and thirty one
i
i'm happy lasttime applied is due to indicate that at university
and
and today representing a study on are self disclosure in conversation dialogue system
so i haven't recovers also if you can't understand i'm happy i said
so this was the study was done as part of the cmu my is the
i v and is not exactly in two thousand seven
well as in just standing there is close to
so it's because human conversations of in humans and y are many solutions to think
is right here try to achieve the solution and restart it can be in of
all those that so propositional function based
i have any information of the conversation those that so interactional con functions with just
a system like the quantization for one
and so in cost functions which is
trying to build that of the essential between those who participated
so set motion is one of the key social strategy employed in conversation and intimacy
between pocketsphinx and interleaved idea of the conversation
so many definitions on the one wants to self disclosure
i o one is that strong or mean and it in nineteen seventy three which
defined as the one a static
all opinions or but experiences references by you and wasn't it
so
so
a no side distortion is a very interesting phenomenon in between very well studied by
the psychology community
in particular because of
it's actually the ability to use reciprocity and dyadic interaction
so that you must leave the phenomenon vibrates when one participated in a conversation self
disclosure
the other participants in most competitive set of discourse in this form
and there are many explanations on this but the exact called also based on the
screen on that's not what is the one hypothesis is that it's a formal solution
extreme where the party receiving side distortions this feels obligated but also said this goal
i don't know what concerns is that it's a solution conversational where if a
it doesn't sound is close and return they feel uncomfortable
and i and hypothesis is about social just attraction
ones that is close to people and that's close to that of because i
as i do not trust in like a by the exact it's not what a
distance between a pretty well established a
and right of self disclosure and have been reproduced in many studies and shall
to be a very strong
so subsequent studies also show other ask exercise caution like
a self disclosure reciprocity characterizes initial social interaction because people set is closely to try
to corpsman model
i interestingly you look at a distance to be high school
and actually to be better eliciting self disclosure
also is not it your relationship between this process and it
so it's not expected that a higher amount of self disclosure that make someone like
you
so
that's has been studied for a conversation between human
but really interested to know if the same it's that is close to have the
thing i think in human machine i work
and if it does
that would have
implications for systems which came to elicit information from the user point o to maybe
also more pleasant nice experiences all the methods you task completion
but the key axes
two months that machines don't have sort since you things of that one so any
set this can also apply machine guns that is coming across as dishonest
by the starting point is that night maybe force that humans actually something you computers
a solution that so almost macabre thinking about n i the second source you to
stick study was split
in human conversation the human machine and or
okay so i'm
we are talking now about the context in which we can you contacted is that
so everyone who works in dialogue was how difficult it is together data
but
exactly in two thousand seven amazon had an example right channel
we had a noun in some university students to was channel one on amazon devices
and so this was a pretty one because
we actually get that uses the real world instead of
you know i think something expensive test it
so you want one of sixteen dialogue instance that hosted on the alex at the
right
at this could be able to use the united states of the command that's chat
and you that's what i didn't the data and it is seen that i don't
issues so they didn't know which of the dialog state tracking with
so
it looks like interesting because
i think it was that's what she to end the conversation at any time so
a data and happens
you domain so please specify started goals
so the only reason that even continue the conversation was for that when entertainment
so at the end of the conversation user is allowed only at the interaction the
scale of one point five based on the that they would interact with the social
want to get it also three shows more anybody
so three hundred and nineteen out of the fifteen or one thousand five hundred users
decided to anybody
and well known talking about the dialogue agent that we had in the next upgrade
so that i don't agent was based on a finite state machine architecture and what
this means that every step was a of the finite state machine
with essentially the response that we wanted to give as a dialogue system
and transitions what condition don't use the sentiment
so just an example of how this might look
so madness aside dialogue it and it's the entire house and green
in the users say something more stable be reports to like not but not by
then minus encyclopedia anything special happen if response posted but in the user to say
something that sounds naked
in tries to get a sympathetic response adidas them what's wrong so in this way
we had essentially the order of topics of the conversation
so maybe firstly the user then
acknowledge the positive or negative response then you try to talk about the initial
where we asked if they are interested in one of the latest tv shows if
they said that they are not interested that we asked them about a nice in
the if they say they were not interest in the movie i don't want game
in this example intuition that we show the movie we did spend some time chatting
about that it is that it is that you see you with the
then eventually all users what invited to leon you believe that out just choose not
appeared again and again what kind of one of these long winded word idioms and
they could also choose to exhibit limit in there
now a deterministic which was right
i don't need to impose taking initiative but al
all from the stage one with a conversation initiator shifted to the user
and they could talk to a dialog agent about anything and we try to get
responses from
other sources from the way
so on is
about i dialogue into
but the specifying chat bots that we used in this study would integer different so
we randomly assigned the users who interacted with the system
one of two chat bots or one of them i think in c is a
very high self disclosing chat board from the beginning of the conversation
so in that uses when the machine that's how's it going in the user sees
not i'm not the channel response with the story about itself with this kind of
your because
that's it is directly related to there's been chilling a to b and catching up
with my friends they just got them and whatever you an expression that they shouldn't
one
and the machine i mean of the dialog agent does not really had a frame
is often to work about it
but
