DRAWING WITH WORDS...LITERALLY

What’s in this post: The Inner Workbench; the inner eye; the Google AI Studio Challenge; Wordsketcher; and a few hands-on activities for the classroom.

Many powerful ideas in language learning get a bit lost in the noise of the ‘C’ skills, including the recent addition of Critical AI Literacy (so, 8 Cs so far?). These ideas are often more domain-specific, which makes them less likely to catch the attention of educators and edtech folks ‘reimagining’ improvements to learning at large. Yet, in the end, they might matter far more for language learners specifically.

One such idea is a more active engagement of mental processes, or the Inner Workbench. Adrian Underhill, who introduced it, describes it as the ability to tune in to our inner ear and inner voice as a form of practice and rehearsal.

‘Inner ear 🡢 inner voice 🡢 mime it 🡢 whisper 🡢 say aloud …and now REVERSE’

Such deliberate and purposeful practice helps learners internalize language patterns before they speak them out loud. After all, is there any better way to make language stick than by practising it where it starts?

In today’s post, I’m focusing on one piece of the Inner Workbench and sharing a small project I built around it.

Wordsketcher: Where Words Meet Imagination (Sometimes)

Wordsketcher is a small project I put together for the Google AI Studio Multimodal Challenge run by the Dev Team. I’ve been experimenting quite a bit with Google AI Studio, and had an idea for a practical activity where its multimodal powers could really add value.

It builds on one element of the Workbench -the mind’s eye, or inner eye, and on a practical activity I first shared on the blog about 10 years ago (time flies!). Now it comes with a digital twist that, I believe, makes it even more effective.

The activity itself is beautifully simple. See the Games of Imagination for detailed steps. It’s all about ‘drawing’ with words (or flashcards), imagining the picture, and then taking a language ‘snapshot’ of it.

It’s easy to set up: give learners slips of paper, cards, or post-it notes, along with a set of vocabulary or a topic, and ask them to create their own word sketches. No actual drawing skills are required.

The procedure depends on your learners and their proficiency level:

Look at the words 🡢 imagine a picture 🡢 place the words 🡢 *say aloud / describe

By including tech, in particular a multimodal model, we can take this activity further, going beyond ‘engage the learner’ by digitalization (or AI-zation – has anyone used this word yet?) of a tried and tested classroom activity.

The Wordsketcher offers 3 modes:

Challenge: a specific scene to create (e.g., ‘a cozy house under a smiling sun’) with a predefined set of words;
Topic: a theme (e.g., ‘At the Beach’) with a bank of related words for users to build a scene. For both Modes, there’s an option to add learner’s own words; and
Freeform: a blank canvas to imagine whatever comes to mind.

Now the features where tech adds value:

Text-to-speech

Text-to-speech lets learners hear all the words in the Word Bank. Words they add themselves are instantly added and voiced.

What’s the Word?

This works like a reverse dictionary: learners describe the word they have in mind (e.g., ‘a big house where kings live’), and the model suggests the match.

Image generation

When learners click ‘Sketch it!’, the model generates an image based on the words on the canvas, their placement, and the selected style.

A quick note on styles: I deliberately skipped cinematic and the like styles, choosing only those that are recreatable. By recreatable, I mean styles that seem easy to copy, that is, a learner with basic drawing skills could produce something similar.

Or at least imagine, like the plush toy and paper cut-out styles (they could be a hit with younger learners).

This keeps learners focused on the language and their word sketch, avoiding the distracting (unnecessary) ‘wow’ factor, and makes it more likely that the generated image matches what they’ve imagined.

The image appears instantly, so learners can compare their mental picture on the canvas with the generated one (‘spot the difference’).

It also sets up a reverse activity: they can use the generated image to make a word sketch and then see how the two match up.

So, the updated workflow might look like this:

Look at/listen to/write the words 🡢 imagine a picture 🡢 place the words 🡢 *say aloud / compare / describe…REVERSE

I hope this project sparks some ideas! Think of it as a starting point to explore the Inner Workbench and experimenting with what multimodal models can offer.

If you’d like to try Wordsketcher, I’ve made the source code available in Google AI Studio (follow this link). Just log in with your Google account. It’s free to use.

Which ELT idea do you find powerful and practical, and think should be more widely shared?