Home » Featured

Art-Science Partnerships Help Students Gain Deeper Understanding of Research

1 October 2017 177 views 3 Comments
Eric Laber, North Carolina State University
    Posters illustrating the video games created by Laberlabs. Laser Foxes was illustrated by Danny Schmidt.

    Poster illustrating the video games created by Laberlabs. Laser Foxes was illustrated by Danny Schmidt.

    Posters illustrating the video games created by Laberlabs.

    Poster illustrating the video games created by Laberlabs. Flying Squirrel was illustrated by Tea Blumer.

    Laber Labs was created to foster collaboration between artists and statisticians in the pursuit of cross-disciplinary research and novel approaches to outreach and scientific translation.

    The lab currently has 30 members, including 20 statistics PhD students; five undergraduate students in fields such as computer science, graphic design, industrial design, and film; two high-school interns; and three staff members with expertise in graphic design, animation, web programming, music, and sound engineering.

    The diverse expertise in the group allows us take on a wide range of problems. Currently, the lab is focused on four major research thrusts:

    • (T1) Precision medicine
    • (T2) Human-computer interactions and artificial intelligence
    • (T3) Data-driven monitoring and disruption of sex trafficking
    • (T4) Real-time decision support for infectious diseases

    In addition, we have an outreach program designed for middle- and high-school students. In this program, students learn about statistics, machine learning, and data-driven decision-making by creating autonomous artificial intelligence agents that play video games. To date, we have developed the following five video games, each designed to illustrate key concepts in data-driven decision-making.

    Zombies vs. Treadmills: A simple turn-based game in which players must orient a grid of treadmills to redirect a hoard of zombies away from humans and into active volcanoes. This game is a (hopefully) more engaging variant of grid-world—a canonical, sequential decision-problem from computer science. Through gameplay, students learn basic definitions and concepts in data-driven decision-making and gain the skills needed for more advanced (and exciting) games.

    Flying Squirrels: A side-scrolling game in which the player dictates the diving and soaring actions of a flying squirrel to gain speed and avoid enemies. Using the game as a simulation environment, students learn to use classification to teach a computer to mimic a human player and thereby enable the computer to play autonomously.

    Space Mice: A fly-and-shoot game in which the player must maneuver a flying cat to shoot invading mice from outer space. Students learn by teaching the computer to play using a Bayesian variant of approximate dynamic programming—a technique so successful that the computer ultimately plays orders of magnitude better than the best human player.

    Laser Foxes: A two-player fighting game in which a human player battles an adaptive computer agent that learns from gameplay experience. In this game, students learn about adversarial decision-making and replicate the behavior of the computer agent.

    Snack Attack: A live-action food battle game in which players build armies and decide how, where, and when to deploy troops to destroy the opposing army. Students handcraft artificial intelligence algorithms for a computer player and pit their computer creations against each other.

    We build the games, artificial intelligence, and other educational materials in-house, so our outreach involves graphic designers, sound engineers, and statisticians. Lab members have also collaborated on series of short videos, called 2MinStats, which aim to teach core statistical concepts without math. These games and videos can be found at Laber-Labs.com.

    Outreach is an obvious avenue for collaboration between designers and statisticians; however, our nonstatistician lab members are also involved in each research thrust. For example:

    • In (T1), we built interactive visualizations to help communicate risk and trade-offs across competing treatment options.
    • In (T2), we created 3D models of physical environments to train a robotic arm to move pieces on a game board.
    • In (T3), we used image data generated by projections of 3D human models to train computer vision to identify physical attributes such as hip-to-waist ratio.
    • In (T4), we build interactive decision support systems that allow decision-makers to manipulate estimated optimal treatment allocation rules.

    Adding designers to the lab changed the way we conducted and communicated science and expanded the scope of research translation and dissemination.

    Such lab environments provide statistics PhD students with a substantially different educational and training experience than a traditional departmental environment. Students work in teams on projects and attend regular project-specific group meetings in which members generate, critique, and refine ideas. There are weekly lab meetings at which a student presents their research and lab members brainstorm about alternative solutions, potential weaknesses with the proposed approach, and connections to other problems. These brainstorming sessions give the presenting student experience defending their research and the rest of the lab experience quickly generating potential solutions to a complex statistical problem. In some instances, these brainstorming sessions have changed the course of a student’s dissertation research or spawned new research projects.

    Students are encouraged to generate their own research directions and pursue projects outside of their thesis work. Because their graduation is not contingent on the success of these projects, students tend to be more ambitious and creative with them. Recent student-driven research projects include building a real-time cat-tracker and using natural language processing to create a digital clone of the 1980s painter Bob Ross.

    As a faculty member, running an interdisciplinary lab made up of artists and scientists has been rewarding through both better student development and research output. A commonly voiced concern is that having statistics PhD students engaged in outreach projects or art-science partnerships will detract from the quality, rigor, or depth of their thesis work. Our experience has been that these activities lead to deeper investment in research projects, stronger engagement with the scientific community, and greater imagination in problem solving.

    The lab’s scientific mission is to close the research-practice gap in data-driven decision-making through methods development, education and outreach, and research translation. Students who buy into this mission and invest in the lab tend to do better science.

    If you are interested in learning more about Laber Labs or getting started on your own lab, contact Eric Laber or, for more information, visit Laber-Labs.com. You can also follow @LaberLabs on Twitter.

    1 Star2 Stars3 Stars4 Stars5 Stars (No Ratings Yet)


    • critic said:

      This mostly seems like a poor use of a stat profs time. Are most of his students art majors?

    • Ted Hamton said:

      it seems unlikely a professor could provide adequate funding and project help when they have that many students. Also, none of these listed projects seem like they would help a STEM student find a job later.

    • Matt said:

      Saw their display at Engineering Fair two years ago, pretty cool!

    Leave your response!

    Add your comment below, or trackback from your own site. You can also subscribe to these comments via RSS.

    Be nice. Keep it clean. Stay on topic. No spam.

    You can use these tags:
    <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <s> <strike> <strong>

    This is a Gravatar-enabled weblog. To get your own globally-recognized-avatar, please register at Gravatar.