Content uploaded by Cindy Mason
Author content
All content in this area was uploaded by Cindy Mason on Jan 14, 2018
Content may be subject to copyright.
Appearance and Reality in AI –
The Case of the Robot Priest
Cindy Mason
BISC, EECS Dept.,
U. C. Berkeley, Berkeley CA 94709
cindymason@media.mit.edu
Abstract
We discuss the practical role that appearance and reality
play in AI environments. When appearance and reality do
not line up, trust is broken. As the pedophilia problems of
the priests in the Catholic Church demonstrate, broken trust
also breaks the institutions and the lives that depend on that
trust. To demonstrate that the clothing of an agent, like the
clothing of a clergy, creates an expectation of what may oc-
cur in reality based only on appearances we invite the reader
to consider building a Robot Priest and to explore public
reaction to a Robot Fashion Show. The expectations can be
held both by people and other bots. Some of the biological
consequences of trust or lack of trust discovered by multiple
disciplines in the human sciences are described. These have
significant implications for the healthy and ethical design of
non-human agents and mixed-realities. We conclude coher-
ence between appearance and reality is essential, proposing
a WYSIWYG or coherence law.
Introduction
Appearance and Reality is an old philosophical problem.
Mostly philosophers argue whether there can be multiple
realities or a single unified reality, whether a set of facts
can completely describe a world, and whether with tech-
nology we are doomed to look at the world through graph
paper glasses (Bradley 1893) (Russell 1912) (Porter 2017).
The problem of Appearance and Reality in AI becomes not
philosophical, but immediate and pragmatic because we
are constructing ‘mixed-reality’ environments that we will
share, e.g. human-robot societies, virtual reality-AI envi-
ronments, augmented reality, and so on. Trust is based in
part on the coherence between appearance and reality.
Cliff Nass’s work on user experience indicates that users
who repeatedly interact in a certain way with a gadget will
then turn around and behave that way to others (Nass and
Moon 2000) (Reeves and Nass 1996). In modern life
many of us are in a symbiotic relationship with our gadgets
on an almost 24/7 basis (cellphones, touch pads, laptops,
etc.) If our robots/agents and mixed-AI systems appear to
be trustworthy but in reality are not, then the increased
exposure to such systems and agents means the distrust of
the systems can add to our distrust of one another.
What kind of laws or policies would help us to create
trustworthy mixed-reality? In a mixed-reality human-
robot/softbot/vr society should we base our trust on what
we see or do we need to look deeper into the character or
software? What happens when we lose trust? A robot that
appears to us as a priest should be trustworthy. What
should a trustworthy robot priest look like? We address
these questions in the paper. My hope is that the material
presented here will motivate some of you to begin to work
on the answers for AI and the future to come.
The paper is organized as follows. First we give an AI
reader some insights into the growing body of multi-
disciplinary human science discoveries involving trust –
the biology of trust, its medical significance to our health
through touch and its foundation for happiness as human
individuals and in communities. The discoveries indicate
our lives depend on trust in every possible way and that we
are plastic. We change not just our brains but everything,
from wound healing to gene expression (Mason 2014).
Next we look at failed trust – specifically the failures of the
Catholic Church and introduce the idea of Robot Priest.
The implications of Human Sciences for Human Robot
Interaction (HRI) goes beyond Robotic Priests. We demon-
strate the problem of Appearance and Reality in AI by ex-
ploring what happens when the public witness a Robot
Fashion Show. Spoiler – they don’t bother to ask about
hardware or software, but created expectations of behavior
based on appearance. This leads to the conclusion of the
paper with the proposal of a WYZIWYG Law or the Law
of Coherence.
Copyright 2017 The Author. AAAI/ACM Conference on AI, Ethic and Society, 2017. Submitted.
Copyright 2017 The Author. AAAI/ACM Conference on AI, Ethics and Society, 2017.
Human Sciences and Trust
The Biology of Trust
In humans, trust and touch are deeply connected. The short
explanation for this is loving touch stimulates production
of oxytocin and oxytocin enhances trust (Kosfeld et al.
2005). We used to think oxytocin was associated with
birthing and nothing else. We were very wrong. Oxytocin
receptor sites are in the heart, pancreas, testis, ovaries, pi-
tuitary, thymus, adrenal medulla, vascular endothelium,
osteoclasts, myoblasts, pancreatic islets, adipocytes and
more (Gimpl and Fahrenholz 2001). There is evidence that
oxytocin plays a role “in the control of cardiovascular
functions, thermoregulation, pain threshold and fluid bal-
ance”(Lupoli et al. 2001).
