Faculty member here, conducting research on Natural Language Processing, Machine Learning, and Artificial Intelligence. For me, 99% or more of all research that is worth reading these days is open access and published for free in non-profit, community-run journals and conferences. The only time I stumble upon pay walls is when I venture into older work or neighbouring research areas.
The only notable exception is Google DeepMind, that has a nasty habit of using Springer Nature. I am assuming it is because of the perceived prestige and disregard for the standards we otherwise have set as a community.
On a personal level, I refuse to review for any entity that maintains closed access journals. It is not sufficient that you offer open access, it should be the norm.
Lastly, all is not fun and games when it comes to open access. Even if your field and publications are always open, you are likely to be forced to upload your work to approved repositories (which always come with horrendous, slow, and fragile user interfaces). You must remember to do this within a given time window or suffer punishments akin to if you had published your work as closed access, marking the work as ineligibility to count towards funding, etc.
Question: how does one stay competitive if a field mostly focuses on high impact factor journals (eg nature) and one makes trainees publish in lower impact factor journals as a matter of principle.
Won’t one be hurting the careers of students by not publishing in the venue with the broadest audience?
Speaking for NLP in particular, and based on my experience, one doesn't publish in conferences "as a matter of principle". One does it because it's how everyone else in the field works. No one I know in the NLP field cares about Nature - if you want to get the broadest audience, you submit to one of the ACL conferences.
The one problem I'm aware of is those who want to get positions in universities that don't know how the field works, because their selection committees tend to complain that you have no journal papers. If that's your plan then you can submit to journals such as CL. But again, that's not where the broadest audience is.
If you are in a field “driven by” closed journals, you are highly unlikely to make career progress if you stay principled as funding, promotions, and so on will depend on publishing where others perceive the work as being good. Maybe if you are an utter genius that brings about a revolution you would not, but how likely is that?
However, I do not expect that the audience that matters will not have access to your work in some way, even without open access. If your university is a good one, they will have subscriptions. Plus, there will be networks of researchers illegally passing around documents. Rather, closed access hurts researchers at less privileged institutions and the general public, as they are unlikely to gain access.
This is more or less my experience as well. I’m not as involved as I used to be, but the only time I couldn’t get access to something I wanted in ML was Google’s TPUv4 paper. But a quick trip to https://www.reddit.com/r/scholar solved that.
The only notable exception is Google DeepMind, that has a nasty habit of using Springer Nature. I am assuming it is because of the perceived prestige and disregard for the standards we otherwise have set as a community.
On a personal level, I refuse to review for any entity that maintains closed access journals. It is not sufficient that you offer open access, it should be the norm.
Lastly, all is not fun and games when it comes to open access. Even if your field and publications are always open, you are likely to be forced to upload your work to approved repositories (which always come with horrendous, slow, and fragile user interfaces). You must remember to do this within a given time window or suffer punishments akin to if you had published your work as closed access, marking the work as ineligibility to count towards funding, etc.