This video describes the visual design of the “avatar” for Watson – the Jeopardy-playing AI that recently debuted on the Jeopardy show.
This is a lovely example of generative art. Fun to watch as it swirls and swarms and shimmers. But I do not think it is a masterpiece of avatar design – or even information design in general. The most successful information design, in my opinion, employs natural affordances – the property of expressing the function or true state of an animal or thing. Natural affordances are the product of millions of years of evolution. Graphical user interfaces, no matter how clever and pretty, rarely come close to offering the multimodal stimuli that allow a farmer to read the light of the sky to predict rain, or for a spouse to sense the sincerity of her partner’s words by watching his head motions and changes in gaze.
Watson’s avatar, like many other attempts at visualizing emotion, intent, or states of human communication, uses arbitrary visual effects. They may look cool, but they do not express anything very deep.
…although Ebroodle thinks there is something pretty deep going on with Watson, as in… world domination.
Despite my criticism, I do commend Joshua Davis, the artist who developed the avatar. It is difficult to design non-human visual representations of human expression and communication. But it is a worthy effort, considering the rampant uncanny valley effect that has infected so much virtual human craft for so long, caused by artists (or non-artists) trying to put a literal human face on something artificial.
What Was Watson Thinking?
Watson’s avatar takes the form of a sphere with a swarm of particles that swirl around it. The particles migrate up to the top when Watson is confident about its answer, and to the bottom when it is unsure. Four different colors are used to indicate levels of confidence. Green means very confident. Sounds pretty arbitrary. I’ve never been a fan of color for indicating emotion or states of mind – it is overused, and ultimately arbitrary. Too many other visual affordances are underutilized (such as styles of motion).
Contradiction Between Visual and Audible Realism
Here’s something to ponder: Watson’s avatar is very abstract and arty. But Watsons voice is realistic … and kinda droll. I think Watson’s less-than-perfect speech creates a sonic uncanny valley effect. Does the abstraction of Watsons visual face help this problem, or make it more noticeable?
Is the uncanny valley effect aggravated when there is a discrepancy between visual and audible realism? I can say with more confidence that the same is true when visual realism is not met with complimentary behavioral realism (as I discuss in my book).
Am I saying that Watson should have a realistic human face – to match its voice? Not at all! That would be a disaster. But this doesn’t mean that its maker can craft abstract shapes and motions with reckless abandon. Indeed, the perception of shapes and colors changing over time – accompanied by sound – is the basis for all body language interpretation – it penetrates deep into the ancient communicative energy of planet Earth. Body language is the primary communication channel of humans as well as all animals. Understanding these ancient affordances is a good way to become good at information design.
Hmm – I just got an image of Max Headroom in my mind. Max Headroom had an electronic stutter in his voice as well as in his visual manifestation. Audio and video were complimentary. It would be kinda fun to see Max as Watson’s avatar.
What do you think, my dear Watson?