Archive for April, 2010

Towards an Information Science of Zombies

April 2, 2010

Inspired by amazing posts like this one and this one, I would like to try my hand at zombies and the young discipline of Information Science. We are not as established as International Relations or Sociology, but we are very broad. So we have plenty of ways to interpret, analyze, and design for zombies.

Mainstream Design – One of the strongest regularities in designing ZUIs (Zombie User Interfaces) is the logarithmic relationship between time to zombie eradication and size of interface plus distance to zombie. Holding distance to zombie constant at a comfortable (but reasonable) fifteen feet, we exploit this relationship to design a maximally effective Zombie User Interface. The interface, codenamed BFG, showed a 15% (p<.05) decrease in zombie eradication time when tested in a random sample of 20 male university students.

Critical Design – the BFG is a severely flawed device that constrains the user in the classic paradigm of “kill the zombie.” We explore several alternatives such as “be killed by the zombie,” “invite the zombie out to tea,” and most significantly, “engage the zombie in an appropriate social context so as to better understand it.” The high death rate among our subjects should not be taken as a sign of failure, but as a provocative statement that will hopefully stimulate further design in these exciting new directions.

Cognition – current approaches are too focused on the immediate task of eradicating the zombies around you. We define a Zone of Proximal Zombies(ZPZ) as the space between the zombies right around you (that you can take out with your shotgun) and the larger population of zombies in your area (that you can take out with the help of other zombie fighters). We argue that zombie eradication tools should be designed with the ZPZ in mind to develop zombie eradication expertise and help people grow as zombie killers.

Network Analysis – Consider the network where nodes are zombies Z and there is an edge E between two zombies Z1,Z2 if Z1 and Z2 have a social tie (sharing the brain of a human, shuffling together for more than 5 minutes). We show that this network has a power-law degree distribution, suggesting that some zombies are unusually sociable, shuffling with hundreds or even thousands of other zombies at different points in their zombie unlife. These “zombie hubs” are incredibly important to the connectivity of the network, and their removal will fracture the zombie society into a large number of tiny components that can be focused down and eradicated piecemeal.

Natural Language Processing – by analyzing a corpus of over 400,000 zombie groans, we have come up with a generative model of Zombie. This model leads to a statistically significant improvement in MAP over the baseline, which assumes every document contains only the term ‘brains.’ We also show that the generative model of Zombie is superior to existing discriminative approaches for over 300 Zombie search tasks hand-coded by a team of 5 Zombies (inter-coder reliability = 0.12).

Data Mining and Search – with the recent explosive growth of zombies, the problem of finding a particular zombie has become exponentially difficult as most zombies share a basic set of characteristics (taste for brains, vacant look, absence of higher cognitive functions) and relevant zombies become lost in the sea of undead. Current approaches focus primarily on the individual characteristics of zombies and ignore a crucial relationship that exists between zombies – the Bite, whereby a zombie infects a human who then becomes a zombie and can bite other humans, and so on. We leverage the structure imposed by Biting relationships between zombies to create ZombieRank, a unique zombie search engine that relies on implicit authority relationships between the biter and the bitee to generate a set of highly relevant zombies for any particular zombie search. ZombieRank is now live at!