Comedy & AI Ethics: Navigating Bias, Anonymity, and Industry Tension

Table of Links

Abstract and 1. Introduction

2. Methods
3. Quantitative Results and Creativity Support Index
4. Qualitative Results from Focus Group Discussions
5. Discussion
6. Mitigations and Conclusion and Acknowledgments
7. Ethical Guidance References

A. Related Work on Computational Humour, AI and Comedy

B. Participant Questionaire

C. Focus

7 ETHICAL GUIDANCE

7.1 Ethical considerations

Our empirical study was approved by the ethics board appointed by our institution, who considered adverse impacts of LLMs upon participants (e.g., exposure to harmful and biased LLM outputs), the right to withdrawal without prejudice, and the compensation of participants. Participant anonymity was an additional requirement for the study, which explains why we did not collect demographic data from the participants.

At the beginning of the session, the objectives of the workshop were discussed with the participants.

The study contains some offensive language, which is mentioned as a trigger warning at the beginning of the paper.

7.2 Researchers’ positionality

The question of researchers’ positionality was directly asked by some of the workshop participants, who asked “what side we were on?”, meaning the side of the AI or the comedians’. Workshop facilitators answered that they were passionate about building AI tools that are useful for the artists, that they were open to any criticism and wanted the industry to listen to these criticisms and change behaviour. The professional affiliation of the workshop organisers (Google DeepMind) was disclosed to the participants.

Two of the authors of this study (Piotr Mirowski and Kory Mathewson) have extensively used AI in professional comedy performance since 2016 through their human and AI duet HumanMachine[13] and AI theatre company Improbotics[14], a participatory theatre lab exploring the creative potential and ethics of generative AI. In that process, these two researchers have had numerous opportunities to discuss the subjects of AI with artists and academics. All participants of the workshop were aware of the authors’ dual affiliation and artistic exploration of AI tools for live performance. We all exchanged flyers and pitched our respective shows at the end of the workshop.

7.3 Adverse impacts

We considered reputation damage for the participants, due to the participation in a study on AI for writing comedy material. This risk is exemplified by some participants: “I have friends who don’t talk to me anymore, because they learned that I have an AI show going on here” (p1) or “I have a friend who is deeply upset that I was using generative AI in my flyers” (p2). This risk was mitigated by keeping all participants anonymous in the study, by requesting all participants to adhere to Chatham House rules [47] following the focus groups, as well as by anonymising the names of shows and details of comedy material in the participants’ survey answers and focus group transcripts. Furthermore, we reduced this risk by reaching out to prospective participants who had already advertised using AI in their process or who were discussing AI in their subject material (i.e., on their show listing for Edinburgh Festival Fringe 2023, or on their personal website and social media profiles).

The second adverse impact can be the intentional or unintentional publication of comedians’ work. We addressed this concern by asking the participants to remove details of their shows and personal material from the writing sessions, and by personally removing anything that was omitted from the survey results, writing sessions’ answers, and from the focus groups transcripts.

The third possible adverse impact could be advertising for the dissemination of LLMs as tools for writing. It is addressed, to some extent, by honest reporting of the concerns of the participants.

[13] https://humanmachine.live

[14] https://improbotics.org