ArticlePDF Available

Caught inside the black box: Criminalization, opaque technology, and the New York subway MetroCard

July 2019
The Information Society

DOI:10.1080/01972243.2019.1644410

Authors:

Santa Clara University

This article investigates how an account of hidden, internal properties of an everyday technology became a framework to interpret human action as a serious crime. Using a case study situated in the New York subway system, I examine the criminalization of a practice of New York’s poor known as “selling swipes” performed by so-called “swipers”. A high court came to classify the practice as felony forgery, interpreting it though an expert-witness account of how objects physically manipulated by swipers interact with a secretive, proprietary digital information system. Thousands of felony arrests – overwhelmingly of nonwhite men – have been legitimated under this theory, in which the crime occurs on a plane of technical interactions to which swipers have no access. Through close examination of the underlying technology (known as MetroCard), however, I show considerable problems in the authorities’ understanding of the technology, illustrating the hazards of interpreting human action through proprietary or complex systems, especially as they are represented solely through expert accounts. The case demonstrates fresh connections between technology and unequal outcomes in the U.S. criminal justice system, and suggests an emerging form of social vulnerability, to interpretations of our actions through the logic of technologies black-boxed to us.

The Mattocks account of bent metrocards as forged instruments. Notes. Selling swipes begins with a metrocard in which the lowest of the two value-fields shows zero fares remaining, which will always be rejected by the turnstile computer (illustrated in swipe #0). Cross-examination of an expert-witness in the 2005 People v. Mattocks criminal trial characterized a bent metrocard (as in swipe #1) as capable of producing free entry into the subway because the fare processing system "gives the benefit of the doubt" in case a metrocard had been somehow damaged. Testimony also made reference to the existence, but not the substance, of subsequent steps necessary to achieve that outcome. Those steps -not revealed in this account -suggest ambiguity in the legal classification of a bent metrocard as a forged instrument.

…

Figures - uploaded by Noah McClain

Content may be subject to copyright.

Content uploaded by Noah McClain

Content may be subject to copyright.

Content uploaded by Noah McClain

Content may be subject to copyright.

Full Terms & Conditions of access and use can be found at

https://www.tandfonline.com/action/journalInformation?journalCode=utis20

The Information Society

An International Journal

ISSN: 0197-2243 (Print) 1087-6537 (Online) Journal homepage: https://www.tandfonline.com/loi/utis20

Caught inside the black box: Criminalization,

opaque technology, and the New York subway

MetroCard

Noah McClain

To cite this article: Noah McClain (2019): Caught inside the black box: Criminalization,

opaque technology, and the New York subway MetroCard, The Information Society, DOI:

10.1080/01972243.2019.1644410

To link to this article: https://doi.org/10.1080/01972243.2019.1644410

Published online: 25 Jul 2019.

Submit your article to this journal

View Crossmark data

Caught inside the black box: Criminalization, opaque technology,

and the New York subway MetroCard

Noah McClain

Department of Social Sciences, Illinois Institute of Technology, Chicago, Illinois, USA

ABSTRACT

This article investigates how an account of hidden, internal properties of an everyday

technology became a framework to interpret human action as a serious crime. Using a case

study situated in the New York subway system, I examine the criminalization of a practice

of New York’s poor known as “selling swipes”performed by so-called “swipers”. A high court

came to classify the practice as felony forgery, interpreting it though an expert-witness

account of how objects physically manipulated by swipers interact with a secretive, propri-

etary digital information system. Thousands of felony arrests –overwhelmingly of nonwhite

men –have been legitimated under this theory, in which the crime occurs on a plane of

technical interactions to which swipers have no access. Through close examination of the

underlying technology (known as MetroCard), however, I show considerable problems in

the authorities’understanding of the technology, illustrating the hazards of interpreting

human action through proprietary or complex systems, especially as they are represented

solely through expert accounts. The case demonstrates fresh connections between

technology and unequal outcomes in the U.S. criminal justice system, and suggests

an emerging form of social vulnerability, to interpretations of our actions through the logic

of technologies black-boxed to us.

ARTICLE HISTORY

Received 10 April 2019

Accepted 5 July 2019

KEYWORDS

Criminal prosecution;

criminalization; inequality;

informal economy; law;

magnetic stripe; MetroCard;

New York subway

Introduction

In the early 1990s, the New York City subway intro-

duced a new fare system based on a magnetic farecard

(called “MetroCards”

) to replace its aging revenue

system reliant on metal tokens. By the end of that

decade, a growing number of metrocard users had

learned that an extra, unbought ride could be

extracted from a spent metrocard by bending the card

in a specific spot on its magnetic strip before

“swiping”it at a subway turnstile. This trick became

central to a hustle of the urban poor in which

“swipers”(as they came to be known) stand at subway

entrances with a stock of bent metrocards, selling

entry to the subway at some discount below the offi-

cial fare.

Swipers were subject to significant policing,

but it was unclear what laws swipers were violating.

The state soon passed a law to make the practice

a misdemeanor, carrying a potential penalty of three

months of incarceration, but police and prosecutors

also sought more significant charges, and began

to classify swipers’bent metrocards as “forged

instruments”. When charged as a misdemeanor,

making or possessing a forgery can carry a one-year

sentence.

Yet in one jurisdiction –the borough of

Manhattan –police began, in the early 2000s, to arrest

swipers with felony forgery, a charge which can carry

a much higher penalty. According to police data, over

thirteen years, more than eleven thousand accused

swipers were arrested under felony forgery charges;

92 percent of these arrestees were black or Hispanic,

and 97 percent were male.

When this policing practice emerged, anyone look-

ing for an explanation of just how bending a magnetic

farecard can produce a forgery would have discovered

that the technology underlying the metrocard system

is proprietary and kept secret from the public.

The

inner workings of the metrocard technology, however,

rose to the forefront of a legal case before the highest

court in New York State in which the court affirmed

the theory that a bent metrocard can be a forgery.

Interestingly, the court did not derive its rationale

from what is plainly observable: a bent-metrocard

cause and unpaid subway ride effect. Instead, it found

its rationale inside the technology, or at least in an

CONTACT Noah McClain mcclain1@gmail.com Department of Social Sciences, Illinois Institute of Technology, Seigal Hall #116, 3301 S. Dearborn

Street, Chicago, IL 60616, USA.

Published with license by Taylor & Francis ßNoah McClain

THE INFORMATION SOCIETY

https://doi.org/10.1080/01972243.2019.1644410

expert-witness description of the inner workings of

the metrocard system and how those inner workings

interact with a swiper’s bent metrocard.

Although metrocards are ubiquitous in the pockets

of subway users, the metrocard system is a black box,

generally known only in the simplified terms of what

it does, not how it does it (MacKenzie 2005; Latour

1999; Pinch and Bijker 1984). In effect, swipers were

accused of a specific crime by dint of technical fea-

tures they could neither directly inspect nor challenge

with an alternative interpretation of the technology.

Moreover, swipers came to be defined as felons for

what they technically do, that is, for their actions

framed and interpreted through a context of the tech-

nology. To analyze this scheme of things the term

“technical agency,”which has been used elsewhere to

describe the agency of technology (see Neff, Jordan,

McVeigh-Schultz, and Gillespie 2012), will be re-pur-

posed here as a term of convenience for: human

action interpreted in strict reference to a relevant

technology, independent of the actor’s awareness

or intentions.

The criminalization of technical agency, as we see

in the case of swipers’manipulation and possession of

bent metrocards, raises a number of compelling ques-

tions: What are the politics of public systems which

operate on proprietary and secret underlying technol-

ogy (see Levine 2007)? Can human criminality

become an artifact of technical design? What is the

role of expertise in the joint interpretation of technol-

ogy and criminality? Just how do mundane technolo-

gies intersect with problems of race and social

vulnerability?

The criminalization of selling swipes offers a case

in which a subsistence practice of the urban poor has

been subject to a punitive interpretation through the

frame of a proprietary technology. Defendants have

had no direct access to the technology, or to inde-

pendent expertise, making existing claims about how

the technology works more or less irrefutable, and by

extension, making claims about how a crime is com-

mitted through that technology more or less irrefut-

able as well. Asymmetrical distributions of technical

knowledge can be especially impactful when, as this

case shows, the prevailing account of the technology

is flawed, but still serves as the basis for the elabor-

ation of criminal law. More broadly, the case illus-

trates how understandings of technology may become

a fresh vector through which existing inequalities

travel, when claims about complex, proprietary tech-

nology are leveraged against certain groups and prac-

tices. The complexity or secrecy of technology can

itself be impactful when claims about human behavior

are built upon them.

This article, then, investigates the emergence of

swipers as a policing and legal controversy which was

eventually settled on the “technical grounds”of met-

rocard technology; that is, on dynamics hidden in the

black box of the metrocard system. To understand

and situate that outcome, this article digs deep into

the metrocard technology, precisely where a high

court placed social stakes, and upon which it affirmed

an aggressive pattern of policing and prosecution.

