IGF 2023 Town Hall #114 The narratives we want: Which stories are NLP telling us?

Round Table - 60 Min

Description

A growing number of cultural products are being created using AI, specifically NLP systems. The capability to produce music, books and poems is astonishing. However, there is an inherent issue in the stories NLP systems are capable of telling us. Many NLP systems are trained on text data that contains gender and race biases and stereotypes, leading to reproduction of male perspectives and the according structural exclusion of female alternatives result in biased knowledge. This knowledge is then becoming the foundation for nlp systems, inherently perpetuating stereotypes as well as a limited view on what we believe is the truth about socio political discourses.

Which effect will it have, if we increasingly rely on NLP as a producer of culture? Which role will and can NLP play in cultural transformation? Which stories are NLP not telling?

With this session we will (1) discuss the potentials of AI for cultural transformation (2) examine the narratives of AI generated cultural products such as film, music and literature (3) assess NLP generated narratives specifically from a gender and race perspective (4) deduce potential risks of NLP generated narratives (5) identify policy recommendations for governing NLP systems and implications for cultural policy.


1) Social Media Engagement: Encourage the use of event hashtags and social media platforms to foster interaction and engagement. Participants can share their thoughts, ask questions, and interact with speakers and other attendees through social media channels, bridging the gap between onsite and online audiences.
2) Moderation between online and offline participants through two moderators: One onsite moderator will facilitate and manage the virtual and live interactions and ensure a smooth flow of communication between both groups. The online moderator will monitor online chats and relay questions from online participants to onsite speakers.

Organizers

iRights.Lab
Mariel Sousa, Research & Projects, iRights.Lab, Berlin, Germany
José Renato Laranjeira de Pereira, Co-Founder Laboratory of Public Policy and Internet (LAPIN), Brasilia, Brazil and Research & Projects, iRights.Lab, Berlin, Germany

Speakers

Makoto Shozu, Sound Designer
Maite Taboada, Professor of Linguistics, Simon Fraser University
Boyang Li, Nanyang Technological University, Singapore
Rea Eldem, Diversity consultant, IN VISIBLE
Li Lucy, University of California, Berkeley
Lucas LaRochelle, artist and designer

Onsite Moderator

Mariel Sousa

Online Moderator

José Renato Laranjeira de Pereira

Rapporteur

Mariel Sousa

SDGs

5. Gender Equality


Targets: This proposal is closely linked to the SDG target 5 Gender equality" by addressing the impact of NLP systems on cultural production and the perpetuation of gender and race biases. It aims to foster discussions and formulate policy recommendations to ensure that NLP systems contribute to inclusive and diverse cultural narratives, promoting gender equality and challenging stereotypes.
1) Potentials of AI for cultural transformation: The proposal aims to discuss how AI, including NLP systems, can contribute to cultural transformation. It acknowledges the capability of AI to produce various cultural products such as music, books, and poems and explores the opportunities and challenges that arise from this transformation.
2) Narratives of AI-generated cultural products: By examining the narratives of AI-generated cultural products like film, music, and literature, the proposal aims to shed light on the stories being told by NLP systems. It recognizes that NLP systems trained on biased text data may reproduce male perspectives and perpetuate stereotypes, leading to a limited view of socio-political discourses.
3) Gender and race perspectives in NLP-generated narratives: The proposal specifically focuses on assessing NLP-generated narratives from a gender and race perspective. It aims to uncover the biases and stereotypes present in these narratives, highlighting the need for more inclusive and diverse storytelling.
4) Potential risks of NLP-generated narratives: By deducing the potential risks associated with NLP-generated narratives, the proposal recognizes the importance of understanding the impact of biased and limited perspectives. It emphasizes the need to address these risks to ensure cultural products generated by NLP systems promote equality and avoid perpetuating harmful stereotypes.
5) Policy recommendations for governing NLP systems and cultural policy implications: The proposal seeks to identify policy recommendations for governing NLP systems and their implications for cultural policy. This demonstrates a commitment to ensuring that cultural production through NLP systems aligns with principles of gender equality and addresses potential biases and limitations