Reading for the August issue of The Bombay Literary Magazine has begun. As usual, fiction submissions were capped at 400 entries, a number that was exceeded before the third week of the month had ended. We are a team of eight editors, six of whom read the stories in the first round. The upvoted stories (about a quarter of the whole pile) are then divided between the two remaining editors. After this and another round of reading (in which the two editors have exchanged stories they liked), the surviving stories, usually about 20, are read by the whole group. We then conduct an all-hands meeting, discussing each of the shortlisted stories one by one. A publication list is finalised by consensus.The methodThe above process means that each team member has to read about 60 or more stories. Our publication schedule leaves an editor with about five to six weeks to do this reading. That’s ten stories a week, about 40,000 words. As voluminous as a novella, you might say. Actually, reading a novella’s worth of words each week may not sound too extreme. But we do have lives outside TBLM, you see, lives that do not run on clean timetables, and this is unpaid work, no matter how beloved. Some of us like to be done with the reading as soon as it comes our way. Others (like me) prefer the radiation of a deadline, deferring the work till it can be deferred no further. Either way, it usually comes down to reading about 240,000 words in a couple of weeks.That’s a lot of words for that duration. And the task is much tougher than, say, if one were reading the same amount from a single book. These are, remember, three score or more stories, from different worlds, of different genres, from writers of different sensitivities and styles and experience and finesse. Each writer sends us something that is singular to them, but it inevitably becomes part of a plural for us.Having done this for a few years, I have had occasions to observe myself in the task. After enjoying a story immensely, I find moving on to the next one an onerous task, for I am desperate not to plunge into disappointment right after elation. Likewise, to read a bunch of duds one after the other and then staying alert enough to enjoy the formal experiment or descriptive density offered by the n+1th story requires considerable mental effort. Sometimes, my eyes begin to glide over a block of text, my mind scarcely registering the words’ intent. Does this mean that the story gets rejected? No. If there are enough redeeming features that I have been alert enough to catch, I might ignore the block I couldn’t parse. The risk of a false positive is anyway mitigated by the process: we only publish by consensus; each published story must, in theory, have had 16 approving eyes laid on it.But is the risk extinguished? What if each editor found the same block difficult to parse, but upvoted the story in the way of trusting the group to do the labour that they, in general operational stupor, did not undertake? Groups can have free-riders; I am not talking about that. I am talking about the error implicit in a process that relies on two ideals:Each individual shall pay the highest attention to each granule of a large amount of work (for no monetary compensation, mind), andThe group, through its varied and collective expertise, shall have better taste than an individual.These ideals provide the ideology, sure, but reality tends to reside in deviations. Ultimately, a litmag is a labour undertaken for no commensurate real-world benefit to the editors. Over time, with a growing reputation, some “power” may accrue to the collective, but it is always minuscule compared to the effort involved. The editors of a litmag remain love-of-the-thing workers, given to doing maddening things, like reading 240,000 words by 60 writers within two weeks. Individually, and when seeing a text in isolation, they can spot clunky phrasing and mixed metaphors and the oddness of bland triads. But when seeing a text among a series of texts, their perspicacity apropos complex phrases and spots of un-meaning can vary, as would any human’s. And though being a structured group and reading in tiers does limit lapses, there’s no guarantee that the bubble of mishap cannot rise. It is entirely possible, even in an all-hands meet, for editors to discuss a whole work and its effects, and to miss talking of specific lines or phrases about which there was but only a wee doubt in the back of someone’s mind.The AI scenarioI am proud of most stories we have published at TBLM. But there are some I am not too proud of; some even feel like mistakes. My consolidated experience tells me that this state of affairs is normal, that this is the best one can hope for. But the latest AI fiasco has added a new kind of mistake to the set: one about which people might speak with a strain of expert-like objectivity (faux objectivity, for Pangram or no Pangram, there is still no method that provides enforceable certainty) about something that our subjectivity, and, by definition, my subjectivity, has rewarded with publication. There is the prospect of higher scrutiny and accountability, even as a sister scenario, that of AI-written stories flooding the submission pile, promises to shrink any reward for the editor’s soul.I ask the world to be kind to litmag editors at this moment. This is not to defend the readers and judges involved in the Commonwealth Short Story Prize. Irrespective of whether it used AI or not, the Caribbean region short story doesn’t read as worthy of a high prize. And it is hardly unfair to assume that the prize had the resources to read better. What I want, and perhaps this is too nuanced a point that is relevant for too small an audience, is an acceptance that it does come down to resources – time and money – if one wants to minimise the error function implicit in tiered reading by groups. Most litmags have next to no budgets. With the volumes they have to manage, Pangram is not free. Adding it to the editorial workflow is an additional step. Add to that the fact that there is still some haze around the ethicality of inserting a writer’s work into Pangram for a check, and the only seemingly perfect way seems to be one where consent is taken from every submitter. And then there is the question of whether Pangram is absolutely right or not. Could it also be true that some human writing is itself leaning towards the so-called AI aesthetic? With what confidence can I, in the absence of certainty, reject a work saying it is not human-written?As always, I will be using my own faculties this time too. But a drastic shift has taken place. Earlier, the labour could be undertaken with some deemed allowance for mistakes of subjectivity; today, all at once, under the terrible spectre of the supposedly objectively discernible mistake of publishing something that has some AI-written elements, the margin has shrunk – while editorial subjectivity is loaded with fresh biases (against triads and outlandish metaphors and em dashes and adjectives like “immense”) and no exact method of help is available. It was a labour of love then and it will be a labour of love always, but it’s a tough love now.Also read:Has an AI-authored story just won a literary prize? What does it mean for writing and books?This article originally appeared on Tanuj Solanki’s Substack.