2006.05.18 :: final presentations

seth.jeevan :: one-bit (small-bit) signaling

the skittle experiment
everyone distributed -- then pass around to whatever fancy
compare the quantity of each person

you get these things for different reasons - social interaction, undisclosed, colour favourites, active vs nonactive
no cost for giving, except that you have a finite number in your possession - need to gauge your in vs out items

--> low-cost signaling
what can you deduce, analysis of this signaling system, costs, benefits, etc

can one-bit message acquire meaning without background context, but with existing signaling costs? (ie. finite resources, handicap)
i.e. can this meaningless system acquire meaning through cost-based methods

starfish: web-based graphical system, with each click you increase size of other person, and you decrease
edge thickness - indicate amount of action between two people -- history of last 20 clicks
different dimensions -- judgement based on circle size, judgement based on line thickness

- asynchronous
- semi-anonymous (no indication of specific parties)
- exclusively non-verbal: no text, just graphical

meanings rise from the dust:
clicking signals attention:
check if they're online, see if they're active
teasing and start clicking game
poke to initiate a new relationship
varied - mirror real-world relationships (i.e. can identify medialabber network)
different interpretations of dot size:
rich dot --> poor dot: gift / charity
poor dot --> rich dot: status seeking, want to be included in a community's network, tap the hubs
equal dot --> equal dot: mutual affection
from study, people did not click randomly, had intention signals

* deception? inescapable level of ambiguity in this simple system
deception is all tied up in the couched intentions
is it possible to deceive? or misunderstand? to what extent?
is there a common understanding in an ambiguous situation?
complicit interpretation throughout a community
what to add to make deception a little bit less ambiguous
as time progesses and more bits are available, more consensus on how to interpret this system
what if - what happens if your dot gets depleted? does it disappear/die? is this a competitive game, or more social environment?

-- need more immediate feedback for the user, to tell them what's going on with their click inventory

how do you interpret the different actions and appearances of people in this community? perception from others
bringing conceptual meaning to signaling system


jeff.jesse :: social signaling in 'the sims 2'

game is social simulator, non-violent, popular, all-time #1 game
looking at modeling signaling

create character + environ, issue request to other characters, manage needs

choose many specific details of appearance, physiological and fashion
choose aspirations : define exactly what fulfills their needs

quantify the inventory of 'needs'
everyone you meet gets added to your list of contacts/relationships
other characters are defined with their own set of (invisible) needs
rating on how they feel about you (1D)

rich guy vs poor guy
when they hit on / hug on random women, the women resist / disgust

ambiguous, generic 'language'

giving off impressions:
guy who hasnt showered, stinky and unclean
the women hugs him with feeling, though she does perceive his stinkiness

appearance has no effect
clothing has no signaling effect (free)
if you play music, or serve food, attracts other characters
quality of relationships is purely 1 on 1

- adaptations
clothing can change impression
represent other Sims as a bundle of features (build-in prejudice)
- store identities not as single value but bundle
- compare a stranger to what you know about people with similar features
- i.e. ppl with hat and green shirt are 'typically' argumentative
generalise about strange Sims based on common features
--> add sims + social learning, general impressions
* how do you do this? what kind of structure / process?

you cant change your clothing - no awareness of physical appearance built-in
* how would you build something different to incorporate more awareness of physical features?
-- skill system, gain knowledge / skills, higher status; can attract people by your skills to fulfill their needs
* focus more analysis in terms of skill sets and how it affects / transforms / reframes relationships
-- short-term vs long-term effects in the addition of feature bundles
* incorporate more costs (time, money) -- pay for a wardrobe, or pay for cooking class? reading books or going clubbing? make a forced choice, display of value
reputation: gossip strengthens the bond between the gossiping parties (as akin to gossip / social grooming research, one-to-one exchange of info)
reputation affected only by firsthand witnessing (i.e. guy observing woman flirting with other guy)
adaption: incorporating more nth-hand reputation system, trust network

assessment signals: qualities inventory (read directly), or agent initiates actions to demonstrate
convention signals: user-initiated actions

deception --
get ahead by self-fulfilling things (money, skills)
love triangle - omitting knowledge from other people
marrying for money --> kick them out of house, to keep cash
deception is costly in effort -- if you're going against the inertia of your character, must be consistent in the alternative action
* like ekman, more costly to be more deliberate in showing what you're not truly feeling
- what if you could disguise yourself / change physical appearance / physically deceive?

romance - if you see your lover with someone else, can 'understand' whats going on, affect feelings toward rival

-- deception is tricky, in a constructed system, lots of things are fixed but you work around it
-- deceiving your character vs deceiving others, role-playing is weird; conscious mind vs unconscious mind
* where does the human fit in vs where does the character fit in? actions of character autonomous, or consciously dictated?


