I'd offer the example of the intern dataminer at the Potti lab which was found to have committed a particularly egregious form of model fitting and led to its e

I'd offer the example of the intern dataminer at the Potti lab which was found to have committed a particularly egregious form of model fitting and led to its erroneous use in actual cancer patient treatment: by splitting obs from the total study into fit and test sets to give perfect agreement with the already trained model. Here the impact on ethical issues - life or death medicine - was pretty high. The guy who quit personally sacrificed his medical-school career to draw his foot in the sand, and received empty thanks only years later from Duke [1]. In data science, ethical issues seem to be present in all the interesting applications. But the life or death impact is often much less, like Netflix movie ratings - where a public datamining competition was cancelled due to a compliance-based class action lawsuit for the potential to de-identify anonymized account data [2]. It's hard to where on this ethical spectrum your particular case lies. If it's something that effects others, quit and try to make things right. If it effects lawyers v lawyers let your boss take accountability for that decision and the freedom to make it. [1] http://www.cancerletter.com/articles/20150109_1 http://www.cancerletter.com/articles/20150109_1 [2] http://www.forbes.com/sites/firewall/2010/03/12/netflix-settles-privacy-suit-cancels-netflix-prize-two-sequel/ http://www.forbes.com/sites/firewall/2010/03/12/netflix-sett...