In the prelude to
How We Became Posthuman, N. Katherine Hayles describes two different variations on the
Turing test.
The most famous one, the one many of us may know, involves a person
using some kind of computerized chat interface to talk to either a
computer, or a human in another room. It is the task of the test subject
to determine, from conversation, whether their interlocutor is human or
machine. Passing the Turing test has long been seen as one of the holy
grails of artificial intelligence. When computers are able to pass as
human, the argument goes, one of the distinctions between humans and
computers dissolves.
Hayles also describes another Turing test.
This one starts in the same way as the previous, with a human
participant talking to someone in another room through a computerized
chat interface. But in this one, the discussion partner on the other
side is definitely human. The goal of the participant is, instead, to
determine whether their conversation partner is male or female. If this
second Turing test has similar stakes to the first, Hayles asks, does an
ability to fool the participant negate the gender of the human on the
other side?
One of the crucial questions raised by the gender
Turing test, to my mind, is about the role of rigid, socially defined
gender binaries. The test is predicated on an understanding that there
are two genders, male and female, and that they each behave in a certain
way. If we choose not to take this idea for granted, and instead decide
that there is a vast spectrum of behaviour and appearance running from
that which completely and stereotypically matches a gender, to that
which is entirely opposed, the gendered Turing Test becomes impossible.
How do we decide, from a textual discussion, what gender someone is if
we do not require all people to adhere to a strict social script about
their gender?
That second Turing test does do some valuable work
for us: it highlights the importance of the visual in making judgments
about gender. Many people feel entitled to understand another person's
gender, based on their appearance. We sometimes hear the distressed
question, whispered behind hands, "Is that a man or a woman?" In a space
where those visual cues are not required, where we can present
ourselves textually, or in ambiguous photos, appearance--that popular
tool for determining gender--is not available.
There's a huge
spectrum of ways that gender is represented, discussed, made an issue,
or turned into infrastructure on the internet. Different platforms
construct gender as an issue of varying importance. In some software
development communities (on mailing lists, in IRC), it's generally
considered impolite to ask people for personal details that they're not
readily volunteering. A comment raised by this is the idea that many
women don't get noticed or counted because they don't mention or make
obvious their gender, because the default or un-gendered stated is
considered to be male. If someone does not make it explicitly clear that
she is a woman, she is assumed to be a man. If someone makes it
explicitly clear that they are something other than simply a woman or
man, it starts a discussion, which may or may not be welcome to the
person who has accidentally instigated it. So on one side of this
spectrum, there's communities where disclosing gender is not
structurally necessary and speculation is entirely a private activity by
individuals; on the other side of the spectrum, there are platforms
like Facebook, where including a gender is a required activity in
profile building, and where the default is man or woman, unless you
choose to start writing in an answer, and then there's an authorized
list of possibilities. Gender is built into the bedrock of Facebook. We
take for granted that we can find out what gender someone is on
Facebook. Moreso, we take for granted that we can find out what gender
someone is, in general.
Using the analogy of the Turing test,
I've devised a workshop which explores how we parse gender and gender
representation when others are divorced from our bodies. I am trying to
make legible issues around the gender binary and its supporting
structures. There are a few concepts to work with in service of that
goal: the gender Turing test and the idea of judging gender based on
text-based interaction; technical systems which occupy different places
on the spectrum of gender disclosure and, as a subset of that, the
attitude of those systems to the inclusion of descriptors other than
male or female; the ever-present comparison of the gender binary to
binary, and the real spectrum of gender alignment as more like analog.
The workshop is structured around a kind of Turing test, in this case called the
Strategies game. The
Strategies
game is a deliberately obfuscatory Turing test. Groups take on the role
of either the agent trying to ascertain gender, or the agent trying to
hide gender. All groups start by exploring their own cultural
assumptions about gender and listing things that they see as essential
signifiers. Some groups devote themselves to defining strategies for
concealing gender online, others devise strategies for identifying
gender. The agents trying to ascertain gender use their list to devise a
system of winkling the information out of their opposing agent. The
agents trying to hide their gender do similarly, attempting to devise
systems which prevent the other side from ascertaining their gender.
Strategies are listed on cards, which look like the cards used in card
strategy games, or sports trading cards. Each strategy, whether its goal
is to be revelatory or obfuscatory, goes on a card. Once the
card-development session is done, groups nominate cards to go into a
deck which will be used by one of the players in the Turing test.
Once
the decks are made, based on educated guesses by the groups and each
group contributing an equal number of cards to one of the two decks,
everyone in the room closes their eyes. A volunteer is sought to play
the role of the identifier. They are taken to a chair in the front of
the room, facing a projection screen, and facing away from the rest of
the group. With eyes still closed, another volunteer is sought, to be
the concealer. They are taken out of the room, to a separated area
equipped with a computer. The two players connect to some kind of chat
client or collaborative editing platform (so far, I've used etherpad).
Taking alternating turns, the two players use strategies listed on their
respective decks of cards. Each strategy can only be used once. At the
end of the session, when both players are out of cards, the identifier
is asked if they believe they can identify the gender of the concealer.
In asking participants to consider strategies for concealing
and identifying gender, and in playing out a modified Turing test, this
workshop tackles ideas of gender binaries, cultural gender scripts and
requirements, the violence of forced disclosure, and the differing
conditions under which we identify ourselves as gendered.