A student writes "discussed about the paper" in a practice piece, and our writing feedback marks it. That is what the product does today. When we were asked whether it should, the honest first answer was that we did not know. "Discussed about" is not a slip of the pen. It is ordinary in Indian English, where it is a documented feature of the variety rather than an individual mistake. What our writing feedback found was a difference. What it reported was an error. Nobody sat down and decided that. It arrived as the default and it stayed.
The way we have come to put it to ourselves is that someone who has learned Indian academic English would come over here and need to learn English academic English. That sentence does more work than it looks like it does, because it names two things rather than one thing and a failure to reach it. Braj Kachru's three circles are the formal version. The Outer Circle, which includes India and Nigeria, is where English was institutionalised through colonial administration and has since been codified, with its own reference works and its own literatures. A person writing Indian academic English is following conventions rather than missing ours.
If that is right, a student arriving at a British university is not being repaired. They are being asked to hold a second set of conventions alongside a first set they already command, and to know which one is in play. That is a harder description of the task than the one we started with, and it is also a more accurate one.
Phillipson's five tenets, and the two we shipped
Granting the varieties does not settle where they came from. Robert Phillipson's Linguistic Imperialism argues that the spread of English was policy rather than weather, promoted by institutions including the ones that trained English teachers, and that it served British and American interests. Chapter 7, pages 185 to 215, is the part that matters for us. He takes five tenets, which he traces to a 1961 conference at Makerere in Uganda, and argues that each is a fallacy: that English is best taught monolingually, that the ideal teacher is a native speaker, that an earlier start is better, that more English produces better results, and that other languages used freely will drag standards down.
It is worth knowing where he stood when he wrote it. He had worked for the British Council for nine years. The line he later used as an epigraph, taken from the British Council's annual report for 1987 to 1988 and quoted in his 1994 article in the International Journal of the Sociology of Language, describes English as Britain's real "black gold", worth more than North Sea oil.
Two of those five describe us, and we chose neither of them on purpose. We ship a streak, and a streak is a machine for the assumption that more exposure produces better results. Paul Nation's Four Strands, set out in 2007, is the direct answer to that assumption: deliberate language-focused study is worth roughly a quarter of a learner's time and the rest belongs to other kinds of work. The second is the target itself. The phrase we reached for early on was performance as if the student were working in their first language, which is the native speaker tenet in product language.
A third one we assumed had landed turns out not to have. A good deal of the feedback our tutors give is delivered in the student's own first language, which is close to the opposite of the monolingual tenet, and that was a build decision made long before any of this reading.
Bisong's reply, and the default we had not chosen
Where Phillipson lost us was in what his account implies about the people inside it. Joseph Bisong replied in ELT Journal in 1995, volume 49, issue 2, pages 122 to 131, writing from Nigeria. Nigerians who choose English are choosing pragmatically, from inside a repertoire that already holds several languages, and reading that choice as false consciousness is itself condescending. Our own reaction, arrived at before we had any theory to dress it in, was that Phillipson was condescending and that Bisong was not. Alan Davies made the structural version of the objection in the Journal of Multilingual and Multicultural Development in 1996, volume 17, issue 6, pages 485 to 496: if the dominated say they want English and the theory hears domination in the saying, the theory cannot be wrong about anything.
We should take less comfort from that than we would like to. Phillipson's real point is that treating the current arrangement as a natural fact is itself an effect of the arrangement, since it was made by people and could be unmade by people. "That is kind of just the way it is" is a sentence we use often, and it is precisely the sentence he is arguing against. What we would say instead is narrower. The arrangement will not be unmade inside the three years of somebody's degree, and that person still has essays due in March.
Adrian Holliday's name for the ideology underneath the whole argument is native-speakerism, coined in a 2005 book and restated in a three-page piece in ELT Journal in 2006, volume 60, issue 4. His examples are job adverts and the phrasing of marketing copy. Ours would be product features: a single accent in the AI voice, or a conversation partner that treats every deviation from one target as something to fix. Asked whether our AI voice should have one accent or several, we had no considered answer, and the default we had assumed was British. Offering a choice, and including varieties of English rather than only British and American, is technically straightforward and we have not built it. Saying that is cheaper than doing it, and we would rather record the debt than describe an intention as a feature.
