26 de desembre 2019

Ethical algorithms like hammers?

Main messages:
All decision-making—including that carried out by human beings—is ultimately algorithmic. The difference is that human decision-making is based on logic or behaviors that we struggle to precisely enunciate. If we humans had the ability to describe our own decision-making processes precisely enough, then we could in fact represent them as computer algorithms. So the choice is not whether to avoid using algorithms or not, but whether or not we should use precisely specified algorithms.
Machine learning is a powerful tool that has many extant and potential benefits.  Technology companies such as Google and Facebook of course rely on products powered by machine learning for much of their revenue—but as these techniques grow in applicability, their scope and societal benefits grow as well.
The result is that, at least for a while, the critics of the algorithmic approach may often be right. There are many consequential domains where algorithmic tools are still too naive and primitive to be fully trusted with decision-making. This is because to model the forest, we need to start with the trees. This book offers a snapshot of exciting strands of research aimed at developing ethical algorithms, many of which are still in their very earliest days.