However, we find that there are a host of ambiguities

in how the metrocard technology was understood by

the court, and how it might relate to forgery statutes

–yet they were not identified in the court case (or

heretofore in any venue). These ambiguities suggest

alternate interpretations of technical action that

defendants and their advocates have been precluded

from proposing in their defense.

Opening the black box of the metrocard can show

how the workings of a technological system –and

how that system is understood, or not –can have

heretofore-unacknowledged relevance for sociological

explanation. Our case allows us to see how a claim

about how a technology “really works”affirms and

stabilizes policing strategy firmly leveraged against

minority communities. It is also a step towards

addressing a dual disciplinary myopia articulated by

Rodr

ıguez-Mu~

niz (2016): the reluctance of Science

and Technology Studies (STS) scholarship to address

issues of race so central to sociology, and the parallel

reluctance of sociologists to address the role of the

material and technical in the making of the social.

Rodr

ıguez-Mu~

niz suggests to bridge the divide

through engagement with the “sociotechnical material-

ity of race and racial domination”(p. 214) –the

actual stuff which helps produce stratified racial

orders, from technologies of knowledge production to

medical devices (see Law 1991; Braun 2014). This case

study advances that agenda by showing that, while the

inequalities delivered through policing and criminal

justice practices may hinge on issues long explored by

sociologists and socio-legal scholars, they can hinge

also on accounts of a technology to which criminal

justice actors point to justify their practices.

This inquiry sits at the intersection of important

themes in existing literature and policy debates.

Policing at the subway turnstile has recently emerged

as a focal point in the politics of city policing, but has

long been at the center of “broken windows”strategies

deployed in New York since the 1990s –now the

archetype for national transformations in urban

2 N. MCCLAIN

policing. The nature of expert scientific interpretation,

and the technology and methods of forensic analysis

in the U.S. legal system are important questions in

both socio-legal studies and STS –the criminalization

of selling swipes relies upon an expert account of

swipers’crime and also an entire plane of action

inside the technology where the crime allegedly

occurs. Furthermore, scholars and commentators of

many stripes increasingly point to the emergence of

technology as an axis and vector of social inequality;

this article shows how claims about technology, and

related interpretations of human action, is an emerg-

ing terrain on which vulnerable groups become still

more vulnerable. Before discussing methods and data,

I briefly situate how this case is framed by these

debates, and outline how this case links them.

Criminalization at the subway turnstile

In a vast city where fewer than half of households

own cars and subway fares continue to rise faster than

real wages, the subway turnstile has not only been a

gateway into a major public good, it has also been a

gateway into the criminal justice system. The adoption

of “broken windows”policing strategy in the early

1990s led to an especial focus on fare-beating at sub-

way entrances, where a small fare-beating transgres-

sion offered police a pretense to conduct frisks,

warrant checks, and develop justifications for searches

and arrests (Harcourt 2001; Kohler-Hausmann 2018).

The best available data on the policing of fare eva-

sion in the subway have long suggested that policing

fare evasion contributes to a broader criminalization

of everyday life for poor nonwhite urbanites (see Rios

2011; Stuart 2016). Recent research has relied on tri-

angulation techniques to study racial and ethnic bias

in the policing of fare evasion by comparing frequen-

cies of arrests with neighborhood demographics where

the arrests occur, adjusted for the volume of subway

passengers (Flagg and Nerbovig 2018; see also Stolper

and Jones 2017). In 2018 concern for the bias in

police scrutiny led to a new law requiring the New

York Police Department (NYPD) to report data on

fare-evasion policing,

which made new fare-evasion

arrest data available by race, gender, and age for an

18 month period. While the data (NYPD 2019) is

hardly granular (Lancman 2018; Winston 2018), the

broad strokes are telling: over that period, ninety-two

percent of Theft of Service (TOS) arrests involved

black or Hispanic arrestees. Even these hard numbers

likely obscure the true textures of subway fare

enforcement thanks to how the NYPD records and

reports arrest statistics, by the single most-serious

charge. A police encounter that begins as a stop for

fare-beating might, for example, become an arrest

which also includes a charge for drug possession, and

be counted as an arrest for drugs rather than for TOS.

When confronting fare-evaders, New York police offi-

cers have discretion to make a misdemeanor arrest, issue

a summons returnable to criminal court, or (most pref-

erable to the accused) issue a civil “Transit Adjudication

Bureau Notice of Violation”(TAB/NOV). While consti-

tuting 92 percent of official arrests, blacks and Hispanics

received 69 percent of TAB/NOVs over that same eight-

een-month period (NYPD 2019a). By official policy, a

fare-beating suspect may be disqualified from receiving a

TAB/NOVinlieuofarrestforanumberofreasons,

including not having a verifiable address, or fitting into

the category “transit recidivist”This category includes

thosewhoareonprobationorparole,orhavereceived

more than two TAB/NOVs in the prior two years

(which can be issued for things like laying down on a

subway bench; see NYPD 2019b). This policy thereby

compounds bias in policing by ratcheting up the conse-

quences of the next encounter with the police for some-

one with prior minor offenses, or who is homeless. We

cannot tell, however, what proportion of arrestees for

TOS were disqualified from receiving a TAB/NOV, or in

what proportion of these arrests police might have issued

aTAB/NOVbutdecidedtopursuethemorepunitive

path.

TOS policing is in many ways still mysterious.

The policing of swipers is even more elusive to

decipher. Swipers have been arrested under an array of

charges, including TOS, loitering, and petit larceny, but

there is no ready way to discern from statistical aggre-

gates which of the arrests stem from selling swipes. Even

though powerful local political actors have expressed

concern over the apparent disproportionate policing of

swipers in minority neighborhoods (Stewart 2018), their

efforts to confront the issue have been stymied by lack

of data to even demonstrate the pattern.

The focus on

the deployment of police scrutiny is only part of the pic-

ture, however; the criminalization of swipers’stock-in-

trade, and the basis for felony-level charges, rests not

only on police strategy and uses of discretion; it also

rests on an interpretation of the metrocard developed

through expert testimony, framed by prosecutors, and

stabilized by a high court.

Technology, expert interpretation, and criminal

legal process

The STS literature has explored the ways in which sci-

entific and technical knowledge is deployed and

THE INFORMATION SOCIETY 3

contested before judges and juries possibly ill-

equipped to evaluate it. Litigation such as toxic torts –

when liability may hinge on scientific evaluation –often

encompasses broader issues in the relationship between

science and society, such as how expert authority is

established as credible, how complex, messy procedures

of analysis are represented to courts, and how certain

claims prevail and are accepted as true by nonscientific

audiences (Jasanoff 1997).

With significant exceptions (Jasanoff 1998; Lynch

1998), criminal cases do not generally entail contests

of scientific expertise between well-matched adversa-

ries. Rather, criminal defenses generally have signifi-

cantly less access to expertise than their prosecuting

counterparts (Giannelli 2004; Murphy 2007). Limited

financial resources are one obvious reason, but others

are noted by scholars (Hoeffel 1990; Lynch and Cole

2005; Murphy 2015; Giannelli 2018; Saks and

Faigman 2008). Prosecutors are the key customers of

private crime labs and forensic experts, creating disin-

centives for them to report irregularities or error, or

express doubt. Moreover, physical evidence ends up

in the hands of the state, and, in the case of DNA

analysis, the prosecution’s tests may consume all test-

able material, leaving none for defenses. In effect,

expert witnesses are prone to overstate the credibility

of forensic techniques while courts “have often been

seduced by superficially-plausible explanations of tech-

nique or method”(Mnookin 2008 p. 343). Even the

development of independent expertise is hampered by

high costs of entry and other obstacles, as in the case

of forensic techniques like GPS tracking, data mining

and DNA typing, which entail a “complex interpretive

machinery”and often require access to privately-held

computerized records and databases not readily avail-

able to independent parties (Murphy 2007).

Unequal access to expertise conjoins with formal

restrictions on access to technologies in ways that

may be pivotal for criminal justice outcomes.

Technologies of criminal justice decision-making are

opaque and biased; “smart”products profile certain

identities, places, and situations as somehow suspect

according to inscrutable calculations (Crawford and

Schultz 2014; Joh 2017b). Criminal sentencing and

parole decisions may be informed by an algorithm

which has been observed to rank minorities as riskier

than white counterparts; yet their outputs –even

when based on outright inaccurate input data –have

resisted challenge because aggrieved people and their

advocates cannot prove just how the proprietary algo-

rithm weighs input variables (Angwin, Larson, Mattu,

and Kirchner 2016; Wexler 2018). Moreover, courts

often protect technologies at the core of forensic evi-

dence gathering (e.g. DNA kits) from adversarial scru-

tiny (Hoeffel 1990; Mellon 2001). In a telling New

York case, a judge ruled that a defendant accused of

drunk driving was not entitled to examine the source

code of the breathalyzer device, reasoning that the

state does not “possess”the source code sealed within

the device; that the proprietary code would likely be

unreadable, anyway; and that breathalyzer device

should be trusted by the court because it is trusted by

the state health authority.

In other words, when the

code is black boxed to the state, it can also be forcibly

black-boxed for defendants.