then the humans experiment to see a play today i quite enjoyed
by a group of humans
the see a dialog agent that did not end of story about itself it is
c is all that's create anything a specialization towards technology and is the next question
so now on this was the setting under which we
and i'm not experiment
but not only interested in identifying when you with the we use i with exactly
that are dialog data
actually said is close and stick with defining all what we consider the self disclosure
so in the context of conversations with a dialog agent we said that self disclosure
has to be wanted at and it should be information di otherwise the dialogue agent
should market
so this does not include non-systematic questions for example so in the example given
so if you see when a system see what we collected anything special
and the use this is nothing that changes going on physically with my notes today
so that what constitutes a disclosure
but what exactly fifty times in the movie a time to learn the user sees
i need identity of a bit bigger block on dataset disclosure because it is adjusted
it it's possible question and that you have any extra information
so in this paper a three hundred and nineteen conversations were labeled for a user
self disclosure
and we manage to get a substantial agreement on there
by
we actually had a much larger corpus of conversations which it was not possible
human annotators to allocate every user utterance for self disclosure
so what we do this we built an svm classifier
are trained on this corpus to be able to the entire corpus for all occurrences
of the service goes
so this is the things just is designed to say one going to justify
but the classifier or a accuracy of ninety one point seven percent and or
f one score of sixty seven percent so it was fairly a little bit
okay so now we had
the user utterances which
one instances of self disclosure because we got a classifier to label the whole corpus
i mean or when a machine is close because we designed this is so now
that's allowed us to study the effects of status close to
so are the first one if you want to start with lexical
so
what we did this we studied
in how many users are disclosed and done before to sell disclosing user utterance
is the of the machines at school and we note that users was significantly more
likely to set the schools following of machine that the solution
then when a machine didn't set is close
you know at a door immediately following the for instance of machine self disclosure
and we found its users were much more likely to set it is close even
at the beginning of the quantization of the o
matching after a system set is close
so we set our study other question
the project and a conversations but initial user self disclosure actually longer
an internal yes there were a they were significantly lower
we also want to study if you that's what is close initially do they disclose
of the conversation and i don't know that if someone doesn't so close initially
much less likely to sell it is close to the correlation
so
i don't know question we started was a user will choose not to sell disclose
initially estimated as in just employing machine interest and this is kinda based on the
motion of people having no god it was not
so the way we tested this was a bit a reference to i would do
so we are able to set was initially one more likely to fail to one
and only if they do play the work in for how long deeply working
so yes a the users who a source not a set of is close initially
what actually much less likely to pay to what they also if needed in the
volume deflated what a much shorter time than
a user's we chose to say it's close
next we test the effect is close right in the likability in human
machine interaction
so since avalon provided us with sleeping of well with the people for a given
by the system again we use it as a proxy for online okay
and we thought we had a conversations make you would sell disclosed a lot i
mean the delayed of what vector
and
the and that is we don't know
so we wouldn't really find any correlation between over the user ratings and the a
mole this to disclose in the conversation then and we couldn't find a difference in
the ratings of the statistic located at what's as both within
and he couldn't find a difference in the ratings for conversations which have high service
also insist on with patients with discourse
and so it and a few movies
by we study the effect of set is close in a real-time that skin spoken
dialogue system give model being users in the real world of amazon alex that thanks
to a amazon
and what we found is that indicators of reciprocity have been even in human machine
quantization
and that
by the way how with are authentic as close as efficiently we can characterize the
behavior while the quantization
well what we also identified anything at relationship between san exclusion and like
thank you
first
the question so it's jules so well they're not aspects to self disclosure to adapt
of self disclosure elements of self disclosure and the study considers the only by at
o
future work would be to do this i'll probably but to do both positive and
negative as well or two of the data but the on sentence that i don't
more better we would find or relationship between self disclosure and liking
because i even in psychology it's not remote that
so that was the study in nineteen seventy three i think about what cost
would say that
so he divide itself is closer to three categories do medium and high essentially
and he found that and one set is closer to result in collecting but i
said distortion actually resulted in this and i think
and the reason is that i don't trust
people who say disclose very high
so that maybe
and interesting finding
was
i even like you an extra a i think that someone we gotta because they
have a block so well
i don't know anything that alex is hopelessly but an x and you might be
ignored
but they are this also five shows that people will be back to what we
so that also are similar
and are most of all black students were actually
the titanium believable that the machine could have anything like but
no effect on very clear people responded that wants to that's the
the actually believe that the machine thing
so we don't have instances of battery hundreds of people
what we did not particularly if they were different from the a believable back to
you
really of the reciprocity affect and reciprocity effects or and we would like to these
will store
first
so great question initially our work but only the shaded up to the
and we didn't really ask questions but then people never school us because they just
i know what to talk about rip of order
the initiative to do this architecture may be kept asking questions and we might get
one and i started do not include a direct responses to question
so even something like are not clearly a i even things like i did see
the movie we don't count those that self disclosure available they technically are giving information
because they're
just on sorting the question but the bare minimum of what require
so we only consider things with a probation it wasn't necessary as a as close
to
and that of the data collection
last
do you mean that is us
i think there had so the people are currently work
okay
okay i think