Your information isn't quite accurate anymore, actually in 2022 Elsevier and UC reached an agreement, and they renewed the subscription. Everyone in the UCs currently has full access to Elsevier articles, even the non open access ones. It was indeed difficult and harmful to researchers during the period when access was lost.

1. The old-school method is to simply send a request for a PDF to the contact email address. I usually get 24 hr turn-arounds on requests.

2. Check pre-print servers for the research.

3. Sci-Hub

One final comment on the impact of LLMs: If the need for a very specific paper is not critical AND IF the research research had deep content before the cut-off date of Claude 3.5 Sonnet, THEN one can often get good context on a topic in a useful Q&A format I have come to love.

The implications and impact of LLM on pay-walls are enormous. Most users and scientists do NOT need a specific reference. They need some information or insight on a particular question.

In this context I find Claude highly reliable when my questions are relatively broad (e.g: “Claude, please explain Wright fixation index, F<st>, and its use in comparing genetic similarities and differences among chimpanzee populations?”). For that question and others of this type I get great first-pass answers, better than I would from all but a dozen experts in the world.

In contrast if I ask Claude for specific citations on the fixation index in chimps it is very likely to provide a confabulated reference. I will occasionally luck out but every citation you get from an LLM now needs to be vetted unless you have a RAG front-end.

All of these methods are available to almost anyone with internet access.

I wonder why MIT PI's haven't found the same disastrous effects in the past 4 years... unless the summary at OP doesn't accurately represent actual faculty feelings...

Nowadays you can pay extra for open access even in a closed access journal and most do- so there are very few truly closed articles published anymore. It is mostly accessing older literature where it becomes a problem, but of course there are ways to get anything, especially if you are tech savvy. Most people in academia have access through some other means- their local public library, alumni access through the schools they went to previously, etc. ResearchGate has also semi-automated the 'request a copy from the author' route. People keep requesting my open access articles through that, which is annoying.

> It is mostly accessing older literature where it becomes a problem

Not sure if this was true with UC, but MIT still has access to all Elsevier content they had before, if it was published before 2020. )

> ...leaving users with immediate access to only pre-2020 backfile content...

I'm not sure the details, but I think there are contracts where the university gets to keep access to the stuff they used to have even when cancelling contact; when transitioning from paper journals, the thought was "If we cancel our subscription we don't lose all those paper journals, and don't have to pay for them again, we want something like that."

Now I'm curious the details of this. But the OP write-up was clear that all the pre-2020 "backfile" was still accessible. Also pointing out that as time goes on this will of course be a lower percentage of all requests, since it will remain fixed at pre-2020.

(update: during the period, UC said "most articles" 2018 and earlier were still available from backfiles for them too. Perhaps most packages included such a license but not all https://www.lib.berkeley.edu/about/news/elsevier-outcome )

> ResearchGate has also semi-automated the 'request a copy from the author' route.

While the practice is standard of authors handing out free copies, in most cases I think this is either a copyright or license violation with the publisher? The authors have either assigned copyright to the publisher (and do not have permission to distribute on their own), or assigned the publisher an exclusive license that also doesn't let them distribute on their own. Unless this has changed in the past ~5 years since i was more familiar with it.

The publishers don't really try to enforce this though, as long as it's individual one-on-one which is interesting. ResearchGate is pushing the boundaries a bit there -- not legally, it's all illegal, but in terms of at what point will the scale and ease of access be high enough that publishers try to get their authors to stop giving away for free what the publisher is trying to sell. Obviously it would look bad, and they don't want make enemies of their authors.

The whole thing is a bit odd.

> While the practice is standard of authors handing out free copies, in most cases I think this is either a copyright or license violation with the publisher?

Journals usually allow the author to share a specific number of free copies, and even give the author a system or link to facilitate it.