Teaching Methods

We selected three distinct teaching methods, namely Teaching with Evaluative Feedback, Teaching with Preferences, and a Download condition, to explore their impact on user perceptions, aiming to cover diverse approaches to human-robot interaction and different level of involvement. 

Teaching Tasks

We selected three diverse training tasks: Navigation, Control, and Classification, to assess their influence on user perceptions, providing a spectrum of interactions with the robot with different level of difficulties. 

Experimental Procedure

Participants were assigned to one specific training task taught by a specificly assigned teaching method. First, they watched a teaching video corresponding to their assigned method, followed by a series of tests of the robot in the assigned task. 

Below are some videos of the training. All videos used durring the study can be accessed through the following links:

Testing (all tasks)

Results 

Impact of the Teaching Method

Teaching with Feedback method significantly enhances the robot's perceived responsiveness. This suggests that increased interaction and human guidance correlate with higher perceived responsiveness. Surprisingly, Teaching with Feedback was also deemed more user-friendly than the Download method, despite requiring more effort. This contrast may be explained by the method's alignment with natural human teaching approaches, making it more intuitively accessible. 

Influence of the Personality

Extraverted individuals displayed a notably positive view of the Teaching with Feedback condition, correlating with higher perceived responsiveness, intelligence, and likability of the robot. We propose that this result stems from the dynamic and engaging nature of Teaching with Evaluative Feedback, aligning with extraverts' sociable tendencies. Furthermore, intellect traits were linked to improved robot likability and  ease of use in the Teaching with Evaluative Feedback , and responsiveness in the Teaching with Preferences condition. This suggests that those with higher intellect traits are more inclined to embrace innovative teaching methods involving robots, departing from the conventional Download approach. Lastly, participants with prior robot experience exhibited a subtle yet noteworthy correlation with the percieved anthropomorphism, attributing higher human-like qualities to the robot in the Download condition. This familiarity also led them to perceive the method as more useful in the study's context."

Impact of the Nature of the Task

Participants in the classification task consistently rated the robot higher in perceived anthropomorphism, control, and responsiveness compared to those in other tasks. This is likely due to the classification task's higher engagement and incorporation of social elements, distinguishing it from other tasks. Tasks demanding more engagement and complexity appear to boost the sense of control and perceived responsiveness over the robot. Additionally, tasks resembling intricate, human-like scenarios tend to heighten perceived anthropomorphism

Additional Material

Measure of internal consistency: For each metric in our questionnaire, we report Cronbach’s 𝛼, a measure of internal consistency. All measures meet or exceed the acceptable range ( 𝛼 > 0.7), except for Neuroticism and Responsiveness. However, Responsiveness achieves a passable score of 0.6, which is generally accepted in most studies