Trusted touch also releases a host of other hormones
including insulin, gastrin, cholecystokinin, somatostatin,
and prolactin (Lupoli et al. 2001). Prolactin has a modula-
tory role in several aspects of immune function, such as
modulation of immune response (not too much, not too
little). So trust means not only predicting something will
not hurt us, but that whatever it is we are trusting (and
touching) is good for our health.
How does touch create the chemistry of trust? One
square inch of skin has over 1000 nerves, each surrounded
by microscopic blood vessels (Human Skin 2017). Contact
with the skin stimulates chemical and electrical signals that
travel the branches and pathways of nerves. This starts at
the surface of the body and travels to many locations, in-
cluding brain, organs, glands, etc.
Tactile Power of Trust
Trust and touch go hand in hand. We recoil from the touch
of someone we don’t trust, and run to the arms of someone
we do. Human beings are born of touch. As infants, we
die without it. Just how far does our need for trust and
touch extend? There is evidence to suggest relationship
trust is the foundation of our species – our health and lives
depend on it.
One of the most famous examples of this occurred in
1995 and is famously known as the “Rescuing Hug”. Dur-
ing the first week of life a set of twins, born premature,
were in incubators at Worchester Hospital, Massachusetts
(Townsend 2001). One of the twins was not expected to
live and had low blood pressure and low heart rate. She
would cry until blue in the face and exhausted. “A hospital
nurse fought against hospital rules and placed the babies in
one incubator. When they were placed together the healthi-
er of the two threw an arm over her sister in a endearing
embrace. The smaller baby's heart rate stabilized and her
temperature rose to normal,” (Townsend 2001). These two
babies seemed to instinctively know what to do.
At Stanford Hospital patients undergoing bone marrow
stem cell transplants, an invasive procedure were all bone
marrow is irradiated followed by a lengthy recovery period
where the cells in your body ‘re-boot’, halved their ex-
pected recovery time (56 days in hospital) when they en-
gaged gentle touch therapy (Mason and Mason 2001)
(Shannon 2002). Similar positive results exist for open
heart surgery patients, also at Stanford (Gasner and Mason
2008). Rigorous peer reviewed studies exist on the positive
health impacts of touch therapies on over 40 different dis-
eases. For the bibliography of touch therapy studies see
(Mason and Mason 2001). In most if not all of these stud-
ies trust and touch work together to significantly improve
health.
The Human Need For Friends
Part of a healthy response to life’s burdens involves shar-
ing them with a trustworthy person. To be human is to un-
burden the soul, to confess, to find hope, to forgive, to
have joyful belonging and to be encouraged when we want
to give up or give in. We worry, we celebrate, we laugh
and we cry. We are preoccupied with fear or depression.
We grieve and need to tell our stories over and over. There
is a real human need for trustworthy friends. Part of the
role of a priest/monk/rabbi/etc. is to fill this role for us,
often in ways few others can. What happens when there is
no one?
According to the NIMH, Anxiety Disorders affect
18.1% of adults in the United States (approximately 40
million adults between the ages of 18 to 54) (NIMHa 2017)
and 9.5% percent are on anti-depressants (NIMHb 2017).
In the Journal of Clinical Psychiatry, it is reported Major
Depressive Disorder prevalence rose from 13.8 million to
15.4 million adults between 2005 and 2010 (Greenberg
2015). I’m not saying a Robot Priest/Monk/Rabbi/etc.
could be a fixer for all these problems, but maybe it could
help. My physician and friend Ed Weiss once quipped,
what’s the difference between a friend and an emotionally
intelligent robot? You can tell your truly very sad story
over and over and over to the robot and it will cry with you
every time. For more multi-disciplinary studies on human
sciences and technology design see (Mason 2014).
Catholic Priests – Where Trust Failed
Catholic priests wear recognizable clothing such as a
cloak, collar, or symbol that signals a role in society asso-
ciated with openness, dependability, and absolute trust.
When the Boston Globe uncovered that over 270 priests in
Boston were pedophiles and that thousands of children had
been victims, people lost trust in their priest. It happened
in cities all over the world (Boston Globe 2015). Further
ripples of distrust resulted when it was learned that the
Copyright 2017 The Author. AAAI/ACM Conference on AI, Ethic and Society, 2017. Submitted.
