Player is loading...

Embed

Copy embed code

Transcriptions

Note: this content has been automatically generated.
00:00:00
you have to um talk to you know not hopefully stimulate further
00:00:04
discussion with the people um uh that were working full screen
00:00:08
um hum no measuring christmas so this is a a project that i've
00:00:13
been uh working on since the about two thousand and five
00:00:17
um and that's actually how i got in touch with the the up is that i heard there
00:00:22
seen a study published i'm in uh the gentle signs which
00:00:26
showed that people um to a naive about um
00:00:30
um pretend is a two on the political rice were able to reliably
00:00:35
detect you would when i'm snooty by rating the faces of all
00:00:39
and run up so when i saw the side try to replicate these results and actually added in switzerland
00:00:44
um and i try to understand uh why this was happening and we ran the study with little children
00:00:50
and uh lo and behold who actually reliably to able to
00:00:54
detect would win election races looking at fashion appearance
00:00:58
um and that really great keep it told me because um this was right after i got ten yeah then you know i'm a professor leadership
00:01:04
can you imagine a what i'm supposed to to teach if the ability to you'd already
00:01:09
uh depends on on how charismatic one looks so i started this research program to see whether
00:01:15
we could go beyond what um people usually do it's a heuristic process of observing
00:01:20
classifying and filling in the blanks so depending on how symmetrical the person
00:01:25
is on different features on that will very quickly put them in
00:01:29
a box and then uh put a price on the tag if they look like a million dollars they fill in the blanks
00:01:34
and assume the person has lots of positive characteristics if they don't look like a million dollars well they are pretty much
00:01:40
in trouble so the idea was how can we get people to go beyond initial classifications on the base of look
00:01:46
height six age gender ethnicity all these things matter very very much in
00:01:52
in leadership that's why i believe we have lots of incompetent people in positions of power it's precisely because
00:01:59
all all this problem and and plated written about a long time ago in the republican said
00:02:03
imagine the captain of the bow to store line stronger than any of the crew
00:02:07
is a little blind has a similar in for meeting hearing enters
00:02:11
knowledge of navigation is not much better uses allegory to actually
00:02:16
suggest that democracies would be doomed to selecting people who
00:02:20
look the roll but maybe we're not the best
00:02:23
qualified to do so so the whole um a research that i've been doing has been um
00:02:29
um i'm motivated by this we we managed to publish operating signs as a followup paper
00:02:34
to that article about thought robin and that's actually we don't need to huh
00:02:38
uh so the study in he was running some stuff on faces and then it would be tough to herve
00:02:43
a little while ago am i was trying to write in the l. c. ground which was not successful but we managed to
00:02:47
do a anything as follow up and then uh i got in touch with phil and that's where we are today
00:02:53
so let me quickly introduce you to the topic of charisma how we've measured that how we've manipulated and then
00:02:59
maybe see what uh we can do with that with that phil and and he's group so the
00:03:05
in my field charisma and a theory call transformational transaction leadership is very dominant
00:03:13
um but the problem is in my field is typically people use observation
00:03:18
all data to make causal inferences and which is highly problematic
00:03:22
um so for example when when when people ask a research is ask
00:03:28
um fall was of you just write how really to behave so accent how charismatic there those
00:03:34
ratings are very idiosyncratic that depend on many on model courses for example how you look
00:03:40
um for example if they know that you've been successful so they a lot of problems
00:03:44
which we call a indulge in eighty problems and originality means that the variable
00:03:48
better starting is not exhaustion is it's and dodge unless you depends on other variables
00:03:52
in the system of equations and may be omitted from the system up quite
00:03:56
so why isn't isn't as powerful and confident if you don't model that
00:03:59
then uh what correlation you find between charisma not comes maybe biased
00:04:04
um leaders may change the behaviour as a function of follow up before it's all how much they like the falwell
00:04:11
so we have lots of problems yeah um and using questionnaire measures
00:04:14
which which is traditionally done in mention signs in psychology
00:04:18
yeah they asked people to respond to question there and then they use um the
00:04:22
variance in the questionnaires to predict outcomes doesn't really help us very much
00:04:27
thereby spot lots of things related variables and what we call performance pure facts knowledge of outcomes
00:04:34
or knowledge of reputation biases the person in in how they write the um the target
00:04:40
um and in the other problem is that charisma has been defined in
00:04:44
really strange ways which is not amenable to scientific studies so