mirje.pallavi :: signaling in second life

can change appearance of your avatar, everything from clothing to hair
personality usually keeps consistent but physical appearance can change

role-players
entrepreneurs, building stuff and selling them for profit
social experiments

reputation: costly to rate someone
display of property / possessions owned

profile of 2nd life character, but also display of '1st life' character (ie. real human user)

collab: supportive community
competition: strong markets in terms of goods, housing, land, business; very object-based hierarchy

cost: people invest *lots* of time
- design avatar, build homes, shops, objects, have fun, to socialise, play game called tringo
- real money and professional skills are investments that are beneficial, can improve reputation

virtual mask of role-playing reality -- less barrier / vulnerability to feeling invaded, privacy
everyone, stranger is embodied - can observe without speaking to them
versus IM - which you're there as 'yourself' and dont want to be intruded upon, no observable signals
- investigate more deeply to see if people are being honest / deceptive in their messages

* what makes it compelling to care on what people think of you in second life?
* social dynamics of wanting to be well-regarded by people is high --- why? how is this notion of care and value of reputation designed?

merging of second life and real life -- interact with harvard law school lecture, interact via skype
handicapped people reveal show identities and indicate they are active in second life

* legality and legislation in second life, establish a parallel system, testbed for reality


alea.orkan :: anonymous companion agents for autonomous partner selection

p2p finding and matching partners for various connection purposes
- these systems learn about ppl to support their social interactions
- establishing a long term relationship with your agent

interface
looking for a match (ex. person join you at the theatre)
define criteria - explicit features like gender, quantifiable stats, keywords (answerable by agent)
arbitrary/personal specific questions - 'why are you interested in coming?' (answerable by human)
1. criteria defines short list
2. agent defines ranking (using agent knowledge), secondary criteria; assisted commonsense reasoning
3. if results are above threshold, then you email questions to a select group
-- privacy! what do you want to hide in certain situations? signaling vs privacy
* example: saying you're gay if you're looking for a date, but hiding that information when looking for a job
* define the signaling capabilities + complexities

agents know all features of identity without being specifically identitying a real person
* to what extent can it negotiate with other agents? being different than just a complex search function?
* selectively reveal information in smart negotiation, identity costs in that
* agent to present you in a positive and relevant light so that f2f meetups arent traumatic
* signals are deliberate and evolved communication, distinctive from cues

* signaling through proxies -- agent is signaling on your behalf, agents are receiving signals on others' behalf
* less on person vs. agent (not directly socially competitive)
* investigate competitive signaling through mediation, smart proxies
* search engine versus agent, what level of sophistication and social knowledge and level of control
* highly contextual issues! how to define each situation? and severity of each criteria?

2006.05.11

notquitehuman!

turing test
paper on creating a human - in philosophy journal
the 'imitation game':
communicating with a woman and a man

two people communicating through written communication (typing notes back and forth)
question on gender - embodied aspect
remove the cues of the body - cant see them, voice, handwriting, movement
use language solely as a cue as embodied quality of a person

intellect more important than body?
is intellect separable from the body?

need it to be challenging to be a game
distinguishing and expressing gender - based on cultural knowledge
easier for someone who is living, human, in the culture
what answers are more 'male' or more 'female'?

what does the game say about the question of gender?
- easy to use language
the game, challenge, is a little ambiguous

person + computer ... both claiming they are human

what is the distinction between quality and signal?
is the quality of being human important? the quality or the ability to replicate the quality?
in the game, passing as a woman means that you have the knowledge, but you're not really a 'real' woman

what is the role of language? is the ability to speak in a certain way enough to say that you are?
in intelligence, is language more of a signal?
signalled knowledge ?= real knowledge

surl = phil. at berkeley
the chinese experiment:
every phrase you can translate (just a set of useful responses) from some comprehensive reference
- no knowledge of *meaning*, no learning, just relayed symbols
appears you are speaking chinese, but you have no knowledge
machine only in syntactical domain, not in semantic domain (like humans, interpretation)
- if you make the thought experiment arbitrarily large, does it become just like a complex machine
the human brain - is it merely a machine of highly complex input/output?
what is the degree? what is the metric of 'is it a human'? how far do you go to be convinced?

loebner prize: how convincing is the computer as a human?
loebner
salon article (do not miss)
on wikipedia

is language an assessment signal on how we think?
signal as quality of how thinking works?
shared experience is one thing, but also cultural sharing, but also tiger-standing-on-box thing...

what is the cost of being deceptive?
parsing a grammatical sentence -- for ELIZA, the cost is relatively low
when is it too costly to create something to pass off as human?
expressing subtle reasoning + understanding + shared experience -- how costly to fake it?

is there a
you can signal thinking without 'thinking'
but you must demonstrate + representation of a disembodied quality
can you have real artificial intelligence?

how do you define thought, knowledge, and thinking?
are we more concerned with the signal, or the quality?

robotic pets for therapeutic needs (PARO)
human pet interaction
signaling need for affection
affection --> liveliness?
tamagotchi -- people are obsessed, felt responsible for them, felt sense of failure when it died
but you dont pet it, it's not that 'alive', it's plastic
however, a PARO - mixed of signals, by seeming dependency, needing affection, but also more prototyped as a real being
paro vs. bringing in real animals

no right answer - cannot check quality to signal -
discuss / debate the distinction between quality and signal and the relationship between them and the interpretation of them

how are symbols grounded?
- in physical being
- in
deb roy: robot, sensory experience of the world, build intelligence from sensory experience?

