Creating the original avatar for IBM's Jeopardy! Juggernaut
Watson was IBM's experimental cognitive computing system, built to compete on Jeopardy! against the show's two greatest champions. To turn a room of refrigerator-sized server racks into a televised contestant, the project needed a face — something abstract enough to avoid the uncanny valley, but expressive enough that viewers could read what the machine was "thinking" in real time.
My title was Senior Art Director, but the work was closer to creative technology. I pitched Joshua Davis as the right collaborator for the avatar, worked inside his generative code to refine its behavior, designed the on-screen answer panel with IBM's research team, and ran installation and testing on-site at IBM's TJ Watson Research Center. I also oversaw a documentary series on Watson's development.
The avatar took the shape of a globe — a nod to IBM's Smarter Planet identity — overlaid with 42 dynamic threads, a Hitchhiker's Guide reference that gave the visualization a wink of personality. A swarm of particles raced across the surface, led by a single bright particle, suggesting thoughts streaming through Watson's processors.
Joshua Davis built the generative system; I worked inside his code to map 36 distinct states to Watson's confidence levels and processing stages, tune color and motion to read clearly on broadcast, and integrate live data from the DeepQA architecture so the avatar moved in sync with Watson's actual reasoning.
Working with IBM Research and lead scientist David Ferrucci, I designed the on-screen answer panel: Watson's top three candidate answers, each with a confidence percentage and a buzz threshold indicator. If confidence cleared the threshold, Watson rang in.
The panel did more than display answers — it made Watson's reasoning legible. Viewers could see the system narrow thousands of possibilities to three, and watch the moment a high-confidence guess turned out to be wrong. That transparency was what turned Watson from a black box into a character.
The televised matches aired February 14–16, 2011. Watson defeated champions Ken Jennings and Brad Rutter and won the $1 million first-place prize.
Viewers reported emotional responses to a globe of particles — rooting for it, feeling sorry for it when a high-confidence guess whiffed. That was the goal. By making the machine's internal state visible (confidence, hesitation, certainty), we made Watson feel like a participant instead of a calculator.
The visual language we developed in 2011 — the globe, the threads, the particle motion — still echoes in how IBM represents Watson today.