Some people have an innate fear of robots for one reason or another, whether it’s the fear of being injured or the thought of AI taking over. Whatever the case might be, researchers from UCLA have developed a robotic system designed to put humans at ease by generating different real-time explanations for its actions, meaning we feel more trustworthy of robots if they explain what they are doing when performing tasks.
To find out whether or not humans would reduce their anxiety around robots, the engineers taught their Baxter robot to unscrew a medicine bottle cap, and explain the process several different ways while performing the task. The first type of explanation was described as symbolic (or mechanistic), which involved having the robot display each process (grasp, push, twist, etc.) while executing it in real-time.
The second explanation, known as haptic (or functional), described the general function of each step as it was carried out, using descriptive words such as approaching, pushing, twisting, and pulling. The third explanation involved simple text messaging of the actions carried out and described what the robot was going to do before undertaking said actions.
The team gathered 150 volunteers to observe the robot as it removed the top from the medicine bottle while doing so using the three explanations. The results showed that the participants had the highest trust ratings when Baxter used haptic and symbolic explanations, while the lowest came via text messaging. The researchers found that humans were more trusting if they were given more information on what the robot was doing, and state their next step is to teach robots why they are performing those tasks.