InterSynth Lab is a research environment funded by WASP-HS and the Wallenberg Foundations. We bring together researchers from Linköping University, Södertörn University, Uppsala University, and Lund University to ask one deceptively simple question: what does it take for synthetic data to be trustworthy?

The name says it: interdisciplinary work on synthetic data. Computer scientists build the generators; social scientists, historians, and philosophers interrogate what those generators do to knowledge, representation, and accountability. Neither perspective is sufficient on its own.

What is synthetic data?

Synthetic data is data produced by an algorithm rather than collected from the world. A model — often a neural network — is trained on real data and then asked to generate new examples that look like the originals but are not. The technique is being used to train AI systems, share information about sensitive populations without exposing them, test software, and bridge gaps where real data is unavailable.

Why is it useful?

Synthetic data lets researchers and institutions work with data that would otherwise be impossible to share — clinical records, financial transactions, demographic minorities. It can balance underrepresented groups, accelerate model development, and support reproducible science.

Why is it contested?

Synthetic data inherits the assumptions, gaps, and biases of the data that produced it. When the originals undercount certain groups, the synthetic copies do too — sometimes more so. When generators are opaque, downstream users cannot tell what they are really working with. We study these dynamics and design ways to make them visible.

How can society benefit?

When done well, synthetic data can protect privacy, broaden participation, and enable research on questions previously locked behind data access barriers. Our work aims to help society realise those benefits without absorbing the harms.