00:04:48
how does my means gift a gift of grace originally these people had a nice line to some
00:04:53
variety and could predict the future and had a gift uh to do um uh amazing feats
00:05:00
um so you know identifying charisma as a quality ability will gift of the that doesn't help because what's the nature of the skiff
00:05:06
i'm suggesting that these people like exceptional extraordinary exemplary
00:05:11
useful to logical by definition how charismatic
00:05:13
leaders inspirational and locating um you know as a whole bunch of stuff which
00:05:17
um interview that we recently did we found that that they were
00:05:21
not very useful top come so we never down things
00:05:25
that we could reliably detect and exhaustion is to manipulate
00:05:29
um to determine how where the charisma could actually metaphor anything um uh
00:05:34
at the organisational level so there's a whole bunch of stuff
00:05:37
um so i mean divine related i mean people are actually write papers about that i mean how can we test some of these things
00:05:43
um so the way in which we have to find kristen with my colleagues is that it's to do with signalling it's a
00:05:49
signal that is designed to communicate just like an animal communicates
00:05:53
a peacock communicate certain things by how which struts around
00:05:58
oh uh_huh symmetrically does so um you know the p. hey nice looking for certain things
00:06:03
which will suggest that uh the the the the the the peacock as um
00:06:08
something that's uh that i can help a and b. been officially reproduction of the same thing fall for leadership
00:06:14
m. m. signal has to be honest and the signal has to be costly you know flying around with a big tail
00:06:20
um is is not the most uh the best way to fly around if you're peacock you better of having a small it up
00:06:25
but if you have a big tell that symmetrical and beautiful and you can afford to have
00:06:29
produced that tail and then carry around suggest that you are strong and and fit
00:06:35
signalling charismatic leader also um has certain aspects to it which is costly
00:06:41
because um the signal is not amenable just twenty one to produce a one has to be able um to
00:06:47
have a vast repertoire of of artistic communication means which i will show you know a little bit
00:06:52
um uh ever started these you wrote a lot about um um the art
00:06:56
of of using metaphor one in any in the poetic it said
00:07:00
if i'm the those who are able to master the art of producing
00:07:06
symbolic and metaphorical communication mechanisms other ones who he would
00:07:10
in his books classifies being very very smart
00:07:13
um so it's it's the ability to communicate symbolically stand for
00:07:18
something represents something something that would trigger an image
00:07:21
um based on some values that you defend and this is very costly because the more these values
00:07:27
um target specific sector of of the population of people of the more likely you are to be
00:07:32
hated by by people who don't shave and you say you know oh bomber is objectively charismatic
00:07:38
um but you know he's completely hated by people who don't share use 'em ideology okay
00:07:44
so so that it based communication mechanisms also cost in a sense
00:07:49
that um it does sort the market but it does improve the thickness of the people listening
00:07:54
to speak why what the ones you like really that um no what the stands for and the ones you don't
00:08:01
like the data or also may appreciate that because the the this communicating coming information to them just like
00:08:07
you know if we're gonna be two balls about to fight over cow you know if i look heavier than you
00:08:12
um even though you know you're not gonna be able to reproduce with the cow i didn't help improve your fitness
00:08:19
by the fact that you know i'm posturing in a in a you know why that
00:08:22
signals high um have you full don't even try um to mess with me
00:08:26
so you know you'll go find another car that will be happy to uh deal with young
00:08:30
apologies for using such file and grotesque metaphors is just trying to get the point across
00:08:36
um so we have evidence um that charisma can been trained so we can take people
00:08:43
and we can teach them to behave in in ways
00:08:46
that resemble what we see as being charismatic okay
00:08:50
which means we can teach them to use small um effectively body language but also rhetorical techniques
00:08:56
and that's what i'm gonna focus the token today is how we speak what we said
00:09:00
um we know that it makes people seymour prototypical is you just um we know that in abstract lab
00:09:06
settings people exposed to charismatic leaders perform better than people who are not exposed to charismatic leaders
00:09:12
um it matters much in situations of attribution ambiguity
00:09:16
um what does that mean um we have a model to predict the us presidential elections on the basis
00:09:21
of macroeconomic data and also who's more charismatic and and what turns out is that when macroeconomic data
00:09:28
do not sufficient to give a strong enough signal as to
00:09:31
whether we should reward or punish the incumbent party
00:09:36
that was when performance signals of fuzzy about the state of the economy then people look to see who would make
00:09:42
