The Challenge
Hundreds of hours of video content is being streamed across multiple social media platforms every hour. More than 100,000 moderators from across the globe are currently working to identify and remove inappropriate content. The work is laborious, time consuming and highly stressful. As the volume of content continues to rapidly rise, relying solely on human moderation of content is becoming increasingly impossible. We explored a unique, blended content moderation service whereby our AI technology will moderate and analyse content, providing support and key insights to the human role.
What was Involved
Working with AI Analytics and Software Engineering teams, Fjord were asked to envision the next generation of content moderation and enrichment methods for efficient and effective content review at scale.
It was a 8 week project and the core team was made up of 3 Designers, 3 AI Analytics, 5 Software Engineers, 1 Researcher and a subject matter expert. The Advanced Analytics team were focused on developing a Proof of Concept application which streamed live video and interprets the footage in real-time. We worked closely with them, sharing insights from our user research and asking many questions about the feasibility, maturity and constraints of the proposed technology solution. In tandem, we develop a proposed workflow for video moderation and enrichment that includes all relevant human and automated processes and tasks which followed the lifecycle of a live video and highlighted how AI would augment the human reviews role.
Outcomes
The team developed a technical prototype which demonstrated the technology capabilities and highlighted what was happening under the hood to illustrate how the technology could be utilised.
We also designed a proposed future workflow for the process of content moderation to illustrate how introducing this AI solution would alter the existing role of human moderators while increasing the efficiency and effectiveness of the current process. This highlighted the tasks and cases that AI would not be able to complete and how the human and AI needed to collaborate by giving feedback to train and aid eachother. We created wireframes to communicate the impact it would have on the end users and supervisors.
I was heavily involved in the research and development of this piece of works and responsible for communicating the outputs of the 8 weeks to stakeholders and clients. The video below describes out proposed workflow.