Arvin Raj Mathur
← All writing
Lingways · Design notes

Conversation beyond the script

A Lingways design note on evaluating practice in unfamiliar conversations.

Lingways is in development. This is a proposed approach, not a report of measured results.

A learner practices asking for directions. Then the other person replies with a question: “Are you walking or taking the bus?” The learner now has to make a choice, understand a follow-up, and respond in a way that keeps the exchange moving.

That is a useful starting point for how I want to think about Lingways. Finishing an exercise is one observation. Handling a conversation that changes direction is another. I want the product’s design to make room for both.

Make the goal observable

“Become conversational” is a meaningful ambition, but it is too broad to guide a single practice task. A more concrete goal would be: ask for a route, understand one clarification, and confirm the next step.

For that task, I would want to record whether the learner communicates the intended meaning, what support they need, and whether they can repair a misunderstanding. Speed and grammatical accuracy might be useful observations, but neither would be the entire definition of success.

Change one part of the situation

A possible evaluation would begin with two comparable groups receiving the same amount of practice. Both would learn similar language for directions. Later, each learner would face an unfamiliar exchange containing a small change: a closed street, a different destination, or a request to clarify the mode of transport.

The task would need careful design. If the new version introduces unfamiliar vocabulary as well as a new conversational demand, a poor result would be difficult to interpret. I would start by matching vocabulary difficulty and changing one conversational requirement at a time.

Look at the help, too

A correct response after a complete translation tells us something different from a response after a small hint. A practice system should preserve that distinction in its observations.

One design I want to explore is a support ladder: first allow an independent response, then offer a short prompt, then a more explicit example. The point would be to understand what help unlocks the exchange, while still giving the learner a way forward.

Return to the question later

I would repeat a comparable task after a delay, using different details. That would help us ask whether the learner can use the language beyond the immediate practice session. It would not, by itself, establish that one method works for every learner or every setting.

Lingways is in development. These are proposed design and evaluation choices, not completed experiments. The question I want to keep asking is simple: what evidence would show that a learner is becoming more able to communicate with another person?