The metrocard technology has long been veiled in

secrecy by the Metropolitan Transportation Authority

(MTA), the agency which operates the subway. Expert

examination of bent metrocards used in prosecutions

of swipers can be feasibly performed by only the

smallest set of insiders in the MTA, who have exclu-

sive access to equipment necessary to evaluate if any

specific bent metrocard has the proper data configur-

ation to count as a forged instrument. More funda-

mentally, the criminalization of selling swipes relies

on an expert account of an entire technological plane

of action in which a bent metrocard can be seen as a

forged instrument at all. Charles Goodwin (1994)

offers the example of how a specialized way of seeing

can reframe action in a trial setting: how a person

(shown in a video) prostrate on the ground receiving

blows from a police baton is framed in a courtroom

as displaying aggression warranting the blows accord-

ing to an expert police perspective. Because “common

sense”has no credentialed experts, such an expert

claim can (and often does) prevail without even being

contested. While swipers are denied access to the met-

rocard technology and the plane of action it putatively

constitutes, they are nonetheless vulnerable to a crim-

inal interpretation of their stock-in-trade which is

more or less impossible to challenge without that

access, and a competing expert interpretation.

Opaque technology and social vulnerability

Scholars of digital inequality have noted how a lack of

access to, and related sophistication with, Internet-

connected devices become incapacities to act effect-

ively in an increasingly online world (Dimaggio,

Hargittai, Celeste, and Shafer 2004; Robinson et al.

2015). The point also applies to decision-making sys-

tems in domains of consumer credit, insurance, hous-

ing, employment, and receipt of public benefits, which

produce consequential outputs through processes

4 N. MCCLAIN

concealed from, but pivotal for, the identities they

process, rate and score (Barocas and Selbst 2016;

Eubanks 2018). The rise of big data poses the add-

itional challenge of so-called “franken-algorithms”

which are ongoingly self-modifying, and increasingly

ungovernable even by programmers who set them in

motion (Smith 2018).

Other technologies also grow in prevalence in

everyday life without parallel growth of public under-

standing, much less direct engagement (Pasquale

2015). The criminalization of selling swipes illustrates

consequences from a lack of access to the black-boxed

interiors of a public technology ubiquitous in one

major city. It poses the problem that whatever lies

beneath a user interface can come to define action

taken through a technology, even while these internals

are not inspectable, not under democratic control, and

might even be subject to modification beneath an

unchanging human interface. Swipers, without access

to even a description of the inner workings of the

metrocard system, became vulnerable to the accus-

ation that their bent-by-hand metrocards are forgeries

on an electronic plane according, at least, to one

expert account of electronic interactions. But how?

Background and data

This article has roots in a series of technical and

sociological puzzles: In the mid-2000s, local New York

City press reported that swipers –then proliferating

in subway stations –were increasingly being arrested

under forgery laws, seeming to imply that swipers’

manipulation somehow altered a metrocard’s value.

Could imprecise human hands actually modify digital

data? How did police and prosecutors come to under-

stand bent metrocards as forgeries? Why was the issue

confronted with policing instead of with re-engineer-

ing a publicly-commissioned technology?

Obstacles I encountered in efforts to understand the

technology underlying selling swipes became important

clues towards the obscured nature of the technology as

itself an integral part of the research object. The fram-

ing of human acts through an opaque technology

opened up fresh questions about the distinction

between reflexive action on one hand and “technical”

action on the other, and just who has access to tech-

nical modes of interpretation and how those modes are

deployed in the criminal justice system.

In interviews with subway station employees who

sell metrocards (twenty in total; some also observed in

fieldwork

), informants were unable to explain the

technical mechanics of swipers’method. My

examination of their computer terminals, coding devi-

ces, and observation of operational and troubleshoot-

ing procedures illustrated how these interfaces

betrayed little about the underlying technology. These

workers’instructional materials

also hovered at the

surface of the work interface. Grey literature offered

important details of the fare system’s overall architec-

ture, but failed to clarify the precise technical proper-

ties of the system or of selling swipes.

The first –and only –official account of how a bent

metrocard interacts with the MTA’sfareprocessingsys-

tem made available for public dissemination was con-

tained in a criminal case, People v. Mattocks, argued

before the Court of Appeals in 2009 (see Chan 2009).

However, the technical account of selling swipes in the

Mattocks case seemed inadequate in two ways. First,

swipers –as I observed them in fieldwork –inevitably

encounter a pattern of error messages from the turnstile

computer as they execute their procedure, but those

error messages were unmentioned and unexplained in

the Mattocks account of the process. Second, the account

appeared at odds with a technical principle described as

essential to a metrocard-like fare system in a patent filed

by the metrocard’s developer, Cubic Automatic Revenue

Collection Group (Cubic, hereafter).

Questions about the Mattocks account, and about

the actual operation of the metrocard system were

eventually satisfied through the cultivation of a key

informant, one of few individuals with extensive direct

knowledge of the metrocard system. The informant, a

retired MTA official, had been the official keeper of

the metrocard transaction database and tasked with

monitoring records for anomalies indicative of fraud

or other problems from the system’s inception until

the late 2000s, and was one of a small coterie of MTA

personnel dispatched to give expert-witness testimony

in the trials of accused swipers. Our several interviews

and correspondence (spanning 2013 through 2016)

were instrumental to building the understanding of

the technology I share in these pages, which reflects

the fare control system as it was more or less stable

from about the year 2000 through at least the mid-

2010s (although it may have also been subject to tem-

porary changes not explored here).

My understanding of the metrocard system is

cross-validated against the work of self-identified

computer hackers who took up the challenge that the

metrocard cannot be read by commercial card-readers

(see Michael and Michael 2009), and worked in the

1990s and 2000s to reverse-engineer the metrocard

and its data (Balaclava 1994; blueski-mask and the

wrapper 1997; Redbird 2006; see also Harmon 1997).

THE INFORMATION SOCIETY 5

Their work culminated in the discovery that metrocard

data can be decoded when treated as sound, by passing

its magnetic strip across a tape-recorder head interfaced

with a PC sound card to produce “music”convertible to

binary (Redbird 2005a; b). The organization and mean-

ingofmuchofthedatawasthendiscernedthrough

one-to-one comparisons of metrocards with certain

known properties, such as time or location of its most

recent use, date of purchase, and so on.

The conceptual illustrations of the fare-processing

technology and the way a bent metrocard interacts with

it in Figures 1 and 3employ the black box as a visual

metaphor to also indicate important technical processes

that are hidden from users. Figure 2 extends that meta-

phor in blacking out processes which were neglected in

the Mattocks account, illustrating the understanding of

swipers’manipulation now embedded in law, and what

is missing from that understanding.

While I formed a critique of the Mattocks account

of how swipers deceive the fare processing system

with great help from my key informant, I must note

that my key informant was also the prosecution’s

expert witness in the Mattocks case. This may appear

paradoxical, but the two accounts emerged in two

very different contexts. Both cross-examination and a

sociological interview lead to interactively-produced

knowledge. Yet the procedural conventions of court

examination do not give witnesses much capacity to

offer answers outside of the scope of the questions

posed to them, or to otherwise shape interactional

agenda (Molotch and Boden 1985). As I describe

below, when the expert witness was questioned in the

Mattocks trial, he made a surface-level reference to an

interaction between a swiper’s manipulated metrocard

and a turnstile computer necessary to obtain an

unbought ride. The expert’s cross-examiners did not

seek –and likely did not think to seek –a deeper

understanding of that interaction, which remained

glossed even through the highest level of appeal in the

Mattocks case. My research interviews, by contrast,

did seek to understand that interaction, and found

that it complicates the prevailing understanding of

swipers’transgression at the same technical level at

which the Mattocks case was finally decided.

Conversations with staffpersons in the Clerks of

Court office in four subway-served boroughs of New

York City strongly suggested that felony forgery

arrests of swipers were predominantly processed in

Manhattan, and might be explained by the prosecutor-

ial philosophy of the District Attorney’s office in that

borough.

In turn, I interviewed the individual who

led the criminal court division of that office in the

early 2000s, and was involved in implementing new

policy governing the prosecution of swipers during

his tenure.

Official data on the arrests of swipers under forgery

charges come from the New York City Police

Department via a Freedom of Information request.

Other statistical data was furnished by the New York

State Department of Criminal Justice Services, and the

Manhattan D.A.’s office through direct correspondence.

The metrocard and the emergence of swipers

The MTA began to explore a transition to electronic

farecards in the early 1980s as a solution to chronic

turnstile jumping and widespread use of “slugs”

instead of an official token.

The MTA has long

treated every instance of turnstile-jumping or slug-use

as missing revenue, making fare evasion salient to any

budgetary shortfall. The agency had hoped that mag-

netic, refillable “stored value”farecards would under-

mine the market for slugs, cut transaction costs by

encouraging riders to buy multiple fares at once, and

allow the MTA to replace station clerks with farecard

vending machines.