Copyright 2017 The Author. AAAI/ACM Conference on AI, Ethics and Society, 2017.
institution had looked the other way and/or protected these
priests.
As AI developers and researchers, we can neither allow
the development of non-human agents that cause harm nor
should we participate in looking the other way and/or cov-
ering it up. The effect of the broken trust human priests
caused was devastating and many victims never recovered.
Tens of thousands became drug addicts or never lived
normal lives and others committed suicide.
To have somewhere and someone to trust in times of
trouble or celebration is a treasure to us humans. If people
no longer easily trust their human priests, and there is less
of religious community life (Layman 2015) (Putnam and
Campbell 2010) could they trust a robot priest? Could
you trust it?
Figure 1. Two Robot Priests have same hardware but different
appearance.
The Psychological Factor in Communities
In considering a robot priest in society, it also matters to
consider the expectations from social contexts for such
robots. When trust in the institution of a church or a priest
breaks the community is also lost, and its functions must
be replaced. Tom Layman, Harvard Divinity School, says
that religious communities help “people aspire toward
goals, transform themselves, and work toward change
while holding each other accountable to make things bet-
ter. They are inspiring creativity and inspiring people to
find their purpose and mission in life. All these kinds of
key functions that a community can serve"(Layman 2015).
Theologian Casper Ter Kuile says non-religious com-
munity is stepping in where religious community once
was, to fill the need for joyful belonging in the present time
of certainty when millions are uprooted and disasters are
increasingly in the news (Ter Kuile 2015). Its time that the
AI community begins to investigate whether it can provide
support in these areas of life.
Appearance and Reality in AI
We have seen how trust is essential to human life, mostly
through touch and relationship. There is also a visual side
to trust. For example, we develop expectations of behav-
ior, skills, etc. by looking at what someone is wearing e.g.
physician, mechanic, police, school boy. To hide who we
are we wear a costume or mask. Literature is full of these
kinds of appearance and reality scenarios, e.g. Hamlet. In
mixed-reality environments where touch is limited people
and other agents will make snap decisions of trust based on
appearances.
The shared beliefs we hold about dress and appearance
in the world around us generally keep us sane and keep us
safe together but they rely greatly on trusting and trustwor-
thiness. Occasionally someone can impersonate a cop or a
doctor by dressing like one. In dealing with a robot priest,
monk, rabbi, or a softbot banker, physician, teacher, etc.
we naturally bring an expectation of trust to the relation-
ship. Imagine the distrust that will develop towards AI if a
robot impersonated a cop or a doctor. To explore the prob-
lem of appearance and reality in AI we staged a simple
robot fashion show at the Stanford Shopping Mall less than
a mile from the AI Center at SRI Intl.
Robot Fashion Show
During the 3rd week of March, 2015, we held a fashion
show at Stanford Shopping Mall’s Nordstroms with one
robot and 5 outfits. As shown in Figure 2 the clothing was
chosen to illustrate a variety of identities we might relate to
in the world: A) a businessman, B) a frilly girrly, C) a
bride, D) a superhero, and a monk (not pictured).
Figure 2. Four images from Stanford Robot Fashion Show
We used the same hardware and software for each outfit,
except at one point during change of outfits, we attached a
Rasberry Pi computer to the robot chassis under the cloth-
Copyright 2017 The Author. AAAI/ACM Conference on AI, Ethic and Society, 2017. Submitted.
Copyright 2017 The Author. AAAI/ACM Conference on AI, Ethics and Society, 2017.
ing and put small electronics in pockets. These were not
visually discernable after the robot was dressed.
Individual reactions of the public were generally to at-
tribute the expectations of behavior to the robot’s outfit.
None of the observers asked about hardware nor did they
question the software. Some people assumed there was
more than one robot.
Most of us know the software and hardware is what de-
termines how a robot will behave. The fashion show illus-
trates we humans do judge a book by its cover, and when
robots become part of human society or we engage more
and more in mixed-reality, we are likely to judge things by
the way they look. This is a key insight for policy/law
makers.
