice user interface and conversation agents, maybe I would pick up similar topics in my final master project. Finally, this project actually reflected my insufficient ability in the data processing. I plan to follow some data processing courses next year and learn the basic software like SPSS.