the better commander in chief who to trust me essentially who's more charismatic so we actually do pretty well
00:09:48
in predicting our sample because presidential elections um on the model that we've
00:09:52
developed that goes back to nineteen sixty now obviously we have very
00:09:56
few data points um so we have to be very very k. from econometrics all that because elections only happened once every four yeah
00:10:02
um it correlates with effect in the south comes up the score let's means nothing to me because the correlation
00:10:09
if it's antigen eighty plague doesn't mean anything that correlation could be due to limited
00:10:13
calls and in fact maybe there is no correlation it's just the speeches correlation
00:10:18
so we don't have any colette causal evidence in my feel believe it or not even know max weber
00:10:23
first wrote about uh charisma uh about forty fifty years ago um zone uh wrote about it
00:10:28
in biblical writings about a hundred and two hundred years ago it's also being found in
00:10:33
in um in different references uh in in ancient greek mythology the graces and in care as well was one of them
00:10:40
but there hasn't been a well designed field experiments showing that it really matters
00:10:45
closely for performance outcomes sorry do show unit field experiment we did
00:10:50
some experiments that we replicating the lad and then some observation all data that
00:10:54
we also gathered and that's when working on with with l. i'm too
00:10:58
to show how we haven't revel what charisma is and and how
00:11:02
we can measure it um and and manipulated so the idea
00:11:05
i guess we you use more about the identifying features and how we can reliably detect them using some kind of a elder them
00:11:12
um so in the first study this was done by um uh my colleagues uh china data uh um
00:11:19
christian send them a better whether they're economists a foot and it really
00:11:22
matters to show the economic case about why why charisma ease is
00:11:26
essential is very nice to say well you know like this person's
00:11:29
possible great product typical are you willing to exert costly effort
00:11:33
to help this um the the really fight will conquer ties the division so what we did is we went to birmingham still
00:11:41
charity in england conveniently so because she was doing a post doc inch was located in bombing them at the time
00:11:47
and used in a post doc with uh roberta weber and um we help
00:11:51
them to raise some money um by hiring workers to stuff envelopes
00:11:55
to stand up to do a big mess mailing in england so what we did is we were to add
00:12:00
the code at the code um has um um i don't know if you know the blue uh
00:12:05
like the the help of people um uh to to get intemperate jobs and job placement
00:12:09
so they have a database of lots of people who are out of a job and you don't want to do a little uh job here and there
00:12:15
um so we went to add the code and um they
00:12:18
put advertisements through the usual channels to hire temporary workers
00:12:22
what what we called for a mail sorting a task the what
00:12:26
is the don't know who they were participating in experiment
00:12:29
and but still we do not use deception i think i need to say this very explicitly
00:12:33
um because economists and cognitive and a dash of psychologist like like uh like i am
00:12:38
uh typically um when we do an experiment we try to be very transparent about about everything and
00:12:44
we don't lie to people by by saying things that an auction in here we ops
00:12:49
for kate to the fact that we um we using different motivational resumes so you
00:12:54
know imagine you come to work for me and i might expose you to
00:12:57
one kinda manager who briefly in one way and another can't imagine bush another
00:13:01
way we wanna see whether that will affect one person's performance um differentially
00:13:05
um so what we did is we put a recruitment campaign up um we were gonna
00:13:09
pay the minimum wage twenty eight pounds for for three hours of of work
00:13:14
plus training class we paid um um for um a transfer compensation for one hour travel
00:13:20
and the idea was that they knew that we're gonna come in workforce put three hours and they would never see them again
00:13:26
so the idea is can we set up an experiment um that will really
00:13:30
test in a strong way the closer the impact of charisma on outcomes
00:13:34
so what we did is that workers had to assemble different things i'll show you some some pictures in a minute
00:13:39
a fundraising letter envelopes the nation cards in a large envelope um you know we pay them for a certain amount
00:13:46
of time and that was guaranteed wall workers got the minimum wage okay um they were briefed by nectar
00:13:53
but there was sorted in one of three conditions randomly and they assemble the mature at home
00:13:57
so it was done without supervision that's very critical um so if we wanna test
00:14:03
with the something really works and and and creates this warm glow intrinsic
00:14:07
motivation that people suppose well then we're gonna do it without
00:14:11
um i'm checking on what they're doing but we will check the output and we will check the quality of the output later
00:14:17
so this is basically the room in which we brief the work 'cause i'm the acting came in here they were