Although not a formal study, the significance of the
show is that it demonstrates that coherence between ap-
pearance and reality is essential to ensure public trust. The
appearance of a robot/agent plays a role in expectation of
what will happen in reality (behavior, speech, personality,
anthropomorphic behavior, etc.) Appearance and reality
have challenged us in many ways throughout time – not
just as a philosophical problem (Bradley 1893) (Russell
1912) but also pragmatically. For instance, the Royal Soci-
ety’s struggle to determine the ontological authenticity of
the ‘vegetable lamb’ from Tartary lasted centuries (Apple-
by 1997) and we currently struggle with fake news, fake
currency and email attachments. In mixed-reality AI envi-
ronments we face the issue as well, especially so if robots
wear clothes. We create expectations of behavior and
trustworthiness based in part on appearance, yet the reality
is software and hardware determines how a robot will be-
have. If we now add clothing, it becomes crystal clear we
need policies or perhaps laws about coherent appearance,
like WYSIWYG.
WYSIWYG Law of Robotics
Many people are currently working on a modification to
Binder and Asimov’s original ‘Law of Robotics’ (Asimov
1950) (Binder 1939). As proposed by the Engineering and
Physical Sciences Research Council and the British Stand-
ards Institute currently the 4th Law of Robotics reads, “Ro-
bots are artifacts; they should not be designed to exploit
vulnerable users by evoking an emotional response or de-
pendency. It should always be possible to tell a robot from
a human,” (Laws 2017).
To address the issues of trust in non-human agents and
mixed-realities (including VR) law 4 on transparency must
go one step further, as evidenced by the public reaction to
the fashion show. Namely, coherency takes transparency
one step further. A modification to 4th Law of Robotics.
would address the idea that a robot or non-human agent
should look consistent with its purpose. The meaning of the
proposed regulation is that a robot is designed to look in a
way that is consistent in human terms with its software:
The WYSIWYG or Coherency Law: A robot or softbot’s
external appearance should be consistent with its pro-
gramming and therefore its decision and actions.
Summary and Discussion
The Implications of Human Sciences for Human Robot
Interaction and mixed-reality goes beyond robotic priests.
What are the ethics in designing robots and interfaces in a
human-bot society? Right now its like the law west of
Pecos, with famous voices spreading fear and doubt. The
position of this paper is that design decisions, laws, and
ethics policies should be based on scientific knowledge of
human sciences. As programmers, professionals, and re-
searchers we spend our days working on AI with comput-
ers, but what is our awareness of human nature? We have
shared some of the relevant discoveries and encourage
those who work mainly in computer science to look into
other disciplines. Because AI and gadgets are becoming
much more a part of our daily life, we owe it to ourselves
to look at this blind spot.
If a human or non-human agent is dressed like a priest,
talks like a priest and walks like a priest, should we trust?
Among human society, even if you are certified as a priest
and have a congregation it doesn’t mean you are trustwor-
thy. The cost of broken trust can be devastating for people,
society, corporations and the environment. We proposed a
WYSIWYG or Coherence Law modification to Law 4 as
proposed by the Engineering and Physical Sciences Re-
search Council and the British Standards Institute.
We are at a unique position in time in AI where we can
have a ‘do-over’ of the world… write the software so that
behaviors of agents we interact with inspire us, because
they will ‘rub off’ on us. So if non-human agents are
trustworthy, ethical, etc. the hope is we become more so.
If there are errors in the agent behavior, unlike with us hu-
mans, in the next version its relatively easy to fix.
Acknowledgement
The author wishes to thank the AI Center at SRI, Intl. for
resources and support. We also thank the many clerks and
security guards at the Nordstrom’s of Stanford Shopping
Mall.
References
Asimov, I. (1950). "Runaround". I, Robot (The Isaac Asi-
mov Collection ed.). New York City: Doubleday. p. 40.
ISBN 0-385-42304-7
Copyright 2017 The Author. AAAI/ACM Conference on AI, Ethic and Society, 2017. Submitted.
Copyright 2017 The Author. AAAI/ACM Conference on AI, Ethics and Society, 2017.
Binder, E. 1939. “I, Robot.” Amazing Stories. January,
1939.
Boston Globe, 2015. A History of Secrecy and Cover-Ups
in the Boston Archdiocese. Accessed November 14, 2017.
https://www.bostonglobe.com/news/special-
reports/2015/10/13/bcom-
spotlight/H1GowF3c56Yq7n68ONhJcK/story.html
Bradley, F.H. 1893. Appearance and reality. A MetaPhys-
ical Essay. London: George Allen and Unwin Ltd. Online:
https://archive.org/details/appearanceandrea00braduoft
Gasner, C. and Mason, C, 2008. Haptic Medicine at Stan-
ford Cardiology. Accessed November 11, 2017.
https://www.youtube.com/watch?v=zxlyjWzhO5E
Gimpl, G. and Fahrenholz, F. 2001. The oxytocin receptor
system:structure, function, and regulation. Physiology Re-
view, 81(2):629-683.
