AI at KwizIQ

KwizIQ has long taken pride in its use of AI. Even as early as 2015, way before I joined, and before large language models like ChatGPT were a thing, KwizIQ was already describing itself as “the A.I. Language Coach” and KwizBot as “an artificial intelligence”.  But AI has changed a lot since then, and this new generative AI, capable of generating text, images and sounds, has become a very powerful tool that can impact our lives in many different ways, some better than others.

What I see in KwizIQ today is a KwizIQ with the same optimistic view and a commitment to adopt AI as a tool for the benefit of our users and our team. And this post is an explanation of how this commitment is not only possible, but a reality.

Better features

We’ve just released our first generative AI feature at KwizIQ: comprehension exercises. It’s difficult to express how exciting it was for me to see this feature become a reality. Because I really wish it had existed when I was learning English.

I don’t know if you’ve seen these before, but comprehension exercises are commonly presented as a written text, followed by some rudimentary single-choice questions that are fairly unsatisfactory. For a start, I could frequently cheat my way around them just by scanning the text for literal matches of the choices. And when I got them wrong, the explanations wouldn’t go further than “the correct answer is A”. These unsatisfactory single-choice questions were the only way you could really automate grading and feedback back in the day.

Think for a moment, how would you (as a human) check if a student understood the text? The answer is simple, you’d ask them a question and expect them to put the answer in their own words. We can’t really rephrase something if we didn’t understand it. And when we rephrase it, it’s easy to identify the nuances of what was clear to us and what was fuzzy.

Thanks to AI, this is exactly what we do now. After each KwizIQ reader, you can answer a set of questions in your own words. And we use a large language model to provide rich and nuanced feedback that goes much further than a single-choice question. AI is enabling us to teach languages at scale in ways that are substantially superior to what was possible before.

Better content

But before building features with AI, we’ve been using AI to generate content for a while.

You probably noticed when our language team started generating images and music with AI. This allowed our teachers to get creative and exercise artistic skills beyond what they did in the past. As an example, all of our newsletters now contain multiple dedicated illustrations, something that we simply couldn’t afford before. Without them, newsletter entries like this one about Don Quijote would look dull and boring.

Another example is how we occasionally release themed albums to practise listening skills. Although KwizIQ has some talented musicians on the team, there is no way we could have recorded an album for Valentine’s Day.

And the same thing can be said when we developed the Minigames through a Hackathon, the software development equivalent of a jam session, where we used AI coding agents to help us create them.

All of this paints a different picture than the one I see in other companies. We don’t want our team to shrink, we want it to grow. Rather than using AI to replace humans, KwizIQ is using AI to expand what’s possible.

A better team

The above is just the visible part. Our teachers aren’t just using AI to get creative outside of their expertise, they are using AI to generate quizzes, writing exercises, readers, etc. The challenge here is to ensure the quality of the end result. This is the reason why AI-slop is on the rise all over the internet. At KwizIQ, however, our language team is encouraged to use AI wherever they think it can help. And the same strict human review processes we had before are still in place today.

As a consequence of all of the above, our team is progressively steering away from repetitive boring tasks and evolving into more creative and interesting work.

This adoption doesn’t happen overnight. Getting professional-grade results out of generative AI is hard. We make sure our team has the room to learn through individual experimentation, workshops, and hackathons. We see this new kind of digital literacy as an opportunity to support them, and not an extra problem to push to them.

It’s in KwizIQ’s DNA

Over more than 16 years in my career, I’ve worked at different companies. All had things I liked and things I didn’t. KwizIQ has them too. And KwizIQ’s attitude about adopting AI is one of the things I’m proudest of here. Because such powerful tools come with a big responsibility. And we work hard to live up to that responsibility.