Investigating Human Perceptions of Trust and Social Cues in Robots for Safe Human-Robot Interaction in Human-oriented Environments

Abstract

As robots increasingly take part in daily living activities, humans will have to interact with them in domestic and other human-oriented environments. This thesis envisages a future where autonomous robots could be used as home companions to assist and collaborate with their human partners in unstructured environments without the support of any roboticist or expert. To realise such a vision, it is important to identify which factors (e.g. trust, participants’ personalities and background etc.) that influence people to accept robots’ as companions and trust the robots to look after their well-being. I am particularly interested in the possibility of robots using social behaviours and natural communications as a repair mechanism to positively influence humans’ sense of trust and companionship towards the robots. The main reason being that trust can change over time due to different factors (e.g. perceived erroneous robot behaviours). In this thesis, I provide guidelines for a robot to regain human trust by adopting certain human-like behaviours. I can expect that domestic robots will exhibit occasional mechanical, programming or functional errors, as occurs with any other electrical consumer devices. For example, these might include software errors, dropping objects due to gripper malfunctions, picking up the wrong object or showing faulty navigational skills due to unclear camera images or noisy laser scanner data respectively. It is therefore important for a domestic robot to have acceptable interactive behaviour when exhibiting and recovering from an error situation. In this context, several open questions need to be addressed regarding both individuals’ perceptions of the errors and robots, and the effects of these on people’s trust in robots.

As a first step, I investigated how the severity of the consequences and the timing of a robot’s different types of erroneous behaviours during an interaction may have different impact on users’ attitudes towards a domestic robot. I concluded that there is a correlation between the magnitude of an error performed by the robot and the corresponding loss of trust of the human in the robot. In particular, people’s trust was strongly affected by robot errors that had severe consequences.

This led us to investigate whether people’s awareness of robots’ functionalities may affect their trust in a robot. I found that people’s acceptance and trust in the robot may be affected by their knowledge of the robot’s capabilities and its limitations differently according the participants’ age and the robot’s embodiment. In order to deploy robots in the wild, strategies for mitigating and re-gaining people’s trust in robots in case of errors needs to be implemented. In the following three studies, I assessed if a robot with awareness of human social conventions would increase people’s trust in the robot. My findings showed that people almost blindly trusted a social and a non-social robot in scenarios with non-severe error consequences. In contrast, people that interacted with a social robot did not trust its suggestions in a scenario with a higher risk outcome.

Finally, I investigated the effects of robots’ errors on people’s trust of a robot over time. The findings showed that participants’ judgement of a robot is formed during the first stage of their interaction. Therefore, people are more inclined to lose trust in a robot if it makes big errors at the beginning of the interaction.

The findings from the Human-Robot Interaction experiments presented in this thesis will contribute to an advanced understanding of the trust dynamics between humans and robots for a long-lasting and successful collaboration.