‹date/time&=
#8250;
‹footer›
‹#›
------=_NextPart_01C42787.07941530
Content-Location: file:///C:/EF8AB227/Turing_files/pres.xml
Content-Transfer-Encoding: quoted-printable
Content-Type: text/xml; charset="utf-8"
This research was supported by the following grant: NSF ITR Award IIS 0205740<=
![endif]>
![](3D"master47_image001.gif")
![](3D"master47_image002.gif")
He
or a She? Nonverbal Gender Detection
in Immersive Virtual Envir=
onments
Rosan=
na E.
Guadagno, University of California, Santa Barbara
Aleksandar Dimov,
University of California, Santa Barbara
Jerem=
y N.
Bailenson, Stanford University
Andrew C. Beall,
University of California, Santa Barbara
Jim Blascovich,
University of California, Santa Barbara
Weste=
rn
Psychological Association Convention, Phoenix, AZ, April 2004
This research was supported by the following grant: NSF ITR Award IIS 0205740<=
![endif]>
![](3D"master46_image003.gif")
![](3D"master46_image004.gif")
Abstract
<=
span
style=3D'font-size:50%'>oThe purpose of this study was to examine whether individuals can dete=
ct
and transmit meaningful information non-verbally in an Immersive Virtual
Environment (IVET). Participants were 32 (16 male, 16 female) introductory
psychology students who engaged in a series of non-verbal interactions with
each other. Each participant entered an immersive virtual environment with=
two
others – always one male and one female. We sought to determine if
participants could successfully identify the gender of the other interacta=
nts.
Each person was represented by an avatar that was designed to be gender
neutral so as not to provide any cues as to the gender of the interaction
partners. There were three phases of trials in the experiment: a baseline
block in which no specific instructions were provided, a block of trials in
which participants were asked to try to determine who among their interact=
ion
partners was the male and who was the female, and a block in which
participants were asked to continue to detect the gender of their interact=
ion
partners and to also non-verbally transmit their gender to their partners.=
The
location of the interaction partners in the virtual world alternated rando=
mly.
Results indicated that participants were not able to detect the gender of
their interaction partners significantly above chance in the baseline bloc=
k.
Mean accuracy for this block was 47%, t (31) =3D
-0.49, n. s. Nor were participants significantly better than chance=
in
the next block of trials where they were asked to detect gender only. Mean
accuracy for this block was 54%, t=
(31) =3D 0.81, n. s. However, participants in the=
final
block of trials were significantly better than chance at accurately detect=
ing
the gender of the interaction partners. In this case, mean accuracy was 64=
%, t (27)=
=3D 2.83,
p
< .01. There were no gender differences in accuracy scores. Thus, it
appears that men and women can detect gender via nonverbal interaction in =
the
absence of visual cues to gender, but only if the target was trying to con=
vey
his or her gender.
This research was supported by the following grant: NSF ITR Award IIS 0205740<=
![endif]>
Introduction: Prior Research
oPrior
research has reported that:
nMen
and women have very different patterns of non-verbal behavior
(e.g., Bente, Donaghy, & Suwelack, 1998)
oNo
prior research has used Immersive Virtual Environment Technology (IV=
ET)
to examine this issue
oIVET
provided us with a unique way to mask visual gender cues (e. g=
.,
people in the virtual environment appeared as same
gender-neutral avatar)
This research was supported by the following grant: NSF ITR Award IIS 0205740<=
![endif]>
Introduction: Purpose of Experiment
------=_NextPart_01C42787.07941530
Content-Location: file:///C:/EF8AB227/Turing_files/slide0023.htm
Content-Transfer-Encoding: quoted-printable
Content-Type: text/html; charset="us-ascii"
oThe purpose of this stud=
y was
to examine whether meaningful informa=
tion
could be transmitted non-verbally
between participants in a Virtual Environ=
ment.
oIn the absence of visual cues for ge=
nder,
we asked participants to try to determ=
ine
the gender of the other participants=
in
their virtual environments.
