About four years ago, Kevin Dunn and Peter Stevenson,co-founders of Trustnode, approached me with a novel idea forselling insurance and financial products. They wanted to talk withme because of my background in user experience design and humancognition and perception.

Their idea was bold. They wanted to replace human sales agentswith interactive 3D avatars who could sell insurance and financialproducts. These avatars would be programmed with deep knowledgeabout these complex domains and be able to answer questions clearlyand patiently. Their interactive agent would never sleep, haveinexhaustible patience, be available to anyone with a computer24/7, and, most importantly, be trustworthy. Plus, they wouldn’tcharge huge commissions on each sale (a bonus no consumer wouldargue with).

I was showed a couple of demos. One was a 3D realistic avatar, aquasi-human looking guy in a suit and tie who was looking at me andtelling me about insurance. The look of the 3D realistic avatar waswell executed, but there was something creepy about him. While heappeared human and, on the surface, “intelligent,” I knew he wasn’thuman, only pretending to be. Also, as I know enough about naturallanguage processing; I knew I could trip him up simply by asking aquestion that wasn’t scripted. I simply didn’t trust the artificialguy.

Then I was showed a second demo. Instead of a quasi-realisticavatar, this was a cartoonish guy repeatedly walking down the samestreet and experiencing “bad things,” such as being struck bylightning, falling into a manhole, and the like. It was funny andengaging. I knew this cartoon wasn’t a person, but he was conveyingresponses I could relate to personally.

There was something fundamentally different about my response tothe second design that led me to believe that cartoons, if properlydesigned and programmed, would be a better path to creating aneffective virtual sales agent. It’s no secret that we lovecartoons—just ask Disney, Pixar and Warner Bros. But what exactlyis it about our brains that accounts for our love of cartoons ingeneral, and how could this innate response to cartoons be turnedinto a powerful interactive sales tool?

Cartoons and caricatures

There is research in education and psychology on theeffectiveness of line drawings, cartoons, and caricatures versushigh-resolution photographs as representations of objects, events,and people. An early study (Ryan and Schwartz of CornellUniversity) directly compared the speed of recognition of fourdifferent modes of representation of common objects:high-resolution photographs; detailed, shaded line drawings;simplified but accurate line drawings; and cartoons.

Participants were shown pictures of various objects (hands withfingers in different positions, knife switches in differentpositions, engines at different views, etc.) in each of the fourmodes. The pictures were projected at brief exposures, which grewlonger and longer until participants could identify the objects.Interestingly, participants needed the shortest time to identifythe cartoon drawings. The line drawings required the longestexposures; there were no significant differences in exposure timesrequired for photos and shaded drawings. This study is significantin the context of using cartoons as interactive agents because itprovides some deeper clues into the way our brains may encodeinformation about the world.

External form to internal schemas

The Ryan and Schwartz experiment indicates cartoons andcaricatures may work so well precisely because they do notaccurately reproduce a scene in precise metrical fashion. They notonly omit, but they transform, distort, and exaggerate theinformation in ways that appear to facilitate recognition.

The efficacy of cartoons may be somewhat less mysterious if wemake some assumptions, based on earlier theories, about the mentalstructures underlying perception: namely, that our brains seek toencode any sensory input by fitting the sensory data to internal“schemas.” A schema can be thought of as a convenient way ofabstracting out the prototypical, distinctive features of objectsor events in a way that conserves, simplifies, and possiblyexaggerates these features without trying to store rich detail.

Caricatures are another example of how exaggeration andsimplification are effective communication devices. (Caricatures doa better job of conserving the set of distinctive features we useto encode real faces.) Think about Richard Nixon as an example. Acartoon of Nixon—with its exaggerated jowls, elongated nose, wavyhairline, and box chin—is easier to recognize than a photo of him.A caricature, although a frozen moment in time, can convey aperson’s characteristic expression better than a photo because ourfacial expression—not the static structural features—is thehallmark of our face.

In other words, the brain interprets scenes, objects and eventsby using idealized, prototypical internal representations—conceptsthat never quite fit the precision of the real-world, but have tobe adjusted to match any specific scene by using transformations.Cartoons likely work as well as they do because, somehow, theirexternal forms (if well designed), come closer to mirroring theinternal structure of our mental models or schemas than dorealistic photos or precise drawings that are more faithful to the“real” thing.

Why we love them

But why do we find (some) cartoons funny, amusing or evenheart-warming? One theory is that we find cartoons funny preciselybecause, in their very stylized, prototypical, exaggerated forms,they make us aware, perhaps unconsciously, of how our minds encodeinformation; that a clever human artist has tapped into thistendency of the human mind to exaggerate, generalize and simplify.We find jokes funny not just because they have surprising orincongruous endings, but rather because jokes make us aware of theflexible nature of our internal classification schemes.

Although not scientifically proven, this concept of“metacognition”—combined with the power of cartoons as triggeringmental schemas—may provide a key to understanding how the cleveruse of cartoons and even visualizations can be crucial designelements in computer interfaces.

A powerful example of this appeared in 1986 with theintroduction of the Macintosh. When users first booted up thedevice, the first thing they saw was a cartoon of the Mac itselfwith a smiling cartoon face. Although simple, the cartoon helpedcreate a powerful bond between the company and the customer. Withit, Apple attempted to convey the metacognitive message: Thisdevice is not about math, it’s nothing to be afraid of, and itmight even be fun to use. And more importantly, the smiley Maccreated a personal relationship between the company and the user,which might have elicited the thought, “Somebody at this companycared enough about me to design this cartoon.”

Putting theory into practice

Creating a good cartoon character—say, tied to an insuranceproduct—can be a powerful virtual sales tool. But to be effectivethe cartoon must follow some perimeters. They must: be easilyrecognized and remembered; trigger innate feelings of trust andlessen the user’s anxiety in confronting uncomfortable issuesrelated to insurance; be built from a finite set of simple elementsthat can be easily mixed and matched to create new characterswithin new situations; and have simple character expressions thatcan be placed under programmatic control to create a range ofinteractive experiences.

In Trustnode’s case, the key design decision was to have acentral character—a virtual interactive guide—that could engage theuser’s attention and trust. Thus, Harvey Keck was born. Harvey,through a sequence of questions, could provide stories andvisualizations while utilizing other characters. For example, acharacter would experience “bad things” and Harvey would explainthe consequences and suggest how the proper insurance plan couldhelp.

It’s a simple idea that works. The proof of its success is inthe results. For example, a subsidiary of a leading internationalinsurance carrier experienced a 200 percent increase in the take-uprate on its direct-to-consumer site (no call center support), plusa 20 percent increase in its call center take-up rate for prospectsinteracting with the Trustnode module before connecting with a callcenter, versus those who go directly to a call center.Additionally, one of California’s largest labor unions witnessed a500 percent increase in take-up rates of a voluntary life andvoluntary disability product offering by a large multi-lineinsurance company.

This provides real-world evidence that harnessing people’sinnate responsiveness to carefully-designed cartoons andvisualizations, combined with interactive knowledge bases, canserve as an engaging and effective sales tool for complex productssuch as insurance plans.

Michael Mills, Trustnode co-founder, has held a variety ofsenior-level positions in the areas of user experience design andhuman interface at companies, including Katango, Intuit, Yahoo!,and Apple Computer. He is a former Associate Professor ofTelecommunications at New York University. He holds a PhD inCommunication and Cognitive Science from McGillUniversity.