The objection we cannot answer
Then there is the paper we cannot answer. Nelson Flores and Jonathan Rosa, in the Harvard Educational Review in 2015, volume 85, issue 2, pages 149 to 171, take the standard progressive position, that all varieties are valid and students should additionally acquire the one appropriate to academic settings, and argue that it concedes almost nothing. Racialised speakers are heard as deficient however closely they conform, because the judgement is happening in the listening rather than in the speaking. They call the audience the listening subject, and their claim is that the deficiency is manufactured there. If that is right, a product that teaches a student to produce appropriate academic English is promising something it cannot deliver, and it moves responsibility for somebody else's prejudice onto the person receiving it.
We accept that this is true to an extent. Asked whether there is a version of this product that works on the listener rather than the speaker, and whether we would build it, our answer was no, and we are not going to soften it now that we have read the literature. We think the mechanism is real and that it sits with the person marking the essay. We do not intend to act on it.
A student can do everything the app tells them and still be marked down for who they are, and nothing we ship touches that. Work addressing the listener does exist, and it is done by people who work with markers and departments rather than with students. We are not those people. The most we can honestly say is that the two kinds of work do not compete for the same ground.
The reason we keep going anyway is not originally ours. Steve Kirk, in his 2019 BALEAP plenary, warns his own field about what he calls a risk of Critical EAP and Academic Literacies oriented pedagogies: "an over-focus on critiquing 'knowledge of the powerful', rather than (also) facilitating access to 'powerful knowledge'". The distinction those two phrases draw is Michael Young's, from 2008. That caution was written for people who already hold the critical position, and it would be cheap of us to pick it up and use it as a shield. So we will state only what we take from it. Command of a convention is worth having even where the convention is arbitrary and the marking of it is unfair, because the students with the least room for error are the ones most damaged by not having it. That is a reason to keep doing the thing. It is not a reply to Flores and Rosa, and presenting it as one would be dishonest.
Where the accent instinct stops working
Accent is the one place where we had an instinct before we had a reading. An accent that is understood is fine, and our accent is an accent. That is an intelligibility criterion, and it is close to where Jennifer Jenkins and the English as a Lingua Franca researchers have been for two decades. The point that applies to our setting is arithmetic: most academic English spoken in a British university happens between people who did not grow up speaking it, so interaction between two second-language users is the ordinary case rather than the exception. Jenkins and Constant Leung argue from that ground against standardised English tests in Language Teaching in 2019, volume 52, issue 1.
The trouble is that we applied intelligibility to accent and did not apply it to word choice, and no linguistic principle supports the split. "Discussed about" is a systematic feature of a variety in the same way that a rolled consonant is.
The line we can actually defend is not about language at all. Accent is not marked in a written assignment. Word choice is. So our real criterion is what gets assessed, which is a claim about a marking process rather than a claim about correctness, and it puts us straight back in front of Flores and Rosa, because what gets assessed is exactly what they argue is arbitrary.
The standard we apply, and the language we explain it in
The goal should be access, not assimilation. Two things follow that we can be held to. The standard the product applies to a student's writing is British academic English, and it is not a setting, because the product exists to prepare people for assessment at British universities. What is selectable is the language the explanation arrives in: the student's own first language, plainer English, or academic English.
Two claims come out of that and neither of them is large. The standard is named rather than inherited silently, which matters because a corrector with no stated standard drifts toward American convention by default, that being mostly what the training data is. Separating the standard from the language of the explanation is itself a position, since the thing being taught and the language it is taught in do not have to be the same. That is a refusal of the monolingual assumption at the level of design rather than at the level of rhetoric. It is a sensible design decision and not more than that.
The second thing is not built at all. What our writing feedback ought to say when it meets "discussed about" is that this is standard in Indian English, that British academic writing uses "discussed", and that the person marking the essay will expect the second. Nothing instrumental is given up by saying it that way, since the student still ends up knowing which form the marker expects. What changes is that the description of what they wrote is accurate. We have not built it. It is the change we most want to make.
What we do not know is whether a student on the receiving end can tell the difference. Being told that your usage is a legitimate feature of a variety and that you should write it the other way anyway might read as respect, or it might read as a longer and more polite way of being corrected, and which of the two it is probably depends on who is saying it and what is riding on the assignment. We have not asked anybody. Until we have, "access, not assimilation" describes what we are aiming at rather than anything a student has reported experiencing.
