Correction: I have updated the ACL registration fee to the early reg rate, I originally gave the late rate.

I have been an NLG researcher for 35 years (I got my PhD in NLG in 1990). Many things have changed since 1990, some good and some bad, but one thing that I really do not like is the fact that in 2025, people generally have to pay to published peer-reviewed research. Which is not fair to researchers who do not have access to good funding, and is not good for science.

Conferences

In the ACL community, most papers are published in conferences. In 1990, ACL conferences were relatively cheap to attend; usually held on universities using university lecture halls and facilities, and often with an option for attendees to stay in student accomodation, which was much cheaper than hotels. So you had to pay to attend an ACL conference, but it was not expensive.

In 2025, ACL conferences are large enough that they must be held in conference centres, which makes registration much more expensive. Accommodation can also be expensive, especially if the number of people attending ACL drives up hotel prices (this may have happened in Dublin in 2022).

In recent years ACL conferences have had a virtual option, which reduces costs and also visa problems (Ive known foreign PhD students who had papers accepted at ACL but were not able to travel because of visa problems). Unfortunately, the virtual option is also becoming expensive. For example in ACL 2025, it cost $400 (plus $100 ACL membership) to present a paper virtually at ACL. I do not understand this; I appreciate the hiring conference centres and paying for food for attendees is expensive, but why does this matter for virtual attendees?

I think that the ability to present scientific results should not depend on how much money the scientist has access to. Unfortunately this perspective is beginning to look somewhat idealistic and out of touch with the real world.

Journals

Its has always cost money to attend a conference. The biggest change since 1990 is that it now also frequently costs money to present research in a journal.

In 1990, journal publication was free for authors; readers and subscribers paid, but not authors. Which is what you expect; publishers in general do not charge content creators, indeed in other contexts they pay content creators.

However, even in 1990 it seemed unfair that readers without money should not have access to research findings, so there was an increasing focus on what we now call open-access (free to read) publishing models. In AI, the first open-access journals were run by charities (such as JAIR) or professional organisations (such as ACL’s Computational Linguistics), and were free to both authors and readers.

Unfortunately, when open-source publishing spread to commercial publishers such as Elsevier, the publishers needed to make money, so they decided that if they could not charge readers, they would instead charge authors. So APCs (article processing charges) appeared everywhere. At least in my experience, if one of my students wants to publish a paper in a journal from a commercial publisher, the charge is usually thousands of pounds. Needless to say, this is far higher than the actual cost to the publisher of publishing the paper.

From this perspective, I really appreciate that the ACL journals Computational Linguistics and TACL do not charge authors; unfortunately they are the exception, not the rule.

Arxiv

I have mixed feelings about Arxiv. Its fine as a supplementary publication venue, but it still feels odd to me to see papers only in Arxiv, which have not been through peer review. However, Arxiv absolutely shines from a pay-to-publish perspective, since it is free to publish in and free to read! Which may be why I increasingly see papers which appear in Arxiv and nowhere else.

So perhaps Arxiv is the future after all…

This is wrong!

My students usually have some funding, but not a lot, so we need to be careful; paying thousands of pounds to present at a conference or publish a journal paper is painful, especially if it happens several times. I was recently asked to submit a paper to an interesting special issue of a journal, but declined when we found out that the journal would charge us £2300 to publish in the special issue.

We have developed some work-arounds, such as publishing in journals which have an agreement with Aberdeen University to reduce costs,. But sometimes this blows up; on one occasional a student was about to submit a paper to such a journal, when Aberdeen Uni announced they were cancelling the deal with the journal, which made it prohibitively expensive to publish in. So we had to change the paper and submit it elsewhere, which was a hassle.

And it shouldnt be this way! I didnt have to worry about pay-to-publish when I was a PhD student and postdoc, it seems wrong that my students and postdocs need to worry about this in 2025…