Video Privacy Notice Project
This first AI in Focus report presents research findings about the utility of social media-style video as a supplement to text-based privacy notices.
Jonathan A. Obar, David Gelb, Desirée de Jesús, Guilherme Cavalcante Silva, Hannah Palmier Blizzard
York University, Toronto, Canada
Report Summary
A meaningful consent process online requires that a digital service provider communicate information about services to consumers. For example, information about data collection, algorithms, etc. should be conveyed to help people understand the risks and implications of service use. This study addresses how video notices can help digital service providers communicate this information to consumers.
This report includes information about four video notice prototypes developed and focus group-tested. The report also includes five video design recommendations resulting from the focus group analysis. As the video content developed for this study addresses privacy issues raised by artificial intelligence (AI), the video prototypes and design recommendations should be relevant to organizations providing individuals with information (i.e. notice) about AI development or use.
This research responds to the concern that few people read text-based privacy notices. Information protections based on notice-consent frameworks require that individuals are provided information about a service to review before deciding whether to provide consent. Helpful notice materials and associated oversight protections linked to meaningful consent processes are vital, especially for members of marginalized communities who are most likely to suffer AI-inequities.
To improve information protections, the Office of the Privacy Commissioner of Canada, in its Guidelines for Obtaining Meaningful Consent, recommends service providers make notice and consent processes more creative and dynamic, while also considering consumer needs. This report’s findings suggest video notices can align with these recommendations.
Video Notice Prototypes
In mid-late 2023, four social media-style video notice prototypes were developed. Each was designed to supplement a text-based privacy policy about AI.
Each video was about one-minute or less, with vertical-orientation, live-action or animated, and designed in a popular format common to TikTok in particular, with accompanying background music.
Video One: How Our AI Works
A 60-second animation providing information about a digital service provider’s AI system used to assess a job candidate.
Video Two: University Application
A 26-second live-action video conveying information about AI systems that assess applications for university entrance.
Video Three: AI Didn't Pick Me
A 52-second, Wes Anderson meme-inspired live-action video providing information about an AI system used to assess a job candidate.
Video Four: Data Retention
A 63-second animation conveying information about a digital service provider’s data retention policy and AI development.
In early 2024, 17 individuals participated in focus groups to discuss the videos. A qualitative thematic analysis organized five recommendations for creating video supplements to text-based privacy notices.
Video Notice Design Recommendations
These recommendations should be considered by digital service providers, policymakers, and others interested in improvements to notice materials.
Recommendation 1: Create Dynamic Video Supplements for Text-Based Privacy Notices
Key Findings: Social media-style video notices are preferred over text-based notices. Innovative and creative approaches to producing and sharing video notices may support more dynamic and ongoing consent processes. This requires that video notices are designed to keep people engaged. The following recommendations provide strategies for facilitating that engagement.
Recommendation 2: Design Brief, Live-Action Videos
Key Findings: Many focus group participants prefer brief, live-action videos over longer, animated videos. Live-action videos seem to offer an innovative opportunity to facilitate engagement by creating connections to the content.
Recommendation 3: Create Videos that Notify about AI Privacy Implications
Key Findings: Study findings suggest brief, social media-style videos can generate awareness of AI privacy concerns via notification. To address the challenge of conveying AI complexities in under a minute, digital service providers should consider creative approaches to providing a series of brief, iterative videos.
Recommendation 4: Ensure Representation for Marginalized Communities
Key Findings: Focus group participants note the importance of representation for marginalized communities in video privacy notices. Digital service providers should ensure representation helps create stronger connections to content, while enhancing notifications for those more likely to experience AI-inequities.
Recommendation 5: Leverage Social Media Trends
Key Findings: Video notices should utilize trending memes to help interest people in privacy and AI content. To benefit from this innovative and creative approach, when leveraging a social media trend digital service providers should ensure content is timely.
To support the delivery of information protections in AI contexts, meaningful consent processes must be improved. The research findings described in this report align with the OPC’s Guidelines for Obtaining Meaningful Consent, suggesting that creative, dynamic, and consumer-focused video notices may help supplement problematic text-based privacy notices. Further research is necessary to determine how social media-style videos can move beyond notification and help explain in detail how AI systems work, as well as the implications of AI development and use.
Acknowledgments
Thank you to Grace Noua, Sam Loiselle, and Hannah Palmier Blizzard for helping our team create the video prototypes. Thank you to Brenda McPhail and Julie Ng for consulting on the video design. Thank you to Gilbert Bélisle for support with French translation. Thank you also to the Canadian Civil Liberties Association and the Canadian Civil Liberties Education Trust for helping to support this project.
The project is funded by the Office of the Privacy Commissioner of Canada, the Social Science and Humanities Research Council, and York University. We are grateful to our funders. Thank you for your support. The views expressed do not necessarily represent the views of the funders.
Report design and illustration: Hannah Palmier Blizzard