AI Shutterstock

Programme Overview
Shutterstock is a leading global creative platform offering full-service solutions, high-quality content, and creative workflow solutions for brands, businesses and media companies. Working with its growing community of over 2.0 million contributors, Shutterstock adds hundreds of thousands of images each week, and currently has more than 400 million images and more than 24 million video clips available along with a comprehensive collection of 3D models and music. In July 2021 Shutterstock announced Shutterstock.AI along with three start-up acquisitions: Pattern89, Datasine, and Shotzr. Shutterstock.AI powers new recommendations, predictive performance products and services that support the entire creative process. This will help customers make more informed content choices to ensure that what they produce is going to resonate with their audiences. The massive data and the renewed investment in AI at Shutterstock have called for the birth of a new Diversity and Inclusion Program.

Purpose of the Programme
At Shutterstock our goal is to build a workforce, contributor network, content library and AI products that are a reflection of the diverse global community we serve. We are proud of our rich mix of backgrounds and perspectives. We want to nurture this diversity by making sure all employees feel included and supported, our product, search and content offerings satisfy our clients’ diversity needs, and that Shutterstock is living its commitment to DEI through the way we do business, operate and build AI technology. Furthermore we aim to become a beacon in helping the AI community to establish guiding principles and common practice towards Ethical AI which starts with diverse data, an audited AI process, and inclusive teams.

Details of Specific Initiatives
Our view is that an inclusive culture is a shared responsibility of all of our employees. Led by our leaders, the focus is to ensure that all employees understand our DEI mission and know what their role is in achieving that mission.

We created a four pillars initiative:
– Creating Diversity in Data. At Shutterstock we source content with diversity via a network of over 2 million contributors from more than 150 countries. But, we don’t stop here. In 2020, we established “The Create Fund” to empower historically excluded artists, help fill content gaps, and further diversity and inclusion within our content library and contributor network. The Create Fund approach diversity in our content in two ways: (i) Artist Investment Program: Through our Artist Investment Program, Shutterstock is providing mentorships, cash grants and other support to empower historically excluded artists in the stock content industry. In 2021, we launched a webinar series in partnership with Diversify Photo and Black Women Photographers to promote the program among the artists in their networks. In 2022 we partnered with People of Color Collective in the UK to provide grants Black and Asian artists and will continue to support their members throughout the year. (ii) Grant Contest Program: Shutterstock is partnering with a variety of leading organizations to help change the conversation about important diversity and inclusion topics. We aim to help lessen the barrier for-and invest in-artists committed to highlighting diversity, inclusion, social justice, and environmental awareness through the content they create. Our partners have included It Gets Better Project and the American Society (ASA) on Aging. In 2022, we published a guide in partnership with the ASA to help our contributors add inclusive metadata in our platform.

– Audited Process. Machine-learning models are commonly trained on large amounts of real world data. This could entail the risk of inheriting human biases as it is observed in everyday life. Machine learning models can be very powerful and without proper testing not only could recycle these biases but could also enhance these phenomena. Therefore, it’s absolutely necessary to establish a rigorous testing process for biases before these models reach production. At Shutterstock, we test our AI models for a number of different biases from gender to ethnicity and behavioral. We do this by utilizing a large curated collection of images which we use to profile our AI models. We perform a number of statistical tests to ensure that there is no preference for a certain subset of characteristics in the data that could indicate the possibility of biased results. If an AI model does show some bias at our bias testing step then it is not deployed to production but goes back to the designing stage for retraining.

– Diverse and Inclusive Teams. This effort starts with building a diverse tech talent pipeline. We have established partnerships with organizations, like Black Girls Code, Black Professionals in Tech Network, Women Who Code and Women in AI that are on the frontline of increasing diversity in Tech. Through these partnerships, we are providing role models and additional support to develop those who have been historically excluded or underrepresented in the tech industry. We have already reached 40-60 gender balance in our first AI team which is an enormous improvement compared to the average 22% participation in this field. This team also brings diverse culture and backgrounds as its members are from South Africa, Spain, US, Canada, Malaysia, Ireland, Greece and Italy.

– Ethical Training: Machine-Learning models have the potential to solve the world’s biggest challenges, but they can also pose risks to individuals and groups. As we develop this technology, we must consider its ethical ramifications. Responsible design and use of AI starts with training the technical teams to question their inventions with wider social, economic and cultural perspectives. We have partnered with the World Ethical Data Foundation (https://worldethicaldata.org/ ) which is developing a specialized training for Shutterstock’s employees to raise awareness of the societal impacts of AI technology and to give our technical employees the tools and knowledge to pursue responsible AI.