00:14:22
briefed uh in batches of eight to ten they had two bags each weighing about five kilos about eleven pounds
00:14:28
um we protested to see how much an average person uh could carry for
00:14:32
about half an hour um because these people must using public transport
00:14:36
so the uh the the experiment to show what job i had
00:14:39
to do back to comes in and just five minute manipulation
00:14:43
um on how they should do the job in the importance of the double show you the video in a minute
00:14:48
um some of the letters we put we had printing errors on on purpose
00:14:53
and actually it happened kind of random is with you know with plan for it it happened nonetheless
00:14:58
um and the idea was that one of the treatment conditions has monetary incentives
00:15:03
to do well so because they monitoring centres it's there in in in their interest to stuff anything in the envelopes
00:15:10
even if it's not a a good quality data so we gonna tell them you know please sort out the bad quality letters
00:15:16
right on the for what the problem was a little problems of all the lobe so probably the for the problems with a brochure
00:15:21
so you know that we we we set up in such a way that it was impossible for
00:15:25
anybody to finish three hundred fifty units in three hours in fact the average worker did
00:15:30
in a pilot test that we rent two hundred units in three hours to just play that in like two
00:15:36
hundred units used in three hours is what the base rate is so what we're gonna do is um
00:15:42
we gonna observe the performance um in these good situation so we slow them down
00:15:47
by putting mistakes inadequate staff things fall them in particular things we have a twenty four
00:15:51
point checklist um and we gonna randomly sample from the workers once they give us
00:15:56
um but the the the labour the fruit available back we gonna randomly sample to see that they followed all instruct
00:16:02
okay so the gullible that staff put it neatly folding branches put it in the box and then come back the next day
00:16:09
so we randomised the workers to one of three conditions everyone has a guaranteed fixed which so you do
00:16:15
what you did you do three hundred and fifty units you'd he's gonna get the minimum wage okay
00:16:21
um what we manipulate is whether they get a standard speech what charismatic speech which
00:16:26
i will explain to you in a few minutes the charismatic um um speech
00:16:30
yeah is now gonna be compared to a standard speech when we get the p. straight save the wood
00:16:35
'cause stuff more than two hundred twenty overlaps we will pay them a little bonus of twelve pants
00:16:40
um yeah that's about twenty cents so five when perhaps give them a frank
00:16:45
so the idea is yeah we we mathematically figure out what to
00:16:49
set the bonus add to keep the cost per unit of the same as yeah because this what is gonna a lot more money
00:16:55
but we need to maintain the cost per unit approximately every talking up a bonus of one pound that
00:17:00
break the back strong with to do it so we figure out the maximum bonus we could pay
00:17:05
to to stimulate productivity such that in economic theory suggested approximately we should get a
00:17:11
twenty percent increase in performance from the street she but the question was
00:17:16
if instead of giving a standard speech if we substituted the standard speech or the charismatic speech and
00:17:23
still pick the fixed wage could we introduce some um extra performs now we do have
00:17:28
a a mathematical model and some theory to explain and what our predictions are it's not here
00:17:34
um i don't have in here just so that we can focus on other things but we do have some some very specific production
00:17:40
um so the subjects were on like this um we randomly assigned them to
00:17:45
uh uh um five uh three session blocks of the course of the day it's session last about thirty minutes
00:17:51
um and then um they summoned signed up to come back the next
00:17:55
day with a with a stuff so the speech regulation followed
00:17:58
some of the previous uh basically the research that i've done we we we
00:18:02
we we train people to be charismatic and we did some lab experiments
00:18:06
and then also in the study that i told you we we predict the us presidential elections
00:18:12
the speech is had the same amount of words said the same thing at approximately the same time
00:18:17
so we have exactly the same the mind effects on the subjects but
00:18:20
what we manipulating the charismatic speech is that we use mall
00:18:25
i'm a symbolic communication mechanisms visionary symbolic communication one show you an example of this
00:18:31
so the idea is we don't wanna have a confound yeah by changing the information content and the delivery
00:18:36
we try to hold the content to saying change the delivery mechanism
00:18:41
in terms of body language use but also the visionary feeds
00:18:44
um the two speeches then um had the same number of words very similar content but the christmas
00:18:50
speech relies more on certain features which we have
00:18:54
identified to help trigger visual imagery is
00:18:58
um frame the message and and defend some kind of substantive what
00:19:03
would be a logical a plaque from so for example
00:19:05
