Facilitating Human-Robot Interactions through Humor
Humor is an important feature in human communication that can be leveraged to create more naturalistic and lifelike interactions with robots and humor potentialities can be increased through the delivery of user-personalized humor in naturalistic settings. However, the lack of context, the disregard of user’s preferences, and the over-reliance on certain formats of jokes, such as word puns or one-liners, are still limitations found in a large number of current approaches to humor in human-robot interactions (HRI). The AGENTS - Automatic generation of humor for social robots project created databases of jokes in Portuguese and English in order to investigate the role of humor in HRI taking into account some user characteristics, including their attitudes towards robots and their humor style. The main study involved the creation of an entertainment task and the adaptation of a humor game for the HRI context, during which users' perceptions of the task and robots, as well as physiological measures were assessed, and interaction behavior was analyzed. Preliminary results suggest that a robot that uses humor is perceived as being warmer and more competent than a robot that does not use humor, increasing its emotional value. These results and the developed datasets will be relevant for future research in HRI and are expected to contribute to the development of more socially effective robots, deployed in complex naturalistic interaction settings.
In 1999, Apple introduced a joke generation system that was capable to interact with the user in their OS 9. Although limited to scripted jokes, this was a pioneer system that gave a 'human touch' to the machine and is still used in Apple's operative systems. Nonetheless, humor is a natural emergent feature in everyday conversations and its complexity is hard to capture through scripted, off-context interactions, such as those. Hence, systems that use scripted jokes limit human-robot interaction (HRI) because of their lack of context, the disregard of user’s preferences, and the over-reliance on certain formats of jokes, such as word puns or one-liners.
Humor seems to have positive impacts in people's health. For example, a literature review on the impacts of laughter on blood pressure and heart rate variability found that laughter frequency is associated with improved cardiovascular health, evidence provided by longitudinal studies. However, this review showed that several studies presented sub-optimal levels of quality, so the authors alert for the need of further research to assess the impact of laughter-inducing interventions in cardiovascular health. Either way, humor is a relevant aspect of human communication that may extend to HRI.
In this context, a systematic review of of studies on humor in HRI conducted by Raquel Oliveira, PhD student at CIS-Iscte, and supervisors Patrícia Arriaga, Ana Paiva and Minja Axelsson, found that humor seems to improve the user's perception of the robot, as well as their evaluation of the interaction. However, they point out that few studies have investigated interaction contexts in which humor occurs in a natural way and few have analyzed the effects of HRI on task and robot perception, crossing different types of indicators of emotional, cognitive, and behavioral responses during interaction. Therefore, there is a need to further investigate this emerging theme.
The AGENTS - Automatic generation of humor for social robots project was funded by CMU Portugal program with the following consortium:
Iscte - University Institute of Lisbon, integrating Professor and Researcher Patrícia Arriaga, and researchers Raquel Oliveira and João Barreiros from the Center for Psychological Research and Social Intervention (CIS-Iscte), with the support of the students from the Science on Emotions masters' program, Patrícia Oliveira and Jasmine Sarrouy;
Carnegie Mellon University through the integration of Professor and Researcher Louis-Philippe Morency.
The end-goal of AGENTS was the implementation of humoristic interactions in the context of a group card game involving humans and robots, which was expected to improve interaction outcomes, and increase the positive perception of the robots and the intention to interact with these social agents in the future.
Approach and Results
During its year of activity, AGENTS' team used a conceptualization of humor that was based on its social function (humor used to enhance oneself vs. used to enhance others) and the valence of the humoristic content (positive, negative). Using this conceptualization, AGENTS created a dataset of jokes (both in Portuguese and English) and the application of supervised machine learning techniques.
More importantly, perhaps, is the study conducted to test the datasets of jokes from AGENTS, comparing the HRI with a robot using the developed jokes and algorithm (funny robot) and the HRI with a robot that did not use de developed system (unfunny robot). This study measured physiological responses (electrodermal activity, heart rate), humor style, attitudes and perceptions of robots, enjoyment with the interaction and the game, future intention to interact with robots, and behavioral responses during the HRI of 58 participants. According to the project researchers, preliminary results indicate that when compared to the robot without humor skills, the funny robot was rated as warmer and as perceived as having greater emotional value. There was also more interest in interacting with the funny robot in the future. Raquel Oliveira, together with the team recently presented the preliminary results at CMU Portugal Summit 2022 in poster format.
In the course of one year, the project AGENTS created datasets of jokes both in Portuguese and English and developed a machine learning techniques. Preliminary results testing the developed materials suggest that people perceived the funny robot as warmer and more competent than a robot without humor skills. Concurrently, people would rather play with the funny robot than the unfunny robot.
These results provide evidence that HRI may be improved with the use of humor. Robots or systems that have a 'human touch' are more likely to be chosen for future interactions. Still, these results are preliminary and future analyzes should help to better interpret these results.
Implications and Recommendations
The end-goal of AGENTS was the implementation of humoristic interactions in the context of a group card game involving more than one human and more than one robot. This was achieved and lead to an increased value perception of the robot and by the increased intention to interact again with it in the future. The project was disseminated in several scientific events, but also in different formats, such as the 90 segundos de ciência podcast.
This project presented a valuable opportunity to collect data on the users’ behavioral and physiological responses of positive emotions and well-being, regarding the humor displayed by the robots, allowing for adjustments in the model used.
It is expected that the generated data may inform future systems that use humor in HRI, being it in people's personal use (smartphones or computers) or in specific contexts (service robots in hospitals or restaurants). Future research in HRI will also benefit from AGENTS findings.