On the Role of Transparency in Building Inclusive and Trustworthy AI Systems
Didar Zowghi · 2025
Transparency is widely acknowledged as a foundation of responsible, inclusive, and trustworthy artificial intelligence (AI). This talk examines how transparency requirements can be articulated and operationalised throughout the AI system lifecycle. As a case study, I draw on the Australian Government’s mandate requiring public agencies to publish AI transparency statements. Based on an analysis of 100 such statements, I highlight the breadth and depth of interpretations and reveal key patterns and gaps in how transparency is currently understood and communicated. I argue that core practices from requirements engineering; especially continuous stakeholder engagement; can provide a structured foundation for embedding transparency and inclusion into the design, development, and governance of AI systems. This approach moves beyond superficial compliance and toward transparency as a socio-technical commitment grounded in shared understanding and traceable justification.