IVOW’s Kee Malesky at the 2018 #AICulture Symposium at Morgan State University.
by Davar Ardalan and Kee Malesky
We are the storytelling species. It’s natural then to want to discuss the future of storytelling in artificial intelligence and the cultural implications that come with it.
To consider “Artificial Intelligence, Culture, and Storytelling” more than 50 thought leaders from a variety of disciplines gathered for a Symposium co-sponsored by IVOW and Morgan State University in Baltimore on April 23. Representatives from the United Nations, the Australian government’s innovationXchange program, and Management Systems International, as well as renowned machine learning experts, educators, journalists, technologists, and business leaders from across the US and Mexico engaged in conversations about AI as a tool for culturally rich and socially conscious storytelling.
By early afternoon, our focus turned to the need for a future “Declaration of Machine, Citizen, and Culture” that could guide engineers, designers, machine learning experts, and users to understand and protect human rights, and to help ensure inclusivity and diversity in the digital world.
We considered the fact that we’ve been modern civilized human beings for about 10,000 years, with evolving levels of self-awareness that have allowed us to ask essential questions, experience individual consciousness and share it with others.
So we asked ourselves, How do we bring the Machine into this discussion of human rights? What issues/concerns are specific to culture-related AI applications? What does human-centered AI look like? What are the rights and privileges of human beings in the digital universe?
As with any new technology, it’s important to create guidelines on the proper ways to utilize these new tools. Do we need to create machines to hold other machines accountable and accurate, or a responsible third party to review new products before launch?
One participant pointed out early, we need to identify specific issues and current inadequacies; the problem isn’t in the algorithms, it comes from people and society. Data is agnostic and amoral and diverse datasets do exist; but people have biases. A multidisciplinary approach is essential to find balanced solutions. Systems will need to be trained to be aware of cultural context. Dominant biases have considerable power to negatively impact the lives of others; we have to keep humans accountable too. AI expert Mark Riedl of the Georgia Tech School of Interactive Computing, suggested that we should look to Europe and new laws around AI accountability.
AI expert Mark Finlayson of Florida International University urged us to pause and consider what the problem is first before making any declarations.
Lisha Bell, from Pipeline Angels, brought up the point that some biases have more power than others, and we need to hold humans accountable for making AI balanced and diverse. An AI system must be interrogatable; we should be able to understand why a system made a decision.
Louise Stoddard, who works with the United Nations Office of Least Developed Countries, asked “What are the barriers to access and how can we remove them? Who owns this process?” She stressed the need to listen to stories “from the ground.”
Ellen Yount, Vice President of Management Systems International (MSI), liked the idea of having a charter that encourages us to “Do no harm.” AI developers should consider the social implications of telling stories, as well as any unintended consequences.
A Lesson From History
Ahead of our conversation, we looked at some highlights in the history of human rights — and the granting or asserting of those rights, which have evolved over time and been extended to women, minorities, workers, children, disabled, immigrants and refugees:
We didn’t intend to produce an actual Declaration of Citizen, Machine, and Culture at the symposium; we wanted to begin a conversation between our IVOW team and a wide variety of experts. We are well aware that others are actively debating how to balance human needs and rights with the challenges of the digital universe, including these current endeavors:
As the declaration workshop came to a close, IVOW’s product designer Nisa McCoy pointed out that design isn’t perfect, and we need to look at the initial intention of any technology, and then review. We need to know more about a product and its impact before it’s launched. Apps and products must be safe, reliable, private, secure, inclusive, transparent, and accountable. We need a cross-cultural understanding of the audience and users.
After the session, one of the symposium attendees, Paris Adkins-Jackson, founder and CEO of DataStories by Seshat, a research, evaluation, and data analysis company, compiled this summary of the highlights of our discussion, putting it in the form of a declaration:
We declare that citizen, machine, and culture are inherently and essentially connected and in communication with each other; and
We declare that those connections produce inequities and amplify biases;
Thus, we declare that when engaging research, product development, or other actions related to Artificial Intelligence and Machine Learning, there must be a thorough examination prior to development of the potential impact of any product on people including access, appropriation, and amplification of injustice; and
We declare that such information will be a large component of the criteria in the decision to develop the product;
Also, we declare that a system of accountability be developed to mitigate any unforeseen challenges that arise which indicate there has been an adverse impact on people;
Lastly, we declare we will make all efforts possible to work with diverse and non-traditional disciplines to investigate impact, and develop and implement accountability platforms as well as to assist with product development.