What if a robot could detect when you think it got too close to you during its approach? This would allow it to correct or compensate for its social ‘mistake’. It would also allow for a responsive approach, where that robot would reactively find suitable approach behavior through and during the interaction. We investigated if it is possible to automatically detect such social feedback cues in the context of a robot approaching a person.

We collected a dataset in which our robot would repeatedly approach people (n=30) to verbally deliver a message. Approach distance and environmental noise were manipulated, and our participants were tracked (position and orientation of upper body and head). We evaluated their perception of the robot’s behavior through questionnaires and found no single or joint effects of the manipulations. This showed that, in this case, personal differences are more important than contextual cues – thus highlighting the importance of responding to behavioral feedback. This dataset is being made publicly available as part of this publication (http://doi.org/10.4121/uuid:b76c3a6f-f7d5-418e-874a-d6140853e1fa).

On this dataset, we then trained a random forest classifier to infer people’s perception of the robot’s approach behavior from features generated from the response behaviors. This resulted in a set of relevant features that perform significantly better than chance for a participant-dependent classifier; which implies that the behaviors of our participants, even with our relatively limited tracking, contain interpretable information about their perception of the robot’s behavior.

Our findings demonstrate, for this specific context, that the observable behavior of people does indeed contain usable information about their subjective perception of a robot’s behavior. As such they, together with the dataset, provide a stepping stone for future research into the automatic detection of such social feedback cues, e.g. with other or more fine-grained observations of people’s behavior (such as facial expressions), with more sophisticated machine learning techniques, and/or in different contexts. ... Read more

Many researchers have started to explore natural interaction scenarios for children. No matter if these children are normally developing or have special needs, evaluating Child- Robot Interaction (CRI) is a challenge. To find methods that work well and provide reliable data is difficult, for example because commonly used methods such as questionnaires do not work well particularly with younger children. Previous research has shown that children need support in expressing how they feel about technology. Given this, researchers often choose time-consuming behavioral measures from observations to evaluate CRI. However, these are not necessarily comparable between studies and robots. This workshop aims to bring together researchers from different disciplines to share their experiences on these aspects. The main topics are methods to evaluate child-robot interaction design, methods to evaluate socially assistive child-robot interaction and multi-modal evaluation of child-robot interaction. Connected questions that we would like to tackle are for example: i) What are reliable metrics in CRI? ii) How can we overcome the pitfalls of survey methods in CRI? iii) How can we integrate qualitative approaches in CRI? iv) What are the best practices for in the wild studies with children? Looking across disciplinary boundaries, we want to discuss advantages and shortcomings of using different evaluation methods in order to compile guidelines for future CRI research. This workshop is the second in a series that started at the International Conference on Social Robotics in 2015. ... Read more