1953 was a good year for science and technology. The polio vaccine was developed, the first open heart surgery performed, and color TVs hit the store shelves. It was also the year Watson and Crick revealed one of the most important discoveries of the 20th century: the double-helix structure of DNA.
Sixty years later, amidst technological advances that have spurred massive innovation and changed the face of science, we decided it was time to re-inspire today’s generation. James Musick, a digital communications lead at Genentech, came to us for help in creating an immersive iPad experience that would celebrate Watson and Crick’s seminal discovery, as well as engage the science-curious. But with three months to complete the project, we needed to hit the ground running.
Ralph and The Cheesy Mess
The idea of using a game or puzzle as a way to bring users into a world filled with genetics became an early favorite in the ideation stage. Puzzle-based gameplay was a great analogy to reflect the scientific process: from trial-and-error learning stages, to information synthesis, all the way through solving the puzzle and answering the question. A game hinging on a real-life genetic mechanic would be novel and stood to introduce the science to a new audience. If designed well, with engaging play and a compelling story, we would be able to integrate the science, narrative, and gameplay into a beautiful app.
Screenshot from Osmos, a popular science game steeped in physics, mechanics, and strong narrative.
We started off diving deep into the world of video games. We indulged in playing some of our tried-and-true favorites, as well as newer games such as Osmos and Eufloria to understand how they bridged science and storytelling. At the same time, we were learning about cutting-edge research in genetics through conversations with Jason Long and other Genentech scientists, listening to podcasts, and a hefty amount of reading.
Throughout this inspiration phase, we kept coming back to a wildly intriguing area of research called epigenetics, which focuses on how environments affect genetic expression. In short, epigenetics is the area of genetics that, given a fixed DNA makeup, tells the DNA which parts of the code should express themselves and how they should express themselves as determined by external factors like weather, stress, diet, etc. So while we may have a fixed genetic makeup, the environment still plays a large modulating factor in our genetic expression.
Our fascination with the content—combined with a teamwide love of comic books and mutant-origin stories—evolved into an epigenetic-inspired noir-puzzle game based on Ralph, an artisanal cheese-maker, and his cheese company Ralph's Killer Muenster.
The mechanisms of genetics
Early on in the ideation phase, Evan Shapiro, who would become a core team member and help prototype game logic, introduced us to phylogenetic trees and maximum parsimony theory. You’ve probably encountered phylogenetic trees in biology class, where they serve to illustrate evolutionary relationships among species. Maximum parsimony, on the other hand, is more obscure. It serves as a computational forensic tool to help us understand the evolution of genetic traits along the phylogenetic tree, as well as help us infer species evolution. With maximum parsimony, the “preferred phylogenetic tree is the tree that supposes the least evolutionary change to explain observed data.” After mulling that over for a bit, we decided creating and solving a phylogenetic puzzle by employing maximum parsimony as a scoring criteria was a great jumping-off point and a compelling challenge.
An example of a phylogenetic tree documenting the evolution of a species.
To test this, our first prototype was a paper prototype where we started with a sample origin genetic trait. We sketched out several generations by mutating one trait at a time, and created our own phylogenetic tree. Given the final generation of traits and no tree structure, the goal was to see if we could work backwards to recreate the tree structure and uncover the origin trait. Basically, reverse evolution. After a lot of finessing we came up with a rule set that actually reflected natural evolutionary rules and allowed us to solve the tree and find the origin species.
Our first gameplay prototype puzzle that kept us busy for the weekend.
With this excitement, we spent a weekend programming a single prototype puzzle using mechanics that would become the main gameplay: players reverse engineer evolution by mutating a given set of species traits backwards to a common ancestor in the fewest moves possible. This was maximum parsimony made tangible. In order for the puzzles to be solvable, we had to use real evolutionary rules describing how the species’ traits can change. By the end of the weekend, we had puzzles and a game mechanic that were keeping us up late into the night trying to one-up each other’s scores, which, to us, meant we had hit the nail on the head.
From a design perspective, one of the significant balancing acts of the project involved weighing the desire to impart specific scientific knowledge against the value of experiential learning. When we used maximum parsimony it became clear that this was a perfect mechanic to give people an experiential interaction with real genetics.
Prototyping and puzzle design
Our daily workflow started with an early morning meeting so we could all share progress from the previous day. We spent the rest of the day implementing and testing new visuals and interactions on paper and integrating them into the Qt prototype architecture. Lastly, we tested them with people around IDEO to see how they reacted to specific design and gameplay decisions.
An early interactive visual prototype to test user comprehension of traits.
We used Unity for the production app in the development environment for many of the same reasons that we chose Qt. But Unity has the added benefit of working with native OpenGL, along with the capability to compile to iOS. As time-constrained as this project was, we still wanted to impart as much narrative and visual interest as possible, and Unity proved an excellent match for this challenge.
