Nelson Flores and Jonathan Rosa published an argument in 2015 that a language product cannot answer. It appeared as "Undoing appropriateness" in the Harvard Educational Review, volume 85, issue 2, pages 149 to 171. Their claim is that racialised speakers are heard as deficient however closely they conform to whatever standard is being asked of them. The judgement is produced in the listening and not in the speaking. They call the audience the listening subject, and their point is that the deficiency is manufactured there.
Nothing in our product reaches that listener.
If they are right, then a product which teaches a student to write more conventional academic English has not removed the thing that is going to be done to them. It has relocated the responsibility for someone else's judgement onto the person being judged.
We think that claim is true to a degree. Not the whole account of what happens to a student in a British university, in our reading of it, but real, and real in a way that is not repaired by teaching anybody anything. The part of it we find hardest to argue with is that a student can do everything correctly and still be read as having done it badly, because the reading was settled before the writing arrived.
The version of UniFluent we will not build
There is a version of UniFluent that addresses the listener instead of the speaker. We are not going to build it.
That leaves us holding an uncomfortable position, and we would rather write it out than let it sit unstated. If the mechanism is partly in the listener, then part of the problem is located somewhere we do not intend to act. Suppose a student used our product for a year and produced work that met every convention a British marker expects. That student could still be marked down by a lecturer who was never going to read them as capable, and we would have done nothing about the lecturer. We are not planning to. Every claim we make about what our product is for should be read with that sentence attached to it.
The reasons are not principled ones, and we are not going to dress them up as principles. Building for the listener means building for staff instead of students, which is a different product, sold to a different part of the university, with a different buyer and a different evidence base. Demonstrating that it worked would mean demonstrating a change in how people judge other people. That is a much harder thing to measure than whether somebody learned some vocabulary, and we have no idea how we would do it. We are a maths student and an engineering student. Saying we would take that on would be a claim we could not support, and the point of this page is to stop making claims we cannot support.
How much weight the access argument can carry
The cost of the refusal is that our whole justification now rests on one argument, and it had better be a good one. That argument is about access. Our own assumption, and it is an assumption rather than a finding, is that international students learn academic English anyway over the course of a degree, and that the most we can do is shorten how long that takes. We are not conferring anything on anybody.
Michael Young's distinction (2008) is the one we have found useful, though we should be careful about how we hold it. Knowledge of the powerful is whatever an elite happens to possess, and studying it is a sociological exercise. Powerful knowledge is knowledge that lets somebody do things they could not do before, including argue back. Leesa Wheelahan (2010) puts the same case for vocational education: epistemic access is the precondition for social access, and a curriculum that criticises the dominant form without giving students command of it leaves them standing outside it. Both were writing about curriculum, inside an argument among educationists, and neither was writing about software sold to a university. Picking the distinction up and standing behind it would be convenient in a way we should probably resist.
What we take from it is only this: we find the access argument persuasive enough to keep working on. It is a reason to carry on doing what we are doing. It is not a reply to Flores and Rosa, and we should not pretend that it is one.
The evidence we do not have, and the work we are not doing
There is a second cost, and it is worse. We have no evidence that the part we do address works. Nobody has used UniFluent for a hundred hours. There is no efficacy data, and until there is, everything above is a description of what we are trying to do and not a report of what happens. So the honest summary of our position is that we have declined to address one half of the problem and have not yet shown results on the other half. Both of those sentences are true at the same time, and the second one is the one we can actually do something about in the next year.
What can be said alongside it is that it is not a language app's job to fix how markers hear students. We think that is obviously true. A tool a student opens on their phone at eleven at night has no access to the person who will mark their essay in March, and pretending otherwise would be dishonest in a way that is easy to detect. But "not our job" is not the same as "not a problem", and we are aware that the first phrase has been used to avoid the second by every company that has ever sold anything into a university.
There is a body of work that meets the problem closer to where it happens, and we want to describe it carefully, because the careless version of this is self-serving. Mary Lea and Brian Street, in Studies in Higher Education 23(2), 1998, argued that academic literacy is plural and discipline-specific, bound up with identity and authority, rather than one transferable skill that carries across subjects. Ursula Wingate (2015) takes that where it goes: if writing is specific to a discipline, then teaching it belongs inside that discipline, embedded in the modules students are already taking, and aimed at everyone rather than at international students. Most of what follows from that position is staff-facing. It means changing what lecturers do with writing in their own teaching, which is slow work done from inside an institution.
That is not the same thing as undoing racialised listening, and we are not going to claim that it is. What we will say is that the half of the problem we have declined sits nearer to that kind of work than to anything a student-facing app does, and that it is done by people with access to lecturers that we will never have. We are not the people to do it. We are not turning any of that into a reason to buy anything.
What we cannot answer
The question worth putting to anything sold into a university is who it is helping students become. Our honest answer stops early. We can describe what the product puts in front of a student and what it asks them to do. Past that point the answer thins out, and about how the work will be received we have nothing at all, which is a large part of what the question was asking.
Which leaves the case we cannot resolve. If a student works through everything our product asks of them and is still heard in the way Flores and Rosa describe, we do not know what we would say to them, and we are not sure the access argument is much comfort at that point. There is also a version of this where we make things slightly worse. A student whose writing no longer carries the surface features a marker can point at may simply be marked down for reasons the marker never has to name. We have not worked out whether that is a real risk or an imagined one, and we cannot work it out from where we sit, because it would take marking data we have no way to see. We would like to be told.