This research was supported by the following grant: NSF ITR Award IIS 0205740<=
![endif]>
Introduction: Turing Test
oInspir=
ed by
the original paradigm Turing (1950) used in his
formulation of what is now ref= erred to as:
formulation of what is now ref= erred to as:
n“The new form of the problem can be describe=
d in
terms of a game which=
we
call the ‘imitation game.’ It is played with three people, a <=
/span>man (A), a woman (B), and an interrogator (C) who =
may be
of either sex. The
interrogator stays in a room apart from the other two. The object of the game for the interrogator i=
s to
determine which of the othe=
r two
is the man and which is the woman. He knows them by labels X and Y, and at the end of th=
e game
he says either ‘X i=
s A
and Y is B’ or ‘X is B and Y is A.’ The interrogator is =
allowed to put questions to A and B”.
This research was supported by the following grant: NSF ITR Award IIS 0205740<=
![endif]>
Predictions
oWe predicted that:
nMeaningful information such as gender could be transmitted
non-verbally in IVET
nIndividuals would be able to detect gender from
other’s head movements, particularly when instructed to=
transmit
their gender to the individuals they interacted with.
nWomen would be more accurate than men because women
are more adept at decoding non-verbal cues (Rosenthal & DePaulo,
1979)
n
This research was supported by the following grant: NSF ITR Award IIS 0205740<=
![endif]>
Method
oParticipants were 32 (16 male, 16 fem=
ale)
undergraduates who were
run in groups of 4.
oParticipants were told that the purpo=
se of
the study was to examine
non-verbal communication in virtual environments.
nThey were additionally asked to inter=
act with
other participants via only head movements.
oThere were three phases of the study:=
nbaseline data were gathered (2 trials=
).
nparticipants were asked to determine =
the
gender of the other participants
(12 trials).
nparticipants were asked to determine =
the
gender of the other participants
as well as try to convey their own gender to the other participants (12 trials).
This research was supported by the following grant: NSF ITR Award IIS 0205740<=
![endif]>
Method
oThe location of the individuals in the
virtual world varied across
trials.
oOnce in the virtual world, participants appeared t=
o each
other as an identical gender n=
eutral
avatars.
oRegardless of participant sex, each participant saw
avatars representing one =
real
male and one real female.
oThus, participants interacted with two other indiv=
iduals
per session but not a=
ll
participants in each session interacted with each other.
oSee Figure 1 for a screen shot of the virtual
world.
oSee Figure 2 for the lay out of the experiment.
This research was supported by the following grant: NSF ITR Award IIS 0205740<=
![endif]>
Method
oThe following dependent measures were assessed: &#=
13;
nGender selection (collapsed across block)
nAfter each trial, participants were asked to
select:
owho looked at them more
owho had the most abrupt head movements
=
owho they liked more
nAt the end of the experiment, participants
completed:
oPersonal Attributes Questionnaire (Spence & Helmreich, 1978)
oA measure of gender identification scale
oA
manipulation check to assess whether participants actually <=
span
style=3D'position:absolute;top:84.25%;left:17.22%;width:76.96%;height:4.5%=
'>believed they were interacting with other real
humans
This research was supported by the following grant: NSF ITR Award IIS 0205740<=
![endif]>
Results:
Accuracy in Gender Determi=
nation
oThere were no gender differences in a=
ccuracy
of dete=
rmining
gender. There were no groups of participants who together were more accurate than =
other groups.
oAccuracy averaged over the trials is =
as
follows:
n% correct selection of male/female:&#=
13;
obaseline: 46%,
n. s.
ofirst block (detect gender only): 53%, n. s.
osecond block (detect and transmit gen=
der): 64% t
(27) =3D 2.83, p<=
/span> < .01.
nSee Figure 3.
This research was supported by the following grant: NSF ITR Award IIS 0205740<=
![endif]>
Results:
Adjusting for detecting but not properly
encoding the signal
oAdjustment for picking up on a signal=
but
incorrectly encoding
it.
nThe accuracy was rescaled to make the=
range
from –0.50 to +0.50 with the chance/mid-point at zero, took the absolute va=
lue of
each accuracy score
on this new scale and looked to see if overall they differed significantly from chance:
n% correct selection of male/female:&#=
13;
obaseline: 3%,
n. s.
ofirst block: 18%,
t (31) =3D 6.22, p < .001.=
osecond block: 22%,
t (27) =3D 6.25, p < .001.=
nSee Figure 4.
nThese results suggest that some parti=
cipants
did detect a difference in head movements but encoded it the wrong way.
