Guest post by Laura Summers, organiser of the Fair ML Reading Group
“Ethics is arguably the hottest product in Silicon Valley’s hype cycle today, even as headlines decrying a lack of ethics in technology companies accumulate.” (1)
You’ve probably heard about the problems of data bias in the popular media – we seem to have reached peak saturation with concerns about Machine Learning amplifying existing structural bias. What you might be unaware of, though, is the existence of an emergent field of research exploring mathematical definitions of fairness. The quest to define a formalism which helps us better understand, define and encode our ethics is a fascinating and appealing vision, and it’s the topic of a reading group we’ve been holding at Silverpond for most of 2019.
Fair ML Reading Group has been mostly concerned with the mathematical approaches to ethics, rather than the cultural, procedural or design approaches. Which is not to say that we are de-valuing those other vectors, but rather that it’s too much to tackle in one group.
Why would or should we try to use a computer to fix a problem caused by computers? Well to put it simply, we just don’t have enough human-power to assess and check all the decisions currently being made by machine learning algorithms, and the scale and breadth of their implementation is only set to increase. We might want occasional human oversight or second-opinions. But if we don’t want to grind these systems to a halt, we need to find ways to write code tests and set fairness thresholds that give us a baseline of confidence that our systems are acting appropriately.
If this sounds interesting, feel free to follow along out reading list or join the group below.
1) Owning Ethics: Corporate Logics, Silicon Valley, and
the Institutionalization of Ethics
Jacob Metcalf (Data & Society), Emanuel Moss (Data & Society, CUNY Graduate Center), Danah Boyd (Microsoft Research, Data & Society)