Vigloless Testing

We just reached our usage limit with ChatGPT for the time being, so I can't continue doing tests within Viglo directly. However, that doesn't mean I should stop testing Viglo. So to work around this problem, my girlfriend (a native Spanish speaker) and I have been doing calls in Whatsapp, and using ChatGPT in a web browser in order to simulate the Viglo call methodology of identifying problems and then creating climbs to work on them.

First, doing this workaround was concerning because if we can create the video call + ChatGPT experience like this, then is there a point to creating Viglo? Even though I was reminding myself that we will be giving the students specific and detailed progress reports, this thought bothered me immensely. That is, until we did more and more test calls. My girlfriend and I realized that though we have the tools, we do not know how to use them. So eventually, Viglo will be delivering the tools with trained Guides who will get the most out of the students.

The first two times we tried it together, she wasn't able to decide which of my Spanish errors deserved a climb. So after the call, she created her own plan for how Guides should conduct themselves, and we tested it out.

She had some prepared topics lined up and we began chatting about them. After about 10 minutes, she picked 3 words I had trouble with and we created a climb with them. The words, in English, were "to exercise," "to show off," and "belly." These are not directly related words, and so the climb that ChatGPT created was just ok. The questions in the climb were good, but a bit scattered. (Just writing this sentence makes me wonder if Viglo's ChatGPT prompts should specify what you're trying to accomplish. For example, we created the climb for the three words, but we didn't tell ChatGPT if I was trying to memorize the words, practice conjugating them, improve my pronunciation, or whatever. So I need to take that into consideration when working on the prompts).

The other thing I am going to change is reducing each climb from 2 exercises to 1 for the following reasons:
1.) Usually when I get to the second climb, I don't feel like doing it. I think it's a psychological phenomena where after working to grasp the new subject, my mind realizes that it's only halfway done. I want the student to feel accomplished, not just tired, after completing a climb.
2.) If one climb isn't enough, the Guide can quickly have ChatGPT created another, so too few exercises is easily fixed. Furthermore if the first climb had exercises that weren't great, the Guide and student can finish the exercise sooner, or ask for a better exercise. (for example, maybe the first climb was a translation exercise, but the student would prefer a multiple choice)
3.) Students can complete more climbs per call. And when we present the students accomplishments, the student will have a greater sense of pride in having finished more climbs. This could lead to increased chances of the student wanting to continue using Viglo.

Ok, I got slightly off topic, as I never concluded how my girlfriend's more structured approach went.

The conversation flowed more smoothly and we had more fun with the climb than in previous chats. But we need to reduce the amount of time it takes to come up with a climb. It took her about 10 minutes, but I want to reduce it to about 5 minutes (and likely try to reduce it further later on). People can use any video chat service to talk, but only Viglo has the ability to generate climbs to fix their specific mistakes; and I want to get the student to what makes Viglo special as soon as possible. So we'll need to come up with a system to rapidly identify mistakes that deserve a climb.

My girlfriend also suggested that we prepare prompts that include various difficulty levels. I love this idea for two reasons:
1.) Obviously you can give the student a climb that's more appropriate to his or her Spanish level.
2.) When learning a new language, practicing new skills only once won't be enough to retain the information. So the Viglo students will need to be repeating climbs every now and then. But review sessions don't create the same feeling of advancement like doing new climbs. However, if we can present that a climb has been completed at different levels, (completed the climb at the beginner level, then the intermediate level the next day, and eventually you will strive to complete the climb at a difficult level) then I believe the student will finish the call with a similar, if not the same, sense of accomplishment as if they completed a new climb. (About the same sense of accomplishment can be felt when a student completes 1 climb at 3 different levels, as when the student completed 3 separate climbs)

We have been identifying the mistakes I make while speaking, we'll need to figure out how to identify problems with the student's listening comprehension. And that's way more complicated. When the student speaks, their struggles are on display. But with listening comprehension, what can we do to pin point the problem? And what kind of climbs can be made to solve them? I tried asking ChatGPT for some suggestions on how to best do it, but I wasn't getting any useful responses. I'll continue doing more chats with my girlfriend to figure out a good system for identifying spoken mistakes made by the student before moving on to listening comprehension mistakes.