The 200-Word Midnight Message: Machine Translation in a Partner-Class Exchange
Mr. Ibarra's Spanish III class of 27 exchanges asynchronous messages with a school in Querétaro. Three weeks in, he notices a pattern in some of his students' posts: 200-word messages, grammatically flawless, using subjunctive forms and vocabulary never taught in the course, mixing formal "usted" with friendly "tú," and appearing about 45 minutes after the prompt, at 11 p.m. Meanwhile, the Mexican teacher reports that her students are losing motivation because the replies read like essays and never ask questions back. One of Mr. Ibarra's strongest students is a heritage speaker who writes quickly and accurately. He must decide what to say to the class tomorrow and how to grade the exchange.
1. What evidence would justify treating the messages as machine-translated rather than legitimate, and how should he verify it?
Read the worked answer
- Probabilistic indicators, not proof: flawless grammar including untaught subjunctive, lexicon far beyond the course, register mixing (usted and tú in one message), latency implausible for the length, and no questions back - messages that answer but never interact.
- The heritage-speaker confound is real. Speed and accuracy alone prove nothing; check whether the content is personal, consistent across weeks, and matched to what the student produces in class.
- Low-cost verification: a five-minute in-class write on the same topic, and a 60-second oral follow-up ("¿Qué te contó tu compañero?"). The divergence between the two performances is the evidence.
- Process matters: gather the evidence first, apply the policy privately, and never accuse a student publicly. A false accusation here would land on the heritage speaker.
2. Is using a translation tool a proficiency problem in this task?
Read the worked answer
- It depends on the mode's purpose. The interpersonal mode is defined by spontaneous, unrehearsed interaction. If a tool produces the message, the product is not evidence of interpersonal ability - so it cannot be scored as if it were.
- But tools are part of real communication. The answer is not a blanket ban; it is a policy differentiated by mode: tools permitted for presentational drafts with disclosure, not for interpersonal exchanges whose whole point is developing spontaneous language.
- Grade the observable performance - in-class oral debrief, timed short exchanges - and treat the partner exchange as formative practice for the students who need it.
- Communicate the reason, not just the rule: the partner students are real interlocutors, and essay-length monologues are failing the communicative purpose as well as the assessment purpose.
3. Redesign the exchange so that interaction is observable and the partner class stays engaged.
Read the worked answer
- Cap message length at 40–60 words and require exactly one question per message. Length is the enemy of interaction here.
- Impose a 24-hour reply window and topic prompts grounded in personal experience that cannot be outsourced: my neighborhood, my commute, a celebration in my family.
- Add a 30–60 second voice note or video every other week. It is harder to fake, it matches real interpersonal interaction, and it gives the teacher audible evidence.
- Add an in-class debrief where each student summarizes two things learned about their partner and one thing they would ask next. This makes both comprehension and interaction gradeable.
- Renegotiate the format with the Mexican teacher so her students get genuine interlocutors; a demotivated partner class is a design failure, not a discipline problem.