Context:
When the OS 9 was introduced in 1999, Apple decided to include a pioneer joke generation system that was capable to interact with the user by delivering a joke in response to the command ’Computer, tell me a joke’. The jokes told by the system were scripted and allowed only for very limited social interaction. However, they were considered a worthy investment because it was believed that they provided the system with a ‘human touch’ (Hempelmann 2008). Besides, humor is also only a circumscribable portion of natural language as a whole and thus, constitutes a well-defined easy engineering goal. This humoristic feature has remained a constant ingredient in later versions of the OS and remains to this day. In fact, humor is a natural emergent feature in everyday conversations and its complexity is hard to capture through scripted, off-context interactions. Moreover, this system (and other similar ones) disregarded the users’ sense of humor and, as often happens in natural language processing, had the user adapting to the system instead of the other way around (Hempelmann 2008). The lack of context, the disregard of user’s preferences, and the over-reliance on certain formats of jokes (e.g. word puns, one-liners) are still limitations found in a large number of current approaches to humor detection, classification, and generation.
Our proposal:
The central idea of this proposal is that (a) humor is an important feature in human communication that can be leveraged to create more naturalistic and lifelike interactions with robots and (b) humor potentialities can be increased through the delivery of user-personalized humor in naturalistic settings. In particular, we argue that psychological models of humor and its’ everyday functions can be of use when attempting to create a top-down approach of humor that can be modeled to match each user’s preferences. We will use a 2x2 conceptualization of humor that involves the social function of humor (humor used to enhance oneself or used to enhance others) and the valence of the humoristic content (positive, negative) (Martin, 2003). Using this conceptualization, we propose the creation of a dataset of jokes and the application of supervised machine learning techniques that will allow us to extract and automatize multimodal humor delivery according to the style of humor of the user. The end-goal of this process would be the implementation of user personalized humoristic interactions in the context of a group card game involving more than one human and more than one robot. This is expected to lead to better interaction outcomes and increase the value perception of the robot, by contributing to greater user’s task enjoyment, more positive perception of the robots, and intention to interact again with these social agents in the future. The practical implementation of this scenario would also present 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 (e.g. laughter) and fine-tune our model with that information. All the materials and data collected will be shared with the community, including in open repositories (e.g. Open Science Framework), for researchers, software developers, and society in general, always complying in full with the General Data Protection Regulation (RGPD).
Relevance to CMU goals and the adequacy of the research team:
The project team is well-positioned to lead this research given their extensive interdisciplinary record of publications in high-ranking conference proceedings and journals in areas related to this project, such as Human-Robot Interaction, multi-modal approaches to humor, psychology, and computer science. Our proposal also adds to the current literature on humor by presenting an approach based on human behavior theories to humor classification and generation and considering users’ preferences in terms of styles of humor as an input variable to determine the type of humor that robots should deploy during social interaction with humans. This outcome will be relevant for future research in HRI and it is expected to have a high impact by contributing to the development of more socially effective robots, deployed in complex naturalistic interaction settings.
Our proposal:
The central idea of this proposal is that (a) humor is an important feature in human communication that can be leveraged to create more naturalistic and lifelike interactions with robots and (b) humor potentialities can be increased through the delivery of user-personalized humor in naturalistic settings. In particular, we argue that psychological models of humor and its’ everyday functions can be of use when attempting to create a top-down approach of humor that can be modeled to match each user’s preferences. We will use a 2x2 conceptualization of humor that involves the social function of humor (humor used to enhance oneself or used to enhance others) and the valence of the humoristic content (positive, negative) (Martin, 2003). Using this conceptualization, we propose the creation of a dataset of jokes and the application of supervised machine learning techniques that will allow us to extract and automatize multimodal humor delivery according to the style of humor of the user. The end-goal of this process would be the implementation of user personalized humoristic interactions in the context of a group card game involving more than one human and more than one robot. This is expected to lead to better interaction outcomes and increase the value perception of the robot, by contributing to greater user’s task enjoyment, more positive perception of the robots, and intention to interact again with these social agents in the future. The practical implementation of this scenario would also present 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 (e.g. laughter) and fine-tune our model with that information. All the materials and data collected will be shared with the community, including in open repositories (e.g. Open Science Framework), for researchers, software developers, and society in general, always complying in full with the General Data Protection Regulation (RGPD).
Research Centre | Research Group | Role in Project | Begin Date | End Date |
---|---|---|---|---|
CIS-Iscte | Behaviour Emotion and Cognition | Partner | 2021-01-01 | 2021-12-31 |
Institution | Country | Role in Project | Begin Date | End Date |
---|---|---|---|---|
Instituto de Engenharia de Sistemas e Computadores:Investigação e Desenvolvimento em Lisboa (INESC-ID) | Portugal | Leader | 2021-01-01 | 2021-12-31 |
Carnegie Mellon University (CMU) | United States of America | Partner | 2021-01-01 | 2021-12-31 |
Name | Affiliation | Role in Project | Begin Date | End Date |
---|---|---|---|---|
Patrícia Arriaga | Professora Associada (com Agregação) (DPSO); Integrated Researcher (CIS-Iscte); | Principal Researcher | 2021-01-01 | 2021-12-31 |
Reference/Code | Funding DOI | Funding Type | Funding Program | Funding Amount (Global) | Funding Amount (Local) | Begin Date | End Date |
---|---|---|---|---|---|---|---|
CMU/TIC/0055/2019 | -- | Award | Fundação para a Ciência e Tecnologia - 2019 Exploratory Research Proposals Under the Carnegie Mellon Portugal | 69435 | 28905 | 2021-01-01 | 2021-12-31 |
No records found.
No records found.
No records found.
No records found.
With the objective to increase the research activity directed towards the achievement of the United Nations 2030 Sustainable Development Goals, the possibility of associating scientific projects with the Sustainable Development Goals is now available in Ciência-IUL. These are the Sustainable Development Goals identified for this project. For more detailed information on the Sustainable Development Goals, click here.