The metrocard was phased-in over several years

beginning in the early 1990s, alongside new subway

turnstiles designed to make turnstile-jumping more diffi-

cult (Weidner 1996). Each turnstile houses a computer

for processing metrocards –users swipe their metro-

cards one or more times until their fares are successfully

processed and they are allowed entry. As this suite of

devices was phased in, official MTA statistics showed a

steep drop in the rate of fare evasion (Reddy, Kuhls,

and Lu 2011). Soon after, the MTA undertook to reduce

or eliminate stations staffing (Pierre-Pierre 1996), a goal

the agency has pursued ever since.

The MTA’s apparent success was threatened by the

end of the decade, however, when riders discovered a

major vulnerability in the metrocard technology: A

metrocard with no remaining value, scratched in a

precise spot, could be used over and over again to

enter the subway an unlimited number of times. The

vulnerability was soon broadcast citywide in the local

press (Coleman 1998; Rohde 1998). The press did not

discover, and the MTA did not reveal, that the vulner-

ability stemmed from a data remnant which remained

on the metrocard after all purchased fares have been

spent; the scratch obscured other data which indicated

the metrocard no long held valid fares. In response,

the MTA worked with the metrocard’s developer to

modify how metrocards are processed by turnstile

computers (Rutenberg 1998a). One modification had

6 N. MCCLAIN

turnstiles reject metrocards after two failed attempts

to read their data (Newman 1998), but the fix was

soon abandoned –it was not uncommon for users to

experience multiple failed reads with a perfectly valid

metrocard, and the modified turnstile software would

reject them, too (Rutenberg 1998b).

A subsequent revision of the turnstile fare-process-

ing software (the version at the center of this analysis)

Figure 1. How a turnstile computer processes metrocard fares, and the routine errors it encounters, produces, and corrects.

Notes. The metrocard’s magnetic stripe has two value-fields. The turnstile computer treats the lowest readable fare-total between

these two fields as the current value of the metrocard. Our examples begin with a metrocard worth three fares. The metrocard is

processed as it is passed through a slot in a turnstile containing a scan head, a write head and another scan head in physical suc-

cession. Swipe # 1 is ideal: The first scanner (A) determines the lowest readable fare total as the current value; then, a write head

(B) records a new, reduced fare total of two fares to the adjacent value field, then (C) the second scanner reads the freshly-written

data to verify the data written in process B. The computer will now allow the passenger entry, and the value field with the fare

total of three becomes obsolete but is left intact. Swipe # 2 leads to an error in which the computer cannot read a fare total on

either value field, and requires the user to try again. In Swipe # 3 the fare totals on both value fields are successfully read, and

the computer attempts to write a new fare total but the new fare total cannot be verified (C), possibly because either because the

verification scan failed, or the ‘write’(B) was corrupt, or both. The computer instructs the user to swipe again at that turnstile. The

turnstile computer then defaults to a mode, seen in Swipe #4, in which it simply re-attempts the write (B) and verify (C) processes;

re-starting with a read (A) risks deleting a user’s fare. Swipe # 5 is ideal, like Swipe #1. If the metrocard is swiped again (as in

swipe #6), it will be rejected because the lowest readable value field is now zero.

THE INFORMATION SOCIETY 7

was stabilized before the turn of the century.

The update indeed halted the acceptance of scratched

metrocards, but was vulnerable to a more elaborate

procedure of bending, swiping, unbending, and swip-

ing to produce an unpaid subway ride (Daly 1999).

The procedure still takes advantage of the data rem-

nant on a spent metrocard, but that remnant ends up

deleted, so the procedure will work only once per

spent metrocard.

Swipers became a common presence in many sub-

way stations in the early 2000s. Near many entrances,

swipers would gather discarded, spent metrocards,

skillfully bend them in a highly specific way, recruit

a customer, and swipe a metrocard through the turn-

stile reader several times until a message prompted

the rider to enter. In some instances, the MTA’s

de-staffing campaign left swipers as the only human

fare-sellers at a subway entrance, and swipers some-

times secured a total monopoly at the entrance by

jamming paper into adjacent metrocard vending

machines, disabling them temporarily (Luo 2004).

Each transaction earns the swiper a dollar or so. In

the argot of subway users, a swipe refers to any

attempt to use a metrocard at a turnstile, but only a

swiper will sell swipes. Swipers’technique is only

viable at subway turnstiles; on busses, where a mech-

anical apparatus (not a human hand) guides a metro-

card through its processing, a bent metrocard will be

rejected as either unreadable or valueless. (As we will

see, this is an important clue: Why would the

manipulation work at the subway turnstile but not at

the bus farebox if bending a spent metrocard, on its

own, makes it appear valid to an electronic reader?)

Why could not the MTA just check every

metrocard against master records and have turnstile

computers reject any metrocard which ought to be

depleted of fares? The reason is that turnstile com-

puters have no way of telling just what a given metro-

card’s value ought to be, because they process each

metrocard based solely on data on the card, without

Figure 2. The Mattocks account of bent metrocards as forged instruments.

Notes. Selling swipes begins with a metrocard in which the lowest of the two value-fields shows zero fares remaining, which will

always be rejected by the turnstile computer (illustrated in swipe #0). Cross-examination of an expert-witness in the 2005 People

v. Mattocks criminal trial characterized a bent metrocard (as in swipe #1) as capable of producing free entry into the subway

because the fare processing system “gives the benefit of the doubt”in case a metrocard had been somehow damaged. Testimony

also made reference to the existence, but not the substance, of subsequent steps necessary to achieve that outcome. Those steps

–not revealed in this account –suggest ambiguity in the legal classification of a bent metrocard as a forged instrument.

8 N. MCCLAIN

first checking records of that metrocard’s value in a

centralized database. In an arrangement designed long

before the present era of stable real-time networks,

each turnstile computer forwards records of its trans-

actions in batches to a station computer, which then

forwards batches of records on to a central main-

frame.

If a node or link in this network fails or is

merely sluggish, transaction records can continue to

accumulate (and fares can continue to be processed)

until the records can be forwarded on to the next

strata, allowing ongoing fare processing in spite of

that failure. One consequence of this loosely-coupled

arrangement, however, is that a rider might add or

use fares anywhere, any time in the vast subway and

bus network, but the record of each transaction may

not reach the mainframe for hours after each

occurred, and not necessarily in order of their occur-

rence, either (Reddy et al. 2011). Data written to the

metrocard itself is therefore treated as authoritative by

the turnstile computer for the purpose of providing or

denying entry.

In turn, a manipulated metrocard

without valid fares only comes into administrative

focus hours or even days after a turnstile computer

has already provided entry.

Without an obvious software solution, the MTA

turned to lobbying police, prosecutors and politicians

for a sterner stance towards swipers. The MTA also

initiated a poster campaign advising riders that selling

Figure 3. The hidden technical logic of swipers’practice.

Notes. Swipe # 0 involves a metrocard that no longer has any value. Before Swipe #1, the card is bent in a very specific way, mak-

ing concave the portion of its magnetic strip containing the value field with a fare total of zero. When it is swiped (Swipe #1),

that portion of magnetic strip is out of reach of the scanner head (A), preventing a successful read of the value field containing a

fare total of zero. However, the computer tenuously accepts that one fare it can read, and attempts to write a reduced fare total

(B) to the opposite value field, but is unable to verify if the fresh data has been correctly recorded (C) because the kink drew the

value field out of reach of the write and verify heads. The swiper is told to try again. Immediately after, the swiper un-kinks

the metrocard and holds it flat as it guides it through the mechanism (Swipe # 2). The computer, defaulted to only re-attempt the

write and verify scan steps, indeed writes (B) and verifies (C) that a fare total of zero is recorded to the formerly-unreachable value

field, completing essential steps to process the one fare the computer read in Swipe # 1. The pattern of Swipes #1 and # 2 will

require the user to swipe the metrocard once more (Swipe # 3) to zero out the remaining field indicating one fare, leaving two

value-fields with fare totals of zero, before the user is granted entry.

THE INFORMATION SOCIETY 9

swipes is illegal. It turned out, however, there was no

consensus over why it is illegal.

Policing and prosecuting swipers

Precisely what law did selling swipes violate? Charges

for trespassing, loitering, and criminal mischief were

each leveled against swipers, but these were indirect at

best. Theft of Service was a bad fit because swipers are

not generally the users of the stolen service. One major

arrest initiative accused swipers of petit larceny, a Class-

A misdemeanor, under the theory that their earnings

were stolen from the MTA. In one of the two instances

where the highest court in the state heard arguments

over appropriate charges for swipers, the Court of

Appeals rejected petit larceny. Partly concerned of set-

ting a precedent that would criminalize unfair or illegal

business competition, the court reasoned that property

mustfirstbelongtoanentitytobestolenfromit,which

was not the case for swipers’earnings.

In 2004, the state legislature responded to MTA

lobbying with a new law prohibiting “Unauthorized

Sale of Certain Transportation Services”(USCTS), a

Class B misdemeanor that carries up to three months’

incarceration.

In the months after the introduction

of the law, local media reported that police used the

charge agaisnt swipers only infrequently (Donohue

2005). However, USCTS was still used in more than

six thousand arrests through year 2018. Blacks and

Hispanics make up 93 percent of these arrestees (65

and 28 percent, respectively). Members of those

groups account for 97 percent of those convicted

under the law through 2012 (the most recent year for

which I have conviction data).

