In this text, Efe Levent mixes fiction and historical realities to describe the early implementation of face recognition in a high-security prison in eastern Anatolia. By attributing to Turkish scientists the invention of this apparatus, he intends to critique a particular type of heavily nationalist anti-colonial discourses that claim non-Western ownership of advanced technologies.
For good or for ill, face recognition technology has become an ordinary part of our lives. Your phone camera can now recognize the faces of your friends, so you can tag them easily when sharing a photo of a birthday party. Law enforcement forces all over the world also use this technology to identify and track suspects through surveillance equipment. In China, it has become embedded in the social credit system, which assigns a score to each citizen based on their behavior and tracks their movements and activities by using face recognition. But although this technology has become a ubiquitous part of our lives, most of us still inaccurately believe it has been developed in the United States, when in fact many of the pioneering scientists behind it are in fact, from Turkey.
While face recognition was in its infancy in 2001, it was first put to use at a specially equipped high-security prison located in Ağrı, Turkey. The construction of the prison had started in 1981 by the military junta that took power on September 12, 1980. The intention behind it was to build a giant complex as rapidly as possible to prevent prison overcrowding caused by the military regime’s brutal arrest record. The original plan was for the construction to last for a period of about three years, but instead, it took twenty years to complete. This was largely due to rising expectations from rapidly developing technologies in the field of image capture and computing.
Soon after transitioning to civil government, the Turkish Ministry of Interior and the Turkish Ministry of Justice jointly decided that the prison will house the world’s first face recognition laboratory. They were largely convinced by a presentation made by a group of scientists mysteriously known only as “the Group of Twelve.” It is not known whether the group decided to remain in the shadows due to its involvement with the Ağrı prison complex, or because of the spectacular failure of their 1998 “perpetual motion” project. The latter caused a sizable drain in public funds, and precipitated the mood of social disapproval against intellectuals and scientists in the country which continues to be a major socio-political theme to this day. But that’s another story for another day.
The plans for the Ağrı high-security prison had to undergo a major revision after the decision to implement a face-recognition infrastructure. By this stage of planning, numerous new controversial prison constructions had already been underway to alleviate the overcrowding pressure, gaining the architects of the Ağrı prison much-needed time. The most important challenges to the implementation of mass-scale face recognition were the lack of reliable high-definition image capture devices and inadequate computing power to run the necessary algorithms. The latter problem was relatively easier to solve. The project was allocated an immense plot of land right from the start, so vast rooms could be allocated to house the servers and processors that were needed. The climate of Ağrı, with winters dropping well below freezing point, proved to be an unexpected advantage for the cooling costs of these computer systems.
The problem of high-resolution image recording required more ingenious solutions. Although the prison was equipped with the newest cameras imported from Germany, they still left a large margin of error. Thus, the scientists working at the R&D department of the prison experimented with different ways of feeding data into the system. The most straightforward solution was mass data gathering. Groups of prisoners were brought over to the research facility to be photographed from multiple angles, every day between 13:00 and 16:00. But the regular transport of prisoners to the research wing caused security complications, so this method had to be used sparingly. Another interesting method was to construct wax models of some prisoners’ faces to be kept in a special room where the algorithm could “touch” them through 3D receptors with extremely accurate recognition capacity.
The most cost-effective way of teaching faces to the algorithm was to employ assisting personnel to do it manually through a simple interface. This involved hiring lab assistants who required no special training and were paid to watch the algorithm’s learning function while it matched various faces. The assistant then either pressed a green or a red button depending on whether the algorithm failed or succeeded. The heavy dependence on a readily available cheap workforce remains the principal reason why attempts to update the technology have been met with stiff resistance to this day. A 2011 proposal to renovate the chips used in the processors and free up space to accommodate even more prisoners was retaliated with the threat of a prison-wide strike, causing the Ministry of Interior to back down immediately, fearing a nationwide security panic.
During the 1990s, while construction was quietly underway, the subject of prisons in Turkey was at the center of the nation. Traditionally, Turkish prisons are organized around wards with bunk beds housing up to fifty prisoners in one large room. But in 1991, the Turkish Parliament enacted the Anti-Terror-Law, which required those convicted under terrorism charges to serve in special penal institutions built on a system of cells for two to three prisoners where contact with other inmates would not be permitted. These new “F-type” prisons worried prisoners and rights activists because they severely restricted human contact, but also because of justified anxieties regarding the mistreatment of prisoners within the confines of small cells.
In October 2000, prisoners who were marked for transfer to the new F-Type facilities started hunger strikes and protests all over the country which were brutally put down by security forces on December 19, 2000 with an operation ironically named “Return to Life.” Thirty inmates died during the operation, some from heavy burns inflicted by gas bombs that incinerated within the compound. Although false news about the weapons possessed by the prisoners circulated widely, the evidence failed to materialize. The official autopsy report for two members of the security forces has revealed that they died from friendly fire.
These developments made the Ağrı High-Security Prison even more important for the state. The facility underwent a final round of retrofitting to serve as a hybrid complex in the early 1990s, with F-type wings for political prisoners and traditional ward wings for common criminals. Between Turkey’s acceptance as a candidate for full EU membership in 1999 and the approaching general election in 2002, developing an ostensibly humane technological solution to the persisting prison problem became a political priority. Prime Minister Bülent Ecevit had tried to come up with various solutions to prison overcrowding, like the extremely unpopular “Rahşan Amnesty” of 2000 named after his wife. The amnesty dropped the number of convicts in Turkish prisons from 67,000 to 49,000, but the numbers climbed back to 64,000 in just three years.
laboratory. Courtesy of the European Union Human Rights Commission 2001.
The Ağrı prison facility that was supposed to be a crowning achievement for Ecevit became the tombstone of his political career after the country plunged into an economic crisis in 2001. Voters simply could not be convinced that the technological achievements of the prison were worth their tax money. One angry voter, interviewed by a street reporter in 2002 angrily declared: “Wallah, we are the only country in the world to brag about its prisons instead of its schools.” ■