Wednesday, 19 January 2000

éTui - robotic toys for learning

I do a lot of work with BAFTA, from judging movies to helping embed interactivity in the major genre like film and games. Back in 2000 I ran an event there to help BAFTA members come to terms with what was happening, potentially, to toys as they became digital too. The event centred around a project that began with EU's i3 funding support and developed some remarkable little toys. Andy Simpson was the project manager. But the best thing about éTui (the toy's name) was the methodology that developed it - by children for children. No surprise then that the process worked and produced some remarkable little toys.

I think, looking back, that I am quite disappointed with how poor toys are today. If we look at how much of our lives has moved forward, and by such huge distances, the fact that toys haven't is puzzling. Of course there are exceptions... but children playing in the Digital Age can't assemble code the way they could assemble Meccano in the Engineering Age, can they? Nor are their toys as smart as they might be...

Anyway, éTui pointed very clearly to a way forward; as with all these things in 10 years or so someone, hopefully, will wake up and grab this progress as their own!. Till then, you'll have to wait.

Scattered along the corridors and filling the cupboards at Ultralab can be found the UK's National Archive of Educational Computing, custodian of the shortest history that any museum might find itself cataloguing. In a span of less than 30 years the clunky old wooden computer of the hobbyist builder had given way, via Clive Sinclair's ubiquitous ZX computers, to today's creative industry workhorses. But along the way whole cupboards full of missed opportunities and bright, but undeveloped ideas, have accumulated in the National Archive.

Amongst the monster modems and disc drives the size of Battersea Power Station lie one such set of ideas in the shape of a family of programmable toys: lumbering trucks, little barrel shaped robots, inverted perspex wheeled bowls, caterpillar tracked buggies and more. Children visiting the lab today are still delighted by these "old" toys, love to control them, to programme them and to set them tougher and tougher tasks each to be surmounted in turn. The toys engage children intellectually in a unique way. Yet shops today are almost totally devoid of these seductive engaging little pseudo-sentient programmable robots. Their passing is regrettable but it was the Furby a couple of Christmases ago that triggered a wish at the 'lab to revisit the programmable toy genre. The Furby pretended to learn from children but actually didn't and legend has it that whilst children were disappointed only the nation's secret services were daft enough to be fooled into thinking that the Furby was actually listening to anyone as they banned these simple toys from their offices. Children wanted toys that could really learn and somehow their disappointment in current electronic toys and the continued magnetism of the old programmable toys in the archive suggested a clear direction for a research and development project. As a result Ultralab, together with Catalan researchers at Barcelona's Universitat Pompeu Fabra and Apple Computer, began work on the éTui Project.

But éTui was not just about designing a great new toy (as if that wasn't ambitious enough). The project wanted to use four to eight year olds across Europe to design a toy whose real purpose was to help children learn about learning; the holy grail of education is this meta-level learning, but could a toy lead children as young as four to it? European funding was duly offered (from the ESPRIT i3 blue sky research pot) and a methodology was established which offers an effective blueprint for toy development everywhere: the children designed the toy. We established an on-line community between the families and schools in Norway's Trondheim, Spain's Deia and (rather less exotically) Essex's Brentwood. Children swapped thoughts, drawing and ideas with a heavy reliance on digital video (remember that at four nobody reads much and the toys were going to need to be programmed without keyboards or notation of any sort). Initially our young developers told us about their favourite toys and we showed them our favourites from the Archive. Once a dialogue was established we began to feed them with ideas in either 3D VRML or as a screen based simulation of possible behaviours. They responded to these ideas and we re-coded so that they could see immediate feedback from their "research". Through the VRML worlds we offered everything from scary spiders to shapeless amorphous blobs and we collected feedback about look, feel, colour, texture and behaviour. Slowly but very surely we moved in a whole series of iterative steps towards the beginning of a definition of this "ideal" toy. Much of this VRML work was done by Josep Blat's team in Spain and it is worth noting the clear differences between children's play culture across our research population; in Norway, with metres of snow covering most of the ground much of the time, indoor activity dominates, although there is a great national obsession with marbles once a year as patches of ground begin to be exposed where snows melts in late spring. In Catalan Spain life is very much more outdoor with children sheltering throughout the lunchtime siesta from the sun. In Brentwood children had their lives filled much earlier and more formally by a sense of curriculum where play is perhaps getting squashed by the pressure to achieve in early testing. All our schools and families were delightful and different and a successful toy would need to fit into the rather different lives, cultures and language of children in each location. It also needed to fulfil their ambitions and once engaged in the design process the children were enormously, but never unrealistically, ambitious for how good "their" toy might be.

How big should éTui be? In many languages the word éTui means something small, a container, to be carried in a pocket and that had been our initial guess about scale, but the children steered us to something larger. Memorably one of them commented that she'd really like to be able to ride the toy home and we found the design heading in a series of iterative lurches towards something in the size range of many of those programmable toys from the archive, about the size of a dinner plate. The simulation software that Ultranaut Kris Popat masterminded in the 'lab was cute and engaging in its own right (and may well spin-off into another project), but it gave us a very rapid way to test ideas and to feedback to the children how their thoughts might work out in practice. Our young researchers had no problems with making the abstract step between screen toy and real toy; these were after all the youngest of what is a very wired generation.

It was clear from the simulation software that the ability to programme our toy in a tactile way (for example dragging it in a remembered pattern) was desirable. It was also clear that our young researchers were chock full of bright, media literate ideas. When we asked adults the difficult question "How do we know what the toy is thinking?" they would suggest LEDs, or screens or various technological solutions, but the children were clear: "You need music" they said, "because when we are watching TV if something is worried and unsure they play worried-and-unsure music and then we know what they are thinking". So Weiya Wang, Ultralab's electronics genius gave éTui sounds and added the ability to be taught tunes by flashing a torch at it. The children also fed back their clear views for the way an éTui might move (it now has a jerky insect-like motion with a bird-like unpredictability). When we asked if it should have eyes again our young researchers were clear: "if it has eyes we'll think it can see like us, and it won't be able to, will it?" so éTui has strange alien sensors that allow it to explore its environment, seek out dark corners (to hide in when startled), do line following and learn tunes.

So have the children defined something useful? The final éTui prototype has now had a number of outings to exhibitions and shows (including a weekend in Paris - we sent two together). Children are absolutely, completely, dottily captivated by it. For hours. A toy designed by children for children has proved to be, perhaps unsurprisingly, just what children wanted. But there have been some other gains too. Trying to teach a little éTui that is keen, but a bit dim (rather like doing sheepdog trials with a Dalmatian), really does help children to reflect on learning in general and on their own learning in particular too; we have observed those elusive meta-level learning gains. Perhaps most surprisingly, adults seem captivated too, suggesting that we have neglected play for adults for rather too long.

Ultralab are now seeking manufacturing partners for this engaging little toy and its interchangeable wardrobe of stretch-on clothes. Meanwhile in that dusty cupboard in the National Archive of Educational Computing the clunky old programmable toys from an earlier era are getting quite excited as the wheel of fashion starts to turn their way again.

© Stephen Heppell 2000