in life, when do we want the signal, and when do want the quality?
what is the cost to the receiver?

do we want to live in a world populated with PARO-like relationships?
affection and relationships that are programmed for emotional needs without 'the mess'

brave new world: future is comfortable, easy, pleasant, but nightmarish emptiness

treasuring things and relationships more
- the investment of energy / time -- a loss
- importance of neutral regard
how much do you care about what the other thing cares about you?
more responsive robots are built seem like they have an opinion of us

in most relationships, want the other person to play their role correctly
not really show what they're really thinking
satisfied with this 'role-playing', being comforted in the interaction
acceptable behaviour

is a person just a conglomeration of mere roles?
existentialism abounds!

but it's almost impossible to reflect what we are really 'are' and thinking through what we can convey
different underlying system to expended costs to express

when do you want to bring in a belief of a relationship or not?
human + machine, or human + human-like machine?

stepford wives - behaving like loving wives, but no meaning / love inside
ritalin, prozac - change inner qualities to reflect a more correct outer quality

more and more, we are questioning -
interaction with humans, interacting in a proper way

2006.05.04

faces!

understanding what we read into the face
how does one consider it in interface design?

long term cues, fairly immutable: skin condition, hair loss
long term, more immutable + deliberate: haircut, lines of facial expression
short term (facial expression)
decoration: hair, makeup, glasses

in-borne feature vs. deliberate feature
metaphorical generalizations not really grounded in anything

cues are superset: anything you use to read any hidden qualities
signals are intentional or evolved (meant to communicate)
evidence, unintentional cues: not meant as communication but can read into

ex. person looking like an animal -> has nothing to do with their personality
how do you interpret this information?
behaviour associated with this animal
associate look with behaviours
role of it as a cue purely exists in the interpretation process (b/c sender cannot control this at all)

signal that gets generalised as unreliable cue
an infant with baby features - intentional cue so that it's taken care of
an adult with baby features - unintentional cue
--> misinterpretation of a cue

more contextsensitive

ekman:
all signals are beneficial
so all facial expression are meaningfully beneficial
-->
voluntary vs unvoluntary expressions
heart attack: having a face from emotions despite fact that no one is around
do ppl imagine audiences? watching movie w/ friend, alone, told to imagine a friend
process has evolved to be innate - just b/c the moose is alone, he still has antlers.
evolution: parts of brain have mapped to behaviour - make expressions at certain times, i.e. becomes involuntary
--> what is intent? evolution? communication? conscious or unconscious?

natural state overlaid with social nuances, social learned behaviour
--> first question, does emotion even exist? or a set of fundamental emotions?

ekman: emotion as a state ("how am i feeling right now")
fridlund: emotion as an interactive sense ("what am i going to do? how am i going to react) --> makes sense that emotion has a signaling component

struggle -- trying not to show ppl too much about yourself
layers of inner feeling plus the cultured coddling of external representations

ekman: why would you be leaking anything? merely a display
fridlund: why a leak? there must a reason.

generally useful to couple the facial expression with communication inner state
not always conscious control

reading faces is a weird thing -- we're not always good at interpreting faces
evolution to take care of babies is pretty important -- treating baby-faced adults as such is just a side effect

'happiness' --> a conflation of many diverse phenomena

involuntary: blushing, crying
evolutionary uses -- to seek help

what's useful about putting physiological / emotion data in communication system?
blushing is easier to effect on some people

affective chat circles - display people's arousal state thru intensity
- online, has a remote aspect, what is arousal triggered by? online or off?
- colour change fabric, wouldnt add much to a face-to-face interaction
- redundant + inefficient + contradictory (is this desirable or ethical? like a lie-detector)
- facial expression is very complex and we're good at reading them, evolved
- skin response isnt a signal
- deep privacy issues -- sharing these originally hidden things about yourself
galvanic skin response isnt developed to be shared, but we have it
plausible case: evolved communication through tactile -- silent handholding in the forest

GSR -- who do i show this to? can i turn it off?

having awareness of gaze, attention --> problems with not being in shared space

response: interacting w ppl in a graphical space in representative + informative

how can we take these involuntary cues to create a more interesting environment rather than merely supplementing faces?
this 'other' space for the outputs? want to make sense out of the data - interpretation

logitech thing - separating rough facial expression from identity
these work better for videoconferencing than regular video because we read too much into real faces -- reduction to basic expressions can convey enough communication.

looking at people in public
in the T --> very awkward, want privacy in a very public place
gaze, activity
neutral face
deep cultural level - uncomfortable to stare at someone, to read something about them - privacy of our face
the subway is a very tension-rich place, close quarters
--> eye contact opens up connection w other people

sense of intensity of relationship in phyiscal proximity
-- any possible incorporation of this into a mediated interface + space?

face-to-face is efffortful, have to focus
want reaction without making interface as demanding as actual presence
lightweight medium - balance of meaning and cost