Twitch Plays Pokemon by Anonymous, and: Mari/O by Seth Bling, and: Dreeps by Hisanori Hiraoka (review)

If performance works with agency, then games are the genre of performance that best highlight the ostentatious display of such agency. In particular, video games and professional spectator sports seemingly transport players and viewers into bodies perfectly oriented toward achieving pointless, artificial goals that replace unattainable, authentic ones. This certainly describes one register of gameplay, and one that is worth a lot of money—as I am writing this, The Witcher 3, starring a be-stubbled, poisoned-sword-wielding loner, just moved 4 million copies in under a month—but it is far from the only mode in which games can function and even prosper. Along with the rest of the past half-century’s explosion of diversity in game forms and styles, there have also been an enormous variety of experiments in modes of agency. This review highlights a few popular examples from the past couple of years that present extreme points on one spectrum of methods for mapping players to digital agents.

The traditional model of idealized protagonist identification in old media finds its video game correlate in a single player mapped to a single agent. If this stands at the midpoint on the continuum of agent/player relationships, then two extreme points would be: as many players as possible mapped to a single agent, and no players mapped to a single agent. Both of these extreme positions have popular exemplars from the last two years, and the most famous of these exemplars are “hacks”—modifications of existing games intended specifically to press them into other modes of agency.

Both of these positions act as synecdoches for categories of postindustrial labor. The second position names authentic automatism: labor replaced by mechanical effort; the first names counterfeit automatism: intensified labor that serves to hide itself and feign automatism. This is the labor involved in “sharing economy” services like Uber, which collapse human effort down to unified interfaces. Its historical prototype is Wolfgang von Kempelen’s 1770 fake robot, the “Mechanical Turk,” a device that appears to be a chess-playing automaton, but is in fact a contrivance for hiding a human chess master. It has given its name to a significant crowdsourcing business model, a service provided by Amazon that pays large numbers of people small amounts of money to perform atomized tasks. These phenomena look like technological coups only because people are working extra-hard. Counterfeit automatism is a deeply theatrical technology: it produces magic through a strict division of stage and backstage, and through a surplus of sweat and tears.

The Hivemind laboriously navigates Red through town in Twitch Plays Pokemon.

Twitch Plays Pokemon (TPP) exemplifies and parodies this position. TPP launched in February 2014. Its first segment completed in March 2014 and the channel is still in operation. An anonymous artist and programmer—only his male gender and Australian nationality are known—opened a channel on the video game streaming platform Twitch with the short description: “I’m a robot playing Pokemon, I don’t know what I’m doing, tell me which buttons to press.” It performed more or less as advertised.

The Pokemon game in question was Satoshi Tajiri’s Pokemon Red, Nintendo’s 1998 English-language release of the 1996 Japanese original, first playable only on the Nintendo Game Boy. A classic of 1990s video games, it is often credited with popularizing the still-successful intermedia Pokemon franchise. The details of Pokemon Red gameplay are stun ningly dull to describe, but in its original incarnation players pushed keys to navigate a protagonist figure through a fictional world. The TPP implementation of Pokemon Red left the game systems relatively...