This article is a response to Mark Hanson’s Do we really need journals?. If I’ve understood correctly, Mark argues that social media isn’t a good replacement for journals because it has its own problematic biases built in. I agree that social media isn’t sufficient, but I don’t think this in any way makes the argument for journals.

As writers and readers of research papers, what we need is an effective process to match papers to readers. There are various properties we’d like that matching process to have. We’d like to make sure the matched papers are relevant, high quality and an unbiased selection. We might want it to be consistent and fair. We’d like to know that it doesn’t miss papers that are relevant, etc. We’d like the amount of time and energy spent on the matching process to be reasonable and proportionate. We’d like the process not to have negative second order effects on scientific careers.

Journals fail on almost every single one of these measures. The majority of papers in a given issue of a journal - even a specialist journal - are not relevant to most readers of that journal. Pre-publication peer review and the journal system don’t ensure high quality (multiple studies show how many errors are missed by peer review), indeed it encourages authors to hide weak points of their papers and in extreme cases to commit fraud. It’s highly biased in favour of well connected scientists at big name institutions in rich countries. (This is also true of social media and it would be interesting to see evidence of whether that bias is stronger or weaker than the journal bias.) The process is highly random, inconsistent and unfair. It regularly misses important papers that are very relevant to readers. The amount of time and energy spent on it are wildly disproportionate to the value of the filtering process. It has terrible effects on scientific careers, both because of the randomness, bias, effects on mental health, overwork, etc.

Instead, we need a diversity of approaches for paper matching. Curation (a more egalitarian generalisation of what journals do) can be part of that mix, as can social media. There’s also algorithmic recommendation (like semantic scholar), collaborative filtering, and arguably many more. On top of that, a fully open and transparent post-publication peer review system to enable us to find errors and judge paper quality.