Report Card: Logan’s Run

LogansRun-Report-Card

For our purposes, Dome City is a service. Provided by the city’s ancestors to provide a “good life” for their cloned descendants in a sustainable way, i.e., a way that does not risk the problems of overpopulation. The “good life” in this case is a particular hedonistic vision full of fashion, time at the spa, and easy casual sex.

There’s an ethical, philosophical, and anthropological question on whether this is the “right” sort of life one would want to structure a service around. I suspect it’s a good conversation that will last at least a few beers. Fascinating as that question may be, looking into the interaction design requires us to accept those as a given and see how well the touchpoints help these personas address their goal in this framework. Continue reading

A disaster-avoidance service

The key system in The Cabin in the Woods is a public service, and all technological components can be understood as part of this service. It is, of course, not a typical consumer service for several reasons. Like the CIA, FBI, and CDC, the people who most benefit from this service—humanity at large—are aware of it barely, if at all. These protective services only work by forestalling a negative event like a terrorist action or plague. Unlike these real-world threats, if Control fails in their duties, there is no crisis management as a next step. There’s only the world ending. Additionally, it is not typical in that it is an ancient service that has built itself up over ages around a mystical core.

So who are the users of the service? The victims are not. They are intentionally kept in the dark, and it is seen as a crisis when Marty learns the truth.

Given that interaction design requires awareness of the service in question, as well as inputs and outputs to steer variables towards a goal, it stands that the organization in the complex are the primary users. Even more particularly it is Sitterson and Hadley, the two “stage managers” in charge of the control room for the event, who are the real users. Understanding their goals we can begin an analysis. Fittingly, it’s complex:

  • Forestall the end of the world…
  • by causing the (non-Virgin) victims to suffer and die before Dana (who represents the Virgin archetype)…
  • at the hand of a Horrible Monster selected by the victims themselves…
  • marking each successful sacrifice with a blood ritual…
  • while keeping the victims unaware of the behind-the-scenes truth.

Sitterson and Hadley dance in the control room.

Part of a larger network with similar goals

This operation is not the only one operating at the same time. There are at least six other operations, working with their particular archetypes and rituals around the world: Berlin, Kyoto, Rangoon, Stockholm, Buenos Aires, and Madrid.

To monitor these other scenarios, there are two banks of CRT monitors high up on the back wall, each monitor dedicated to a different scenario. Notably, these are out of the stage manager’s line of attention when their focus is on their own.

The CRT monitors display other scenarios around the world.

The digital screens on the main console are much more malleable, however, and can be switched to display any of the analog video feeds if any special attention needs to be paid to it.

The amount of information that the stage managers need about any particular scenario is simple: What’s the current state of an ongoing scenario, and whether it has succeeded or failed for a concluded one. We don’t see any scenario succeed in this movie, so we can’t evaluate that output signal. Instead, they all fail. When they fail, a final image is displayed on the CRT with a blinking red legend “FAIL” superimposed across it, so it’s clear when you look at the screen (and catch it in the “on” part of the blink) what it’s status is.

Sitterson watches the Kyoto scenario fail.

Hadley sees that other scenarios have all failed.

One critique of this simple pass-fail signal is that it is an important signal that might be entirely missed, if the stage managers’ attentions were riveted forward, to problems in their own scenario. Another design option would be to alert Sitterson and Hadley to the moment of change with a signal in their peripheral attention, like a flash or a brief buzz. But signaling a change of state might not be enough. The new state, i.e. 4 of 7 failed, ought to be persistent in their field of vision as they continue their work, if the signal is considered an important motivator.

The design of alternate, persistent signals depend on rules we do not have access to. Are more successful scenarios somehow better? Or is it a simple OR-chain, with just one success meaning success overall? Presuming it’s the latter, strips of lighting around the big screens could become increasingly bright red, for instance, or a seven-sided figure mounted around the control room could have wedges turn red when those scenarios failed. Such environmental signals would allow the information to be glanceable, and remind the stage managers of the increasing importance of their own scenario. These signals could turn green at the first success as well, letting them know that the pressure is off and that what remains of their own scenario is to be run as a drill.

There is a Prisoner’s Dilemma argument to be made that stage managers should not have the information about the other scenarios at all, in order to keep each operation running at peak efficiency, but this would not have served the narrative as well.