Three Principles for Testing Safety-Critical Robotics and AI In Public.

We have recently emerged from truly desperate 5mes. With the global COVID-19 pandemic largely in our rear-view mirror, it is surprisingly difficult to look back and clearly recall how, for what felt like an eternity, our freedoms were shelved while we waited for a vaccine that would solve all our problems. There is no need to replay the details of that collective desire for a technological fix—you have likely done that with loved ones during countless post-pandemic get-togethers.

But it is worth recalling, for the purpose of the argument we make in this paper, that even in those most desperate 5mes we clung to a set of principled practices designed to help us find safe vaccines that we knew with some certainty posed few enough risks to individuals that we considered them acceptable to release for widespread use.