Accused swipers have faced a more common –and

more serious –arrest charges in the form of Forgery

in the 2

degree and Criminal Possession of a Forged

Instrument in the 2

degree, both “Class D”felonies,

carrying sentences of incarceration up to seven years.

According to NYPD data, its Transit Bureau made

more than eleven thousand such felony arrests from

2002 through 2014. As can be seen in Table 1, these

felony arrests also follow a powerfully-racialized skew:

Ninety two percent of these arrestees were black or

Hispanic, as well as overwhelmingly male and adult.

Almost two-thirds of the citywide felony arrests of

swipers made by the NYPD Transit Bureau were

processed in Manhattan. According to Rick Castello,

who led the Criminal Court Division of the

Manhattan District Attorney’s Office when felony

arrests for swipers first occurred, these charges were

welcomed for defendants with prior records. When

asked what occasioned the policy to allow prosecutions

of swipers under felony forgery laws, he responded:

Idon’t remember if this is one of those [charges] that

thepolicecametousbeforetheywerebombardingthe

complaint room with them, or if they bombarded, and

it eventually filtered up. I’mprettysureitwasthe

police, that they came up with this potential charge and

then our office decided how to handle it and agreed

that it was a clever but legally-sufficient charge.

In such instances, according to Castello, his office

would seek internal consistency by making a

“command decision”on the acceptability of a charge,

and then instruct junior prosecutors who otherwise

draft complaints more or less independently. The

approach was, according to Castello, consistent with

the Manhattan District Attorney’s general expectation

that the NYPD would deliver arrests under most-ser-

ious charges. Prosecutors, in turn, might exercise dis-

cretion to reduce the charges, or offer a reduced

charge as part of a plea deal. Castello recalled, “Most

of them [swipers with a felony arrest] would plead

guilty to something [ …] If you had a recidivist and

you really wanted to get him six months in jail, and

you charged him with a felony and they thought you

might be serious about indicting that case, they’d take

the six months.”

Table 1. Arrests of accused swipers under two forgery-related

felony charges by New York Police Department Transit

Bureau, 2002–2014.

Arrests Arrestees

Total (%) Total (%)

All arrests 11,142 (100)

Arrest Charge Race or Ethnicity

§170.254,865 (44) Asian/Pacific Islander 261 (2)

§170.10

†

6,277 (56) Black 6,715 (60)

Hispanic 3,571 (32)

Borough of Arrest White 541 (5)

Bronx 1,363 (12) Native American 13 (0)

Brooklyn 2,191 (20) Unknown 41 (0)

Manhattan 7,112 (64)

Queens 474 (4) Sex

Staten Island 2 (0) Female 388 (3)

Male 10,754 (97)

Year of Arrest

2002 467 (4) Age Distribution

2003 644 (6) 10-17 1,139 (10)

2004 1,887 (17) 18-24 3,218 (29)

2005 1,815 (16) 25-40 3,936 (35)

2006 1,504 (14) 41-59 2,765 (25)

2007 1,092 (10) 60þ84 (1)

2008 692 (6)

2009 479 (4)

2010 512 (5)

2011 369 (3)

2012 388 (3)

2013 516 (5)

2014 777 (7)

Criminal Possession of a Forged Instrument - 2nd Degree

†

Forgery - 2nd Degree

10 N. MCCLAIN

Some swipers were indeed tried under felony

charges.

The Manhattan District Attorney’s office

reports that it prosecuted 174 “bent metrocard”cases

as felony possession of a forged instrument between

2001 and 2018, with half of these in the span of 2004-

2008.

Instances include one swiper convicted of a

felony count for each of the 96 bent metrocards found

in his pockets, but allowed to serve 96 sentences of

two-to-four years concurrently.

Another defendant,

convicted of just two felony counts for his two bent

metrocards, was sentenced to serve two consecutive

prison terms of 2-6 years each, although a higher

court later halved the penalty, noting that defendant’s

diagnosis of schizophrenia had been ignored at

sentencing.

The relative small proportion of felony prosecu-

tions suggests that the main purpose of the felony

charge is (as Castello puts it) the “great leverage”it

offers in plea bargaining –leading many accused

swipers to plead guilty to lesser charges rather than

risk a felony prosecution. Even if no felony conviction

or prison time resulted from such arrests, they are

nonetheless consequential for arrestees. A full discus-

sion is not possible in these pages, but we can defer

to Kohler-Hausmann’s(2013,2018) analyses of the

ways New York’s criminal justice system asserts con-

trol even without conviction by “managing”defend-

ants.

Elements include periods in which defendants

are “marked”for added consequences in any future

criminal justice encounter (through designations like

“transit recidivist”), and marked with potential

obstacles to jobs and housing through searchable pub-

lic data on open criminal cases. The courts further

entangle defendants by mandating performance of a

“program”, of service or therapeutic activity, to move

their cases towards dismissal, all the while requiring

defendants to re-orient their lives around the vagaries

of court schedules and procedures, and inevit-

able delays.

Criminalizing technical agency inside the

metrocard system

Criminal forgery allegations against swipers often did

little to specify how a bent metrocard is a forgery, for

example, by saying only that a defendant possessed a

metrocard “bent in a manner that would allow illegal

entry into the subway system.”However, more

evolved language located the criminality of the bent

metrocard in its relationship with a computer, on a

technical plane not observable by the accused, such as

by alleging that the bend “obliterates the encoded data

and alters the fare of the card as read by the turnstile

scanner.”

As we will see, such claims oversimplify

the operations of the fare-processing system and how

a bent metrocard can interact with it (explained in

Figures 1 and 3). Yet legal proceedings held to adjudi-

cate just what swipers “do”in a technical and criminal

sense embedded that oversimplification in the fabric

of law, stabilizing the legal status of bent metrocards,

along with the social impacts of that status.

Catching swipers in the black box in People

v. Mattocks

The treatment of bent metrocards as forgeries was

rejected by at least one court which reasoned that

because the MTA does not sell visibly-bent metro-

cards, a bent metrocard is not “falsely altered”because

(as that term is defined in a key statute

) a bent one

does not “purport to be in all respects an authentic

creation.”

This way of thinking about technology –

a“technology frame”(Orlikowski and Gash 1994)–

reflects what is observable to human actors. Its appli-

cation to the metrocard did not long survive, however,

and was soon replaced, through People v. Mattocks, by

a technology frame empathetic to what the turnstile

computer purportedly perceives.

In a trial, which began in 2005, a Manhattan jury

convicted defendant Jonathan Mattocks of Criminal

Possession of a Forged Instrument –2

degree.

this criminal trial, the MTA Director of Metrocard

Security, James Eastman, was designated an expert in

“metrocard security”. He was asked to explain how a

bent metrocard interacts with a turnstile computer

and to indicate which of the defendant’s had the

proper data configuration to enter the subway without

paying, based on his examinations performed through

MTA equipment. The testimony and its interpretation

later became the authoritative account of what turn-

stile computers see when presented with bent metro-

cards, and thereby was key to stabilizing bent

metrocards as forgeries under the law.

The expert witness account

In Mattocks’2005 criminal trial, Eastman revealed

fundamental elements of the metrocard technology

(consistent with Figure 1): The metrocard’s magnetic

stripe contains not one but two data fields designated

to hold fare data. The totals written to these fields are

not equal –one total is always one fare higher than

the other. When swiped at a turnstile, its computer

always treats the lowest fare total as the correct,

THE INFORMATION SOCIETY 11

current value, and writes the new, reduced fare total

to the other value field (see Figure 1). The two phys-

ical value fields thus take turns possessing the current

fare total of the metrocard. Following this alternating

logic, a metrocard used until no fares remain will

have a value-field that indicates a fare total of zero

and will be rejected at the turnstile. Yet, the other

value-field will still show a fare total of 1. That linger-

ing sum is what makes swipers’manipula-

tion possible.

Questions posed to Eastman produced the follow-

ing account of how a swiper’s bent metrocard inter-

acts with the fare processing system (summarized in

Figure 2), which we must take care to examine closely:

A swiper’s bend to a spent metrocard “damages”the

value-field with a fare total of zero. When swiped at a

turnstile, its computer can then only read the opposite

field, which still shows one fare, and will accept that

one fare as valid, and allow the customer entry.

Eastman described the acceptance of just one readable

fare total as a “backup”feature which gives the “benefit

of the doubt”in case the card had been damaged.

When Eastman was asked if certain of the defend-

ant’s bent metrocards “would work”to obtain unpaid

rides if swiped at a turnstile, Eastman stipulated “yes,

but you’d have to swipe them twice”. This was a crit-

ical answer but it passed without notice; if any party

had interrogated why this second step was necessary

and what is accomplished through it, a more nuanced

account of the metrocard system might have emerged,

suggesting a more complex relationship between a

bent metrocard and forgery laws.

