Predicting Perceived Age: Both Language Ability and Appearance are Important

Abstract

When interacting with robots in a situated spoken dialogue setting, human dialogue partners tend to assign anthropomorphic and social characteristics to those robots. In this paper, we explore the age and educational level that human dialogue partners assign to three different robotic systems, including an un-embodied spoken dialogue system. We found that how a robot speaks is as important to human perceptions as the way the robot looks. Using the data from our experiment, we derived prosodic, emotional, and linguistic features from the participants to train and evaluate a classifier that predicts perceived intelligence, age, and education level.

Predicting Human Interpretations of Affect and Valence in a Social Robot

“…Our set of binary classifiers further the research of [31], in finding that different features carry different weight in communicating emotional valence. Our experiments also confirm claims made in [26] and [22] that speech is an important modality for natural interaction between people and robots, yet robotics researchers often avoid audio as a medium of communication between robots and people [25].…”

“…More relevant to our work are [24] and [26], both of which use facial expressions of the human participants to predict how the humans perceive robot intelligence and age, respectively. As in their work, we use multimodal features, but the features we focus on are derived from the robot and not the participants.…”

“…As robots are employed in increasing numbers across industries, the expectation that these robots will interact naturally with people-lay people who will not understand how they work-is also increasing. However, people often assign anthropomorphic characteristics to robots, for example stereotypical gender [8], social categorizations [9], roles [32], age [26], as well as intelligence, interpretability, and sympathy [24] to the robots with which they interact. As noted in [23], affective expression and interpretation can be a highly effective method for coordinating actions between members of a team and, if correctly mapped to a robot's affordances, should reduce the ambiguity that exists in robot-human interactions [12] and facilitate shared tasks between robots and humans.…”

“…For our work we used the Anki Cozmo robot. Cozmo is marketed as a toy robot for children, is small in size (see Figure 1) and inexpensive, and yet is useful as a research platform (e.g., [26]) because the SDK allows developers to access low-level information about the robot state and it has ample degrees of freedom. It includes animated eyes, a head and lift that can move up and down, track wheels that can be used to turn and move the robot forward or backward, and a speech synthesizer.…”

Learning Word Groundings from Humans Facilitated by Robot Emotional Displays

“…The choice of robot affects how humans will treat it, and it is important for our study that users perceive the robot as a young language learning child. We opted for the Anki Cozmo robot because Plane et al (2018) showed that participants in their study perceived Cozmo as young, but with potential to learn. Cozmo's affordances are likewise consistent with this perceived age and knowledge-level.…”