Learning from other Domains to Advance AI Evaluation and Testing: The History and Evolution of Testing in Pharmaceutical Regulation

Modern pharmaceuticals are inherently dual and uncertain in nature. Like any other technology, pharmaceuticals carry treatment value and the threat of harm – both of which are unknown – and are regulated accordingly. While they promise relief and even curing of serious diseases, they also carry risks even when they are effective. Pharmaceutical regulation is specifically grounded in this duality and uncertainty, aiming to learn about their benefits and risks and then weigh the benefit-risk balance. Modern pharmaceutical treatment co-evolved with systematic testing regimes, and the rise of the modern biotech industry is inseparable from that of the testing and regulatory regimes that have been established to govern its products. Modern pharmaceutical regulation is best understood as a system of approval regulation, that is, experimental minima combined with potential state veto of research and development (where veto is premised upon subset of experimental results).
The link between experimentation and state gatekeeping power is as pronounced in the pharmaceutical realm as in any other domain of economic and social activity. The fact that a new or modified molecule must receive regulatory authorization in order to be sold (and hence generate revenue) underpins a vast system of global experiment. Approval regulation as a regulatory-experimental system evolved due to scientific innovation, public and economic demand for information about products, and political pressure forged through critical historical episodes.