The Court of Appeals interprets bent metrocards

Attorneys working on Mattocks’behalf challenged the

treatment of a bent metrocard as a forged instrument,

as well as the use of felony forgery laws to police and

prosecute an activity already prohibited under the

misdemeanor law specifically and directly passed

against it (Unauthorized Sale of Certain

Transportation Services). The argument eventually

won a hearing before the Court of Appeals, the high-

est court in the state. Arguments on matters of law

were to be based on existing transcripts and materials

produced for Mattocks’criminal case; the existing

technical account of selling swipes thus became stabi-

lized and undisputable for the purposes of the appeal.

The Court eventually rendered its decision drawing

from that expert account, and the ways the account

was framed in the initial criminal trial and intermedi-

ate proceedings. What is most remarkable about the

Court of Appeals case is that it sought to specify the

legal status of a bent metrocard within a technological

context neither the litigants nor the court actually

understood clearly, grasping for metaphors perhaps

appropriate to their understandings of the metrocard

system, but inapt for its actual nuances.

According to New York State law, forgery is com-

mitted when someone “falsely alters”a written instru-

ment so that it then “appears or purports to be in all

respects an authentic creation or fully authorized by

its ostensible maker or drawer”(emphasis added).

In oral arguments before the Court of Appeals,

whether or not a swipers’metrocard appears “in all

respects an authentic creation”–was the central ques-

tion. Mattocks’attorney (from the Office of the

Appellate Defender) offered a simplifying metaphor: a

metrocard with only one readable value-field is akin

to a check with the signature line ripped off, and is

thus incomplete and invalid.

Just because the MTA

programmed its turnstiles to accept incomplete metro-

cards, he argued, does not mean that a metrocard

with missing data is a forgery.

An Assistant District Attorney (ADA) from

Manhattan advocated for the felony forgery charge.

Contradicting Mattocks’attorney, she claimed that “it’s

absolutely not true the turnstile reader accepts –can

detect that the card is invalid, actually detects that its

invalid or unreadable and allows someone to enter any-

way. That’snottrue.”

As she argued, even if the fare-

processing system’s capacity to detect a false instrument

is inadequate, any unrecognized false instrument is by

definition “complete”to the system and is thus a for-

gery. Exemplifying the technological frame sympathetic

to an imagined perspective of technology, the ADA

claimed that to rule against the forgery charge because

MTA computers cannot detect the invalidity of a bent

metrocard would be “blaming the victim”.

The Court of Appeals sided with the Manhattan

District Attorney in ruling that a bent metrocard is

indeed a forged instrument to the eye of the relevant

audience, that of a turnstile computer. Yet the court

did not find grounds for forgery charges in the undis-

puted and observable fact that a properly bent metro-

card can be used to produce unpaid subway entry.

Rather, the court looked inside the technology, based

on a reading of the 2005 expert-witness testimony.

The Court’s decision described how, according to

that expert account, a precise bend “can obliterate

that information [zero fare total - au], so that, when

swiped, the computer is unable to detect that the met-

rocard is worthless …the turnstile computer will

read the other field containing the …backup

12 N. MCCLAIN

information, which gives the user of the card free

entry into the subway.”The decision elaborated:

[T]he damage to the MetroCards thwarted the usual

computer processing of the information contained on

the magnetic strips. By bending the MetroCards,

defendant successfully destroyed the zero-fare

information encoded on one of the fields in the

magnetic strips and was able to acquire free rides on

what were worthless MetroCards. Thus, defendants

misused the “benefit of the doubt”system by

intentionally making the valueless MetroCards

purport to be authentic instruments.

The Mattocks ruling remains stable case law. Yet,

when we examine how bent metrocards interact with

the fare-processing system at a technical level, we can

find considerable ambiguity in the legal status of a bent

metrocard. Moreover, we discover how technology can

be misunderstood in ways that may wrongly attribute

criminality to actions taken with and through them.

How a bent metrocard interacts with the fare-

processing system

As the Court of Appeals articulated the process, if a

turnstile computer can only read one value-field (con-

taining one fare) on a metrocard, the computer will

accept that fare and allow the passenger to enter. This

is not correct. A turnstile computer will not, under

any circumstance, process any metrocard as valid for

a fare unless it has successfully: (1) read a fare on one

value field, (2) written a zero to the other value field,

and (3) read the freshly-written zero. Paradoxically,

the second and third of these steps cannot be com-

pleted with a bent metrocard, however, because the

portion of the magnetic strip containing the relevant

value field is not in full, flush contact with the write

or verify heads set in the turnstile. Thus, the process

must be accomplished across multiple steps: In the

first, the turnstile computer reads one fare. In the

second, when the swiper unkinks the metrocard and

holds it rigid, the turnstile computer writes a zero to

the other value field before then reading it for verifi-

cation. The data that the turnstile computer writes

and then verifies in the second and third steps makes

it treat the data it read in the first as legitimate –as

having been valid in retrospect. The fare-processing

system will record that a fare has been used at

this point.

Understanding this process requires a brief tour of

the metrocard’s engineering. The slot set into subway

turnstile for processing metrocards contains a scan

head, a write head, and a second scan head (for veri-

fying fresh data) in linear succession so each can

perform its function as the metrocard strip is trans-

ported across it in one swipe through the slot. When

the MTA began to consider a farecard loaded with

multiple fares in the 1980s, existing products entailed

farecards where a new fare total was written over (and

thus deleted) a prior fare total each time the card was

used. A motorized mechanism would accept the card

and transport the card across the scan and write heads

properly, at correct speed, alignment and uninter-

rupted contact. These conditions are necessary to

allow the scan head to read the strip’s magnetic par-

ticles as coherent data, and for the write head to prop-

erly distribute those particles on the strip in a way

that can be read as data once again. While these

mechanical transport systems helped ensure those

conditions, they often jammed and trapped contami-

nants, requiring labor-intensive interventions.

Moreover, they slowed passenger-throughput below

levels the MTA deemed necessary for subway entran-

ces.

The MTA thus sought a system in which the

user would swipe the card through its processing by

hand, but this posed a new challenge: if the stripe is

not swiped across the write head under proper condi-

tions, what is written will not be readable as a data

once again –it is corrupt, and the user’s value has

been deleted entirely (Cekander 1994).

The metrocard’s developer, Cubic, did not find a

way to prevent the chronic corruption of fare totals

due to bad “writes”, but did find a way to tolerate

that corruption. Cubic’s solution (outlined in a patent;

see Aubrey 1992) involved using magnetic strips with

two value-fields rather than one, and turnstile com-

puter software which would always treat the lowest

fare total between those fields as the current, proper

value of the metrocard. Each time the turnstile proc-

esses a fare, the current fare total is left untouched.

That total simply becomes obsolete once a lower value

is written to (and verified on) the opposite value-field.

Corruption would still be endemic to this arrange-

ment, but the errors are correctable by making users

swipe again until a proper write is verified; users

would be made to compensate for a technical short-

coming. The metrocard system was designed to toler-

ate metrocards with just one readable value-field

because the metrocard system so often creates them.

As illustrated in Fig. 3, swipers’bend is placed right

where the value field holding a fare total of zero sits,

making it impossible to read from or write to that

region of the strip because it is concave and beyond

reach of the heads.

The turnstile computer will treat

the one fare it can read as valid, but only in retrospect

if it is subsequently able to write and then verify a

THE INFORMATION SOCIETY 13

new, reduced fare-total of zero to the other value-

field. However, that verification fails because (as we

humans know) the value-field is unreachable by the

write head and scan head, meaning expected data will

not be found by the verification scanner.

A failed metrocard verification is normal and

endemic to the system, and can be a result of a cor-

rupt write, a misread for verification, or both (e.g.

Figure 1, swipe 3, process C). The turnstile computer

responds by prompting the user to “swipe again here”.

When the user complies, the computer defaults to a

mode in which it only re-attempts the write and verify

steps. This configuration seeks to avoid a major pitfall:

If the computer repeated the initial scan step (e.g.

Figure 1, swipe 1, process A), it might end up treating

a fresh, properly-written fare total it failed to verify as

the current total, and then debit a second fare without

having provided entry for the first.

As the swiper passes the card through the slot a

second time, they now hold the metrocard rigid,

which counteracts the kink in the magnetic strip.

Once flat, the turnstile computer can now write and

then verify that fare total of zero on the relevant

value-field. Verifying that zero closes the loop of steps

necessary to process a fare, and it is at this point that

the metrocard system records that a fare has been

processed. The passenger will be allowed entry as

soon as the user completes a final step, to swipe yet

again so the computer can destroy the one fare once

again lingering on a value field, thus leaving the met-

rocard without any fare totals at all.

Finding ambiguities in the black box of

the metrocard

Even if we accept that defendants are criminally culp-

able for the hidden machinations of the technologies

they use, we can find ambiguities in the interpretation

of bent metrocards as forgeries that have been beyond

reach of defendants and have never been presented

to courts.

State forgery statutes describe a forged document

as one that has been purposefully altered and which

then purports to be in all respects authorized by its

maker.

