Cabinet of curiosities #3: Human-first Ai

Just finished Kai-Fu Lee and Chen Qiufan’s little beauty of a book ‘Ai 2041’. One word: Wow!?

It’s an easy ride of a read - full of thought provoking use cases and analysis… Some already in market and some close by…?

And while this is a book about what Ai-enhanced life might look like in 2041, I was drawn to a line about how in the future, when we interact with Ai ‘all you’ll have to do is ask it (the Ai) the right questions’.?

But just like change, asking the right questions can be hard. Even after a few thousand years of practice, us humans still don’t always know what the right questions to ask are. Added to this we continue to exhibit biases towards asking questions that are of interest only to us and not to whoever is using what we make - Ai-supported or otherwise. The result is we can lose out and never realise the opportunities shiny (and as we will see, sometimes gorgeously fluffy) new technology gives us.?

In healthcare we see this problem every day. One of the more 'fun' examples is in the ‘companion toys’ market.^ Companion toys are synthetic robot toy animals used to comfort often elderly patients with dementia and no easy access to round the clock care. According to researchers, during Covid because of lockdown-driven lack of human contact the demand for these types of toys has increased.

Other uses for these toys include giving them to kids so they learn how to empathise more and take responsibility for looking after someone other than themselves. And possibly to distract them so they don't bother their frazzled, over-worked parents quite so much!

Folks like Sony with its robo-pup (it's called Aibo and it's Big In Japan*) and Hasbro/ Ageless Innovation (with its creepily named Joy For All) range are the main movers and shakers in what's known as the 'connected toys market'.

And play pays. In 2019 the connected toys market was valued in at US$ 5.8 billion and with a projected CAGR of 16% for 2020-2025. Other companies like Namco Bandai, Mattel, WowWee and LEGO are also in, especially as this market uniquely combines and hybridises disciplines like product design, computer games, robotics, VR/ AR/ MR, data and analytics, conversational UI and yes, you guessed it, Ai.

Pups like Aibo, or Joy For All's Freckled or Golden Pup use an extraordinary array of technology. Even Furby - the fluffiest of companions - is packed with technology. This article in CNET takes these darling fluff balls...

... And applies a less-than-surgical-more-Texas-Chainsaw-Massacre approach to show us what's inside!

Similarly, the most recent version of the Aibo pup uses a sophisticated combination of OLED displays, cameras, motions and light sensors, LTEs, microphones to capture data. It has sets of inertial measurement units and a mix of 1- and 2-axis actuators to help it do its stuff - move its neck, mouth, legs, ears, and of course, wag its cute little tail!

Data Aibo records is uploaded via wifi (yes, Aibo has that too) to Sony's AWS powered cloud. Data is fed into Sony's Ai Engine and deep-learning algorithms to help Aibo recognise images and sounds, and to enable the development of a unique pup personality over time. If you thought Alexa was a powerful personal data sucker-upper, think again!

This technology used moves beyond verticals - it is trans-sectoral and diversifiable, so if you're in games, education, entertainment and healthcare, you can get more bang for your buck. Like Lee and Qiufa, these companies can see how the future might turn out and want to have a hand in shaping it to their advantage.

Coming back to health and companion toy animals, the news is a little disappointing. While companies that make these toys claim their products relieve stress, depression, anxiety, and loneliness, and help keep people more physically active this is true in a limited number of cases. And it's true because in designing those pups, kittens and turtles, the 'right questions' have not always been asked.

Research from a mix of universities shows how the developers of less acceptable, less liked toys failed to deliver acceptability and use value because they didn’t take into account how users wanted their pup/ kitten/ hedgehog/ seal/ turtle/ Furby thingy to be.?They were less emapthy-inducing and at odds with users' mental models of

As all decent design folk will tell you, the sine qua non to human beings accepting, interacting with and trusting a new product, service or technology rests on initially asking questions designed to understand what at a human level, the problem is.?

Questions such as: ‘What human need is not being met? Why not? What are the major fail (or pain) points in the current situation? Where is the opportunity to make a difference? How might we do this??

Then once we’ve got some answers to the above and to ensure that whatever we make is going to work for the people using it we ask things like: Is the (thing we’ve made) useful? Does it address the unmet need and solve the problem the human user has? Can they access it easily? Is it a more desirable option than whatever else is around??

If the answer to these questions is ‘Yes’ then it’s worth making more of an investment. If it’s ‘No’ we bin it. If it’s a ‘Maybe’ we iterate or pivot until it’s a ‘Yes’.?

The developers of the less popular ‘companion toys’ did not ask these simple but critical questions. They did not spend enough time speaking to users and getting to know their side of things. Or if they, did they did not heed what the heard. The ‘robo-pets’ they made did not acknowledge users through eye contact, did not have soft fur, look familiar, or give personalised responses.?

Instead, many of their pets had hard shells, were less life-like and familiar-looking, had limited personalised interaction and eschewed speech and noise-making. Their pets invoked little connection or empathy, were misaligned with users' needs and mental models of what that particular animal looks like or how it behaves.

Thus, many of these developers failed and patients, carers and health staff, friends and family, and those making the investment lost out. Users did not want to buy or know more about what they offered and this all costs everybody except perhaps other companion toy makers.?

Like some of the unsuccessful companion toys there are hundreds of other products and services that have also failed because whoever made them did the exact same thing. They didn’t recognise their own biases and didn’t ask the ‘right questions’ - didn’t ask simple questions that start with human needs and wants.?

This lesson is useful if we want to advance towards the types of technological sophistication illustrated by Lee and Quifan. IMHO taking a ‘human first’ approach from the get go and embedding it all the way though a project is the only way we will create products and services that deliver to an unmet human need and will be acceptable to human users.?

Being human-first in our thinking provides the ‘golden thread’ that links human users to new technology, including Ai. It’s the only way we can create products and services that users will want, trust, accept and use and which, in time, step by step will enables us to get to the vision and value Lee and Quifan suggest is possible in the next 20 years.?

And from where I'm standing, looking at the enormous pressure our healthcare systems are under, anything we can do now of in the future to alleviate that gets a big 'Woof!' from me.

----

^I use the word 'fun' but qualify it. Really isn't it a bit sad that we're replacing real human contact with plastic, glass, metal and nylon made bots? I guess it's a case of 'needs must when the devil drives' - i.e. better than nothing. It's also less of a thorny issue than say, a data migration mess up for EMRs where millions of people - patients and doctors alike - find they are looking at the the wrong health records. Definitely a far from fun scenario.

*Dear Reader, my apologies for this. I cannot resist a compulsion to make a crass joke when it presents itself. For more excruciation see https://www.youtube.com/watch?v=DdWZKb659K0