Nivi Ramesh, our interaction designer and art director, worked to create an amazing game landscape with analogies to microscopy and slide specimens. Working closely with developers at Substantial, we created layered and animated textures paired with OpenGL shaders to achieve the tone we wanted. We had parallel work streams throughout the prototyping phase, with one team working on developing the logic and overarching architecture of the production app while the prototyping team refined interaction elements, game mechanics, and informational displays in the prototype testbed.
What makes a good puzzle?
Before diving any further, we took a step back to make sure we got the puzzles right. Good puzzle games tend to adhere to some basic principles: They are purely cognitive challenges with a few simple rules, nearly infinite outcomes, difficulty that increases with mastery, and some clear goals. Plus, they're easy to learn.
Another prototype testing menu layouts and trait visuals with early signs of the microscope background aesthetic.
In building the game, one of our most critical challenges was creating visual representations of the genetic traits that people would be using to solve the puzzles. People had to quickly grasp how they could manipulate and evolve the traits given the genetic logic. If we couldn't get people to make this mental leap, the game wasn't going to succeed.
Testing trait sequences with fellow IDEO-ers.
One example configuration of possible traits and their sequences that were tested.
A real genetic trait is made up of a snippet of DNA and is a long, winding series of genetic code (or, a lot of AGTC). To make the game approachable, we chose to abstract and simplify the cheese trait representations as visual identifiers that had a bit of mystery and were themselves a puzzle to decode. We tested many different ways of visually representing the traits, the majority of which people could not directly connect to an increasing or decreasing series. The simplest visuals turned out to be the most effective: shapes, dots, and lines.
Final in-game traits.
With simple geometric shapes, people could see that moving from a triangle to a square to a hexagon involved moving through a progression of increasing the numbers of sides. Similarly, lines and dots worked well for understanding how the traits could be mutated. They were easily accessible by players in planning a strategy for mutation.
Less is more
After the weekend of playing the prototype level, we needed a way to determine the optimal solution to these puzzles so that we could give user feedback and help players improve in the game. With Ralph's Killer Muenster, an optimal solution for any puzzle involves making the “most parsimonious” decisions, i.e. using the least number of moves or fewest mutations. We needed to find a way to reward people for doing less, contrary to typical game design where users are rewarded for doing more. Without a clear and appropriate qualitative feedback mechanism, we found that people were finishing puzzles without any sense for how much better they could perform.
Mesquite—the computational biology environment we used to analytically solve the puzzles we had created.
Scavenging the Internet for a tool to analyze phylogenetic trees uncovered Mesquite, a software package built specifically for computational biology and evolutionary analysis. After learning from tutorials produced by Harvard University, we could solve the initial species sets and puzzles we had created, and arrive at the most efficient tree structure and trait mutations.
We used the statistical language R to generate thousands of possible levels, export files for Mesquite to crunch, and output the moves required for each level into a custom level file that we could use to test in our prototype game.
Levels testing and design progressed in parallel with refining the scoring display and making hundreds of other visual and UX enhancements. With our trait visualizations, levels design and scoring mechanism solved, we felt we had achieved something that was both subtle and significant: People who grasped the underlying puzzle dynamics were quite literally using their brains in a way that mimics a core methodology of genetic science. This seemed to hit at the essence of experiential learning.
Muenster launches, Bill Nye approves
Ralph's food truck delivered grilled cheese sandwiches throughout the streets of San Francisco.
The Ralph's Killer Muenster iPad game launched at the 2013 GET (Genomics, Environments and Traits) conference, coinciding with the anniversary of Watson and Crick's seminal paper describing the structure of DNA. At the conference, Bill Nye the Science Guy, a long-time advocate of a science literate society, played the game, and loved it. This made our day. The app also caught Apple's attention, and was listed as a new and noteworthy Education app in the App Store.
To further raise awareness around the game, we created a Ralph's Killer Muenster brand, including a visual identity, packaging, website, Twitter, game trailer, and food truck—all mobilized by Ralph to push the only tool that could save the city (and his reputation) and allow him to lift the health code violations that kept him from producing and distributing cheese. Ralph’s food truck, sent throughout San Francisco, distributed grilled muenster cheese sandwiches.
What we learned
After the launch festivities, our team took a step back to consider what we had learned. Here's the take-away:
Set constraints to help make a hard problem more approachable. Our self-imposed constraint that the game needed to rely on a scientific mechanism helped us narrow in on phylogenetic trees and maximum parsimony as our core concepts.
A good story goes a long way. We spent significant time concocting Ralph's story, and getting the tone and art direction down in order to engage people. We want them to stick around for the more mentally taxing interactions later down the line.
Find the right tool for the right job. We used a variety of prototyping and user testing methods and environments, rather than trying to find or force an all-in-one. Each tool elicited learnings unique to them.
Brad is a mechanical design engineer. He focuses on both deep mechanical design challenges from technical concepting through manufacturing as well as building and designing digital experiences. Brad's passions include flight, technology-enabled design, and suspension of disbelief.