While swipers’bend makes it impossible for

the turnstile computer to read a portion of the data,

the data is itself unaltered (and would be legible again

if the card were unkinked). If the turnstile computer

“sees”through contact, it seems apt to say that bend-

ing the metrocard is more like turning out the lights

when a cashier is examining our currency than it is

like handing over a fake dollar bill. The circumstances

of perception are modified, but the instrument is not.

The multi-step and iterative nature of swipers’

manipulation is a second site of ambiguity. To achieve

unpaid entry, a bent metrocard must be read in three

different states (first, as only partly readable, and

second, as containing a fresh zero in the formerly-

unreadable field, and lastly, as containing no fares on

either value-field). The validity of the first fare-total

hinges on the successful recording of the second;

without the second step, even a properly bent metro-

card will always be recognized as invalid and be

rejected by the turnstile computer. The interaction

will not even generate a record. It is only when the

turnstile computer writes and verifies a fare-total of

zero to the opposite field that the prior total of one is

treated as legitimate, and a passenger is given entry.

However, the data configuration which permits

unpaid entry exists for only a split second, just before

the swiper is required to pass the card through the

turnstile slot a final time, when any lingering repre-

sentation of value is deleted before the passenger is

allowed entry (see Figure 3). Moreover, it is not a

bent metrocard which is able to produce this out-

come, but an unbent one.

The incapacity of a turnstile computer to accept a

bent metrocard, and the issue of unbending, were not

mentioned at any stage in the Mattocks trial. It is highly

unlikely that any prosecutions of swipers have taken

into account these problems. The evidence held against

swipers are bent metrocards with the potential to pro-

duce unpaid entry but which cannot under any circum-

stance produce unpaid entry or be considered complete

by the turnstile computer without further steps.

It is not clear if these ambiguities rooted in an

improved understanding of metrocard technology

could have resulted in a different conclusion in

Mattocks, much less halted the search for severe conse-

quences for selling swipes. It is certain, however, that

accused swipers were denied the potential to develop

important arguments based on these ambiguities.

Discussion and conclusion: Interpreting

technology, interpreting criminality

When the MTA discovered that users could ride free

using spent metrocards, the agency did make some

internal attempts to halt the practice, including a soft-

ware modification which limited the number of times

the manipulation could be performed on a given met-

rocard. Yet, as selling swipes began to proliferate,

swipers’behavior was framed as the source of the

14 N. MCCLAIN

problem, rather than a flawed or misbehaving technol-

ogy. Notwithstanding initial tinkering, the metrocard

technology was taken as an ungovernable given, like

the weather, over which there is no human control,

and about which there can be no politics, oversight or

public intervention (see Jasanoff 2016). Indeed, the

prosecutor advocating that bent metrocards are for-

geries argued convincingly to the New York Court of

Appeals to take the design of the metrocard verifica-

tion system for granted, and treat the (putative)

parameters of computer perception as given, even

when those alleged parameters can lead to a human

action being given a felonious status. Presumably, we

could all be forgers if we inserted a $1 bill into a

vending machine marked to only accept $5 bills, but

the vending machine treated our one dollar as five

dollars upon verifying that yes, the currency is green.

If criminality can be an artifact of technological

design, and the black box of a technical artifact masks

the very human decisions entailed in that design, we

discover that, through technology, human decisions can

have un-confronted, autonomous impact –this is one

foundational insight of STS research. The engineers of

the metrocard system are not unique in failing to

anticipate alternative uses of technology. Indeed, in a

second basic observation in STS, they are

“interpretively flexible”(e.g. Leonardi 2009;Oudshoorn

and Pinch 2003). However, notwithstanding examples

like consumer-safety regulations, we lack clear political,

legal or other mechanisms to problematize the engin-

eering of things, much less to intervene in it. The opa-

city –if not proprietary secrecy –of many technologies

presents a barrier: what can we contest or regulate

when we do not know what to contest or regulate?

Artifacts may have politics, as we have long known

(Winner 1986) but they need to be noticed in order for

an intentional human politics to emerge over them.

The stealth work of opaque infrastructure stands to

be more insidious because infrastructure is inescapable

without also giving up all to which it is central (Star

1999). Opting out of AC electric power means also

giving up toasters and television, just as opting out of

the fare-processing system of your city’s mass transit

means giving up a vital mode of mobility. Another

layer is added to this picture in the case of technolo-

gies deployed by arms of the state in which, as Joh

(2017a) notes of the engineering of police body cam-

eras, “design is policy”. The mechanics of the technol-

ogy itself has the impacts of public policy but is too-

often granted the neutrality of ungovernable and

undisputable things.

The criminalization of bent metrocards offers a fur-

ther twist on this already-fraught problem. Even in a

court proceeding which sought to examine a putative

relationship between a technology and a criminal act,

an inadequate (and quite arguably wrong) account of

that technology was made central to a legal decision.

It is not only artifacts that have politics; accounts of

those artifacts have politics as well.

Accounts of technology rise in significance when

the opacity of a system is not merely the outcome of

cognitive or intellectual neglect, but of active secrecy.

Secrecy grants monopolies of knowledge and access to

those with interests in specific accounts of that tech-

nology in a way that is difficult to dispute. As we saw

in the Mattocks case, a singular account from an

expert witness was glossed and construed to describe

not only the status of a bent metrocard, but also to

conjure the technological context in which that bent

card could have that status. The 2004 misdemeanor

law which targeted the selling swipes, against the

unauthorized sale of transportation services, at least

described the activity in a way accessible to swipers

who might intuit that they were indeed selling trans-

port services without authorization. The use of felony

forgery charges, by contrast, alleges a crime its perpe-

trators themselves cannot witness because it unfolds

within a domain to which they have no legitimate

access. While certain allegations against an accused

swiper might be contestable, the fact of a bent metro-

card being a case of forgery is hardly contestable,

thanks to the secrecy of the metrocard system, and

domination of related expertise and even basic infor-

mation by those invested in that secrecy.

Surely, forms of bias are revealed in the social selec-

tion of selling swipes as a problem warranting highly

punitive arrest policies, which yielded arrests of (almost

exclusively) African American and Hispanic men. They

are also revealed in efforts by prosecutors to defend

these arrests and prosecutions up to the highest levels.

However, those efforts found traction in the putative

interiors of technology. Labeling technical agency may

be a new vector of inequality in criminal justice, espe-

cially when contesting a criminal interpretation of tech-

nical agency requires exactly the resources marginalized

defendants tend to lack, and the expertise and institu-

tional access almost everybody lacks.

Within every black box of technology is many more.

We must worry that each might contain –or be said to

contain –a technical rationale to reframe human action.

Those who cannot anticipate those re-framings are espe-

cially vulnerable to them. Swipers offer one example. A

contrasting case is that of the computer hackers whose

THE INFORMATION SOCIETY 15

capable probing uncovered both how to duplicate valid

metrocards, and how these duplicates would be discov-

ered by MTA computers in fairly short order.

Their

sophistication with technology helped them to anticipate

duplications as a risky –if not futile –forgery.

The

value of this kind of sophistication is likely to grow

when the internal operations of technical systems are

cited to frame just what users do (in a technical sense)

when they do this or that (in a way that is self-aware).

While the practice of selling swipes is much in

decline,

the Mattocks precedent stands ready to gov-

ern analogous practices with other digital media. The

linkage between a bent metrocard and forgery can now

be invoked without being revisited, a black box of its

own. The flawed understanding on which that linkage

is based, however, illustrates the problem of opaque

technology deployed to frame the actions of unwitting

users. No data suggests that any actor relevant to the

Mattocks case put forth what they knew to be a flawed

understanding of the technology, but the incomplete

account of metrocard technology which emerged in the

case is perhaps an outcome of a kind of “structural

secrecy”(Vaughan 1996), achieved when an expert wit-

ness without legal training answered questions posed

by lawyers unable to identify ambiguities in the picture

of the technology which emerged from his answers. Yet

even if the Mattocks ruling were based on a rigorous

understanding of the metrocard system, now that the

legal status of bent metrocards has been stabilized,

there is no nexus to update that legal status concordant

with updates of the internal, hidden aspects of the tech-

nology. A hacker’s 2016 attempt to decode the metro-

card, for example, suggests some modifications to the

way data was previously written to the metrocard,

while users’experience has been unchanged.

Whose

job would it be to inform courts, defendants, and pros-

ecutors that the metrocard now works differently, and

the criminal status of bent metrocards warrants recon-

sideration? With no institutionalized means of evaluat-

ing and updating existing accounts of technology,

definitions of human action can be trapped within

those accounts, with little chance of escape.

Notes

1. Hereafter, I avoid the capitalization of “MetroCard”to

diminish distraction and to reject treatment of public

infrastructure as a private or branded product.

2. This paper describes selling swipes of what are known

as “pay-per-ride”metrocards. Another less-common

type entails selling the use of “unlimited ride”

metrocards, generally policed under civil rules of

mass-transit conduct (see 21 CRR-NY 1050.4). The

latter practice requires no manipulation, but needs an

expensive unlimited ride card, which entails a delay

between uses.

3. A 2002 juvenile case, In the Matter of D.U. [Family

Court, Queens County, Jamaica, New York, 192 Misc.