<=
/span>
o
This research was supported by the following grant: NSF ITR Award IIS 0205740<=
![endif]>
Results: Other Dependent Variables
------=_NextPart_01C42787.07941530
Content-Location: file:///C:/EF8AB227/Turing_files/slide0021.htm
Content-Transfer-Encoding: quoted-printable
Content-Type: text/html; charset="us-ascii"
oParticipants chose males significantl=
y more
than fema=
les
(regardless of whether they accurately chose the =
male) on:
nthey liked the male more
<=
/span>
nthought he had the most abrupt head
movements
nhe looked at them more
oThere were no gender differences resp=
onse to
the scales measu=
ring who
looked at the participants more, whose head movements were the most abrupt, and whom participants liked more.
<=
/span>
<=
span
style=3D'mso-special-format:lastCR;display:none'>
This research was supported by the following grant: NSF ITR Award IIS 0205740<=
![endif]>
Results: Other Dependent Variables
------=_NextPart_01C42787.07941530
Content-Location: file:///C:/EF8AB227/Turing_files/slide0007.htm
Content-Transfer-Encoding: quoted-printable
Content-Type: text/html; charset="us-ascii"
oBut
were the men actually more likely to look at them and make <=
span
style=3D'position:absolute;top:29.5%;left:9.73%;width:84.45%;height:5.25%'=
>more abrupt head movements?
nWe looked at means and standard devia=
tions on
the Yaw, Pitch, and Roll data to get a feel for how much participants moved
around.
nMale
participants had significantly larger standard deviations in Pitch than did females (SD male =3D 7.62 vs. SD =
female =3D
6.13), t =3D 2.19, p
< .04 indic=
ating
that males were more nonverbally active with head movements.
oQuestionnaire data:
nPositive
affect towards one gender (e.g., “In general, I am glad to be a man/woman”) and the M – F=
scale
of the PAQ (which measures instrumentality and ability to assert oneself) were
predictive of uncorrected accuracy scores.
This research was supported by the following grant: NSF ITR Award IIS 0205740<=
![endif]>
Discussion
oThese results suggest that the transmission=
of =
meaningful
information such as gender may be possible nonverbally in I=
VET
but only after practice and specific instructions are
provided
oContrary to predictions, no gender difference in =
ability
to detect emerged, possibly due to the technology’s inabili=
ty
to render detailed facial expressions
This research was supported by the following grant: NSF ITR Award IIS 0205740<=
![endif]>
Discussion
oThere was some evidence that people detected a sig=
nal
but did not properly
encode it (e.g., noticed a difference in head movements but mislabeled males vs. females).
<=
/span>
oRegardless of gender, all participants showed a
preference for men over women
which was possibly a reflection of the greater overall activity of the make participants.=
oSome perceptions of participant’s experience=
in
the virtual environment (e.g.=
, that
men looked more that did women) were
verified with behavioral data indicating that participant’s impressions accurately reflect=
ed
what occurred in the
interaction.
This research was supported by the following grant: NSF ITR Award IIS 0205740<=
![endif]>
References
oBente, G., Donaghy, W. C.,& Suwelack, D. (1998=
). Sex
differences in body movement a=
nd
visual attention: An integr=
ated
analysis of movement and gaze in mixed-sex dyads. Journal of Nonverbal Behavior, 22, 31-58.
oRosenthal, R., & DePaulo, B. M. (1979). Sex
differences in accomm=
odation
in nonverbal communication. In R. Rosenthal (Ed.), Ski=
ll in
nonverbal communication (pp. 68-1=
03).
Cambridge, MA: Oelgeschlager,
Gunn, & Hain.
oSpence, J.T., & Helmreich, R.L. (1978). Mascul=
inity
and femininity: Their
psychological dimensions, correlates, and antecedents. Austin, TX: University of Texas Press=
.
oTuring, A. (1950). Computing machinery and intelli=
gence.
Mind, 59<=
span
style=3D'font-size:73%'>,433-60.
This research was supported by the following grant: NSF ITR Award IIS 0205740<=
![endif]>
Figure
1. A Screenshot from the Virtual Environment
![](3D"slide0010_image006.jpg")
This research was supported by the following grant: NSF ITR Award IIS 0205740<=
![endif]>
Figure 2. Set up of Experiment
![](3D"slide0022_image008.gif")
![](3D"slide0022_image010.gif")
![](3D"slide0022_image012.gif")
![](3D"slide0022_image013.gif")
Perceiver
Female Participant
Male Participant
1) Tracking system, 2) Computer, 3) =
Head
Mounted Device (HMD), 4) Gamepad