It is a decent idea, but it can be really hard to spot scientific fraud. As in REALLY, REALLY hard. Some fields, like machine learning, gets deluged with paper

It is a decent idea, but it can be really hard to spot scientific fraud. As in REALLY, REALLY hard. Some fields, like machine learning, gets deluged with papers coming from all kinds of researchers - there's no unified scientific notation, so to speak. In one paper you could have a pure math authors, in some other you could have econ authors, and then maybe a theoretical physics author. All these could use wildly different notation to describe related ideas, or discuss ideas to solve problems in their respective fields. And in that case, how do you detect bogus science? You could be a leading ML scientist, but not have the faintest idea about the other domain. Or the inverse - be a leading domain expert, but have little ML knowledge. Not to mention, the people that are publishing fraudulent scientific papers aren't morons - they're likely publishing that kind of stuff because of bad incentives. Maybe you'll catch the laziest, least capable fraudsters - but trying to catch savvy researchers sounds like a daunting task. In my mind, this type of task/job is more fit for full-time scientific "auditors".