2d 601 (N.Y. Misc. 2002). https://casetext.com/case/in-

the-matter-of-du] appears to mark the first

prosecution of an accused swiper under forgery laws.

4. The secrecy survived two public inquiries into chronic

errors and “double debits”experienced by users. See

reports by New York City’s Public Advocate

(Sheppard, Chernobilsky and Mintz-Roth 2005), and

an oversight committee of the New York City Council

(Feerick and Gupta 1998), both on file with author.

5. New York City Administrative Code § 14-172 requires

the NYPD to post data on enforcement of Penal Law

§ 165.15(3) and on TAB/NOV summonses. See New

York City Council (2018).

6. In 2017, the Manhattan District Attorney’s Office

announced it would no longer prosecute standalone

Theft of Service arrests (McKinley 2017) –the only of

five city jurisdictions to adopt this policy. However, the

policy does not apply to defendants deemed “threats to

public safety”. Moreover, the NYPD continues to make

standalone TOS arrests in the Manhattan jurisdiction

(1,611 in 2018, according to NYC OpenData, 2019),

stemming from enforcement which may lead to the

discovery of additional charges, being taken into police

custody, and other consequences.

7. Phone discussion, Jarret Hova, Director of Policy, and

Counsel, Office of the Public Advocate, City of New

York, November 1, 2018.

8. See People v. Robinson.

9. Interviews stem from a broader project which also

included extensive ethnographic observation in the

subway system.

10. E.g. MTA New York City Transit (1999), on file with

the author.

11. Specifically, a farecard system patent assigned to Cubic

(Aubrey 1992) operates on the principle that a

turnstile computer would never, in any one “swipe”,

read a fare total from, and write a new fare total to,

the same value-field. The Mattocks account did not

explain how these processes were accomplished

separately, glossing the steps between a “read”and a

metrocard no longer able to provide an unpaid entry.

12. Each of the five counties of New York City has an

elected District Attorney, leading to variation across

jurisdictions.

13. Freedom of Information Law request to New York

Police Department #2013-0212, issued December 2012,

rejected twice, reopened twice on appeal, and partly

fulfilled in November 2015 as statistics on Transit

Bureau arrests of swipers for six forgery-related

charges from 2002 through 2014.

14. One industrial-scale slug-making enterprise was broken

up following a police raid on a workshop containing

hundreds of thousands of fake tokens. Its leader, a

former graduate student in sociology at the New School

(Dwyer 2004; Sullivan 1991) was charged with Criminal

Possession of a Forged Instrument in the 2

Degree, the

same felony leveraged against swipers years later.

15. See Prichard (1991), on file with author.

16 N. MCCLAIN

16. E.g. Metropolitan Transportation Authority (2011), on

file with author.

17. While the MTA is able to instruct turnstile computers

to reject certain metrocards (for example, ones that

were reported as stolen and were traceable through a

purchase record) by serial numbers, this “bad list”is

updated on each turnstile computer just once per day,

and is not usable for a real-time intervention.

18. See People v. Hightower. Oral arguments of November

14, 2011 available as video at https://www.nycourts.

gov/ctapps/arguments/2011/Nov11/Nov11_OA.htm

(accessed on March 13, 2019); see comments of Hon.

R. Smith at 12:20.

19. Penal Law §165.16. Also see the 2004 Executive

Memorandum in support of the bill describing its

purpose, excerpted in Part III of the uncorrected

Mattocks (2009) decision, at https://www.law.cornell.edu/

nyctap/I09_0062.htm (accessed on March 13, 2019).

20. Data refer to arrests in which Penal Law §165.16 is the

most serious, or “top”charge. Data come from New York

State Division of Criminal Justice Services Computerized

Criminal History Database, on file with author.

21. Interview conducted via telephone, January 19, 2013.

22. These cases were generally charged under Criminal

Possession of a Forged Instrument in the 2

Degree,

rather than Forgery in the 2

Degree, very likely

because bent metrocards act as physical evidence, and a

defendant cannot be convicted of both crimes for the

same instrument (see New York Penal Law § 170.35).

23. Email correspondence between the author and Danny

Frost, Director of Communication, Manhattan D.A.’s

Office, Jan 22 and Jan 29, 2019.

24. People v. Curry.

25. People v. Solomon.

26. While Kohler-Hausmann’s(2013,2018) research

follows trajectories of misdemeanor arrests, those

trajectories illustrate, if in broad outlines, felony

arrests which do not result in felony charges.

27. Compare People v. Richardson with People v. Santos.

28. NY Penal Law § 170.00(5, 6).

29. People v. Lopez, cited in Goell (2006/7).

30. People v. Mattocks 2005.

31. New York State Penal Law § 170.00 (6); see

also 170.10(3).

32. People v. Mattocks 2009. Oral arguments available

as video via request from Albany Law School

Government Law Center, http://www.nycourts.gov/

ctapps/forms/VideoForm.pdf (accessed on March 22,

2019), and on file with author.

33. Mattocks’attorney erred. The computer establishes the

presumptive value of a metrocard based on the lowest

fare total even when there is only one legible fare total

on the card.

34. As we shall see, the ADA’s claim was substantially

wrong. The turnstile computer always detects an

instance in which it cannot read a value field.

35. See Jeffreys (1984), on file with author.

36. To process a fare, the turnstile computer must also be able

to read the card’serial number, and other data, which

remains legible to the scanner despite the swiper’sbend.

37. This is precisely the reason why when users, after

experiencing errors at one turnstile try their luck at

another and sometimes find that their metrocards

have somehow lost value (see Blair 1998).

38. New York forgery laws define a “falsely altered”instrument

as one changed “by means of erasure, obliteration, deletion,

insertion of new matter, transposition of matter or any

other manner”(§170.090(6)).

39. Middle-class populations are better insulated from

such a radical labeling of mundane malfeasance.

Compare, for example, the case of selling swipes with

the use of plastic covers that obscure automotive license

plates to overhead cameras used to bill motorists for

tolls. Reportedly popular on the personal vehicles of

police officers (Cohen and Marsh 2017), their use

remains in the realm of civil violation. While they

prevent the collection of tolls, the perspective of the

deceived camera (or some similar technical reading) has

not been made to matter as a version of fraud or forgery.

40. The hackers noted that multiple metrocards bearing

the same serial number would lead to great imbalances

between fares used and fares purchased. Once noticed

by the central mainframe, that serial number will be

placed on a “bad list”updated to every turnstile

computer each day. See video-recorded presentation of

Red Balaclava (1997) especially at 15:30-25:00.

41. In another example, when sophisticated media file

sharers migrated to file swarming (or torrents) in

which users download small fragments of a media file

from many sources instead of whole files from just

one, the distributed nature of the sources and

recipients made the legal interpretation of their

activity more ambiguous, more difficult to investigate,

and largely shielded those users from mass litigation at

least for a period of years (Caraway 2011).

42. For example, we learn that sending an email to a

neighbor becomes “overseas communication”and can

therefore be read by the U.S. National Security

Administration when our message is, for whatever

reason, routed through overseas servers (Goldberg 2017).

43. The market for swipers’services was diminished in the

late 2009 when new emergency gates at subway

entrances offered nonpaying passengers a ready way in

(Lysiak and Goldiner 2009). Soon after, used

metrocards, the fundamental stock-in-trade of swipers,

became scarce thanks to a $1 surcharge added to the

purchase of new metrocards.

44. Hacker W. Woodruff found his (2016) reading of the

metrocard consistent with the prior discoveries of

Redbird (2005a), but could not discern the same

sentinels which mark the beginning and end of

data sequences.

Acknowledgements

The author expresses sincere gratitude to Harvey Molotch,

Ashley Mears, Owen Whooley and Christena Nippert-Eng

for critical feedback across the duration of this project. The

author is thankful for the insight or help of Daniel Bodah,

Diana Graizbord, Marie Hicks, Harry G. Levine, Ben

Merriman, participants in the Northwestern Ethnography

Workshop, and co-panelists and audiences at the meetings

of the Eastern Sociological Association, the American

THE INFORMATION SOCIETY 17

Sociological Association and the Society for Social Studies

of Science. Daniya Kamran-Morley designed the Figures

with the author’s input and is responsible for all of their

virtues and none of their flaws.

Funding

This work was partly supported under grants from the

National Science Foundation (SES-0542777) and the Harry

Frank Guggenheim Foundation.

References

Angwin, J., J. Larson, S. Mattu, and L. Kirchner. 2016.

Machine bias: Risk assessments in criminal sentencing.

ProPublica, May 23. Available at: https://www.propublica.

org/ (accessed on June 6, 2018).

Aubrey, K. R. 1992. Fare card read-writer which overwrites

oldest or invalid data. U.S. Patent 5,298,726 filed January

7, 1992, and issued March 29, 1994. Available at: https://

patents.google.com/patent/US5298726 (accessed on June

13, 2019).

Balaclava, R. 1994. The New York City Metrocard. Paper

presented at HOPE: Hackers on Planet Earth conference.

New York, August.<

Caught inside the black box: Criminalization, opaque technology, and the New York subway MetroCard

Abstract and Figures