1

Treat societally impactful scientific insights as open-source software artifacts

We should rethink the way we publish academic output. Best practices in open-source software development may hold important keys.

Can't trust the feeling? How open data reveals unexpected behavior of high-level music descriptors

When run 'in the wild' by the community, high-level music descriptors may not perform the way they did in the lab.

Oracle Issues in Machine Learning and Where to Find Them

To what extent can we trust 'ground truth' in supervised machine learning to be a reliable oracle?