Reimagining Evidence: Ethical Innovation with Synthetic Data in Collaborative Evaluation

What can evaluators do when data is inaccessible, sensitive, or incomplete? This session introduces the thoughtful use of Generative Adversarial Networks (GANs) to support collaborative evaluation in data-constrained contexts. Instead of using GANs as a purely technical tool, this proposal frames them as a shared opportunity for evaluators and stakeholders to simulate missing evidence, stress-test interventions, and enhance learning. Through three pillars—synthetic data realism, ethical safeguards, and collaborative scenario modeling—GANs can support equity-focused evaluations where direct data collection may be limited. Grounded in the principles of the Model for Collaborative Evaluations, this approach honors stakeholder participation while introducing an innovative pathway for generating trustworthy insights. Participants will explore when and how synthetic data can be responsibly used to expand inclusion, reduce risk, and build shared ownership. This session responds to AEA 2025's call to engage communities and share leadership through ethical, forward-looking evaluation practices.
Click to fill survey.

Michael Harnar; Michelle Rincones-Rodriguez; liliana Rodríguez-Campos