Greenberg PE, et al. 2015. The Economic Burden of
Adults With Major Depressive Disorder in the United
States (2005 and 2010). Journal of Clinical Psychiatry.
76(2):155–162.
Human. 2017. Human Skin. Accessed November 13, 2017.
https://en.wikipedia.org/wiki/Human_skin
Kosfeld, M., Heinrichs, M., Zak, P. J., Fischbacher, U., and
Fehr, E. 2005. Oxytocin increases trust in humans. Nature
435, 673–676.
Laws of Robotics, 2017. Accessed November 12, 2017.
https://en.wikipedia.org/wiki/Laws_of_robotics
Layman, T. 2015. How We Gather – A New Report On
Non-Religious Community. Harvard Divinity School. Ac-
cessed November 7, 2017.
http://hds.harvard.edu/news/2015/11/04/crossfit-
church-examining-how-we-gather#
Lupoli, B., Johansson, B., Uvnas-Moberg, K., Svenner-
sten-Sjaunja, K. 2001. Department of Animal Nutrition and
Management, Swedish University of Agricultural Sciences,
Uppsala J Dairy Res., 68(2):175-87.
Mason, C. 2017. Robot Fashion Show. AI, Ethics and So-
ciety. Accessed November 13, 2017.
Mason, C. 2014. The Multi-Disciplinary Case for Human
Sciences in Technology Design, AAAI Fall Symposium on
The Nature of Humans and Machines, Washington D.C.
Mason, C. and Mason, E., 2001. Haptic Medicine.
Proc. Future Health Technology Summit, Cambridge, Ma.
2009. Studies in Health Technology and Informatics,
2009, 149:368-85, Pub MedPMID 19745495.
Nass, C. and Moon, Y, 2000. Machines and Mindlessness:
Social Response to Computers. Journal of Social Issues,
56-1:81-103.
NIMHa 2017. Any Anxiety Disorder Among Adults, Ac-
cessed November 11, 2017.
https://www.nimh.nih.gov/health/statistics/prevalence/any-
anxiety-disorder-among-adults.shtml.
NIMHb 2017. Major Depression Among Adults, Accessed
November 11, 2017.
https://www.nimh.nih.gov/health/statistics/prevalence/maj
or-depression-among-adults.shtml
Porter, A. 2017. Private Communications. Nov. 10, 2017.
Putnam, R. D. and Campbell, D.A. 2010. American Grace:
How Religion Divides and Unites Us. New York: Simon &
Schuster.
Reeves, B., and Nass, C. 1996. The Media Equation: How
People Treat Computers, Television, and New Media Like
Real People and Places. New York, New York: Cam-
bridge University Press.
Russell, B. 1912. Appearance and Reality, An Analysis.
Problems of Philosophy. New York: Henry Hold and
Company.
http://www.gutenberg.org/files/5827/5827-h/5827-h.htm
Shannon, C. 2002. Jin Shin Jyutsu Outcomes in A Patient
With Multiple Myeloma, Alternative Therapies, 8(5). Ap-
pears in Haptic Medicine:Our Stem Cells Like Compas-
sion. Ed. Mason, C. Accessed November 14, 2017.
https://www.academia.edu/7404110/Haptic_Medicine_Our
_Stem_Cells_Like_Compassion_-
_Cancer_and_BMT_A_Case_Study
Ter Kuile, C. 2015. Knowing We Must Die, How Then
Should We Live? Sunday Assembly Archive, April 27,
2015. Accessed November 13, 2017.
https://www.sundayassembly.com/blog/knowing-we-must-
die-how-then-shall-we-live
Townsend, L. 2001. Premature Twins Thrive with a “Res-
cuing Hug”. In Worcester Telegram & Gazette, November
18, 1995. Accessed November 11, 2017.
http://www.csun.edu/~vcoao0el/webct/de361s52_folder/ta
ctilepower.html
Copyright 2017 The Author. AAAI/ACM Conference on AI, Ethic and Society, 2017. Submitted.