we use metaphors metaphor triggers an image makes it easy to remember you
00:19:10
can understand much better what the person is saying so for
00:19:13
example we compared the letter to a ticket to attend christmas and we went back to the stick it metaphor coupler time
00:19:21
um we use stories and anecdotes you will see a in a couple of minutes when i'll play a section of
00:19:25
the speech you know too much of a story of an old man now when you tell a story
00:19:29
people get captivated by the story if it's told properly they see themselves in the story they identify with the protagonist
00:19:36
and then they remember the moral in a visual and easy to understand why
00:19:40
contrast not yet to do this you get to do that ask not what your country can do for you but
00:19:45
what you can do for your country even you know george bush who we studied also in a in a
00:19:52
in the yeah study on on on a on a on presidents you know
00:19:55
it's had some memorable sayings either you with us on the water or
00:20:00
what does everybody with that's all
00:20:03
against us yeah exactly so you know these things make it memorable and define what you should do this is what shouldn't
00:20:09
be makes things black and white helps to frame the message so force is really important to frame the message
00:20:14
point to where you want the people to listen to but using visual imagery is posing questions using
00:20:20
contrast and then defend the substance and then the liberty in a in a way that's energetic
00:20:25
i'm posing rhetorical questions create sink sleek makes a person look for the answer
00:20:31
saying things in three it's it's much more easy to remember three partners than a twenty seven populist
00:20:36
small memorable it more rhythmic i'm expressing the sentiments of how people feel putting into
00:20:42
words which closes the psychological distance between the order to and the person listing
00:20:48
setting goals that uh makes people feel that they are able and
00:20:52
capable and increases the feelings of of a efficacious nuts
00:20:56
i'm giving concern is that the goals can be achieved and then various nonverbal techniques i'm not gonna focus too much on that because we're not
00:21:03
for the purpose of what we're doing that also interested to uh right now we all would
00:21:06
like one day to incorporate the nonverbal techniques but first we gonna focus on the record
00:21:11
moral conviction is also something that's the values that i told you that the defence
00:21:16
um but because this fundraising campaign was done in the context
00:21:21
over charity we hold constant d. um moral conviction
00:21:26
in both speeches okay 'cause we don't
00:21:29
want that just simply because was a model task in that we use more references
00:21:33
that that was driving there's also with exactly the same amount of more of the monsoon both speeches we hold back on
00:21:39
okay just to show you an example i'm gonna just for back um to um the the speech
00:21:45
made by bark a bomb uh a little part of it at the democratic national convention
00:21:50
so many just play through and then i'm gonna analysts it for you to show you what kind of features we look for okay
00:21:56
um so here we go so this is done now one week after the republicans have
00:22:02
done their um convention and um i'll buy my is talking about the republicans convention
00:22:11
uh oh i guess ah or or or or or uh
00:22:27
oh
00:22:33
uh
00:22:40
yeah
00:22:44
ah ah ah ow
00:22:49
ah
00:22:53
ah so he's in a lot of these uh techniques that we've uh been
00:22:59
identifying over the years um i'll just break it down first uh
00:23:05
i'm in a in a kind of object away and then just recap we set so firstly if you
00:23:11
notice the speech is broken up into three three heartless that's an interesting um thing that he does
00:23:17
um he uses and and anecdotes so immediately he says what our friends down in tech in tampa
00:23:22
yesterday or the week before so everyone knows what the republicans would doing that and typically
00:23:26
what the republicans do at the conference at the convention what's the single thing that wish to
00:23:29
defend is we need to reduce taxes so if economy's doing well what should we
00:23:35
he just acts if she's going badly what should we do reduce taxes says like these people doing so this is one big metaphor
00:23:41
yes the doctor the republican the passion as a disease we given medicine the patient
00:23:45
does not exhibit the disease would give the saying madison this is ludicrous
00:23:49
so it's all one big metaphor okay and in english when the doctor doesn't know what you have what do they say tactile aspirants an
00:23:56
call me in the morning so this is with us last saying come from so yes an anecdote which immediately triggers a visual
00:24:02
and what the republican stand for you does need to explain it explicitly everyone knows what this is gonna do what
00:24:08
any you they separates what they want from what i want to make
00:24:12
his position very clear that's what a contrast they want to vote
00:24:15
but i don't want you to know the plan and that's because all they have to face the same prescription headphones that
00:24:21
so that's where the metal begins with rhetorical question rhetorical question
00:24:24
rhetorical question on so on so on set using metaphor
