[kleiba]

I used to be a researcher in NLP until not too long ago. Over the last few years, increasing pressure has been put on on everyone trying to publish papers related to training data collected through human trials to provide ethics statements as a way to ensure that certain standards are met regarding that data. Not everyone is able to see too much sense in this process, and I've heard comments from non-US colleagues that it reflects a purely American world view. But I think that there are at least some applications where ethical consideration are definitely worth talking about, eg., in applications such as hate speech detection.

And of course, even if you don't agree with the necessity for ethics statements, because it is just one more thing that takes up time you could otherwise spend on your actual job (doing research), you certainly don't want to risk having your paper rejected just because you don't meet whatever ethics standards the conference or journal seeks to uphold.

But remember, I'm talking natural language processing here.

In that light, it is a complete mystery to me how research like the one described in the article could have possibly, ever made it past an ethics review. Unless, of course, completely different standards are applied - which in itself would be rather questionable.

[dekhn]

Why do you believe this work would not have passed an ethics review?

(your comments are a bit obscure, in a way that suggests you aren't familiar with how modern biological research is evaluated)

[kleiba]

I have no doubt that it passed an ethics review. I'm sorry if my comment seemed to suggest otherwise.

I suppose what I was trying to get at was a suspicion that ethics reviews in today's research landscape (not only in the medical field, but others as well) seem to me more like a lip service. And don't get me wrong, that's just an opinion, I'm sure a lot of you think otherwise.

[dekhn]

It is true that a great deal of review is lip service. A cynic would point out that while IRBs contain an ethical review, they often exist only to absolve the university of legal or moral responsibility. The realist in me points out that ethical review has been right-shifted: now we're more likely to approve a wide range of experiments that skirt right up against the common ethical boundaries, and then let the subsequent public response determine the ethics of the action.

None of this really matters until somebody announces we can make viable embryos from stem cells or that we can bring a viable embryo to term in an artificial womb. Both of these seem to be not implausible in the next 20 years.

[numpad0]

It's that modern computer science around Web is just total Stanford prison experiment when it comes to ethics. A/B testing, heatmap collection, user generated data collection, those are all human subject researches and should be required to have an independent board in a University.

These guys have established processes and justifications to do these experiments. It was always just wrong that Web has none, and it also makes no sense to assume that just because the process look substantial judged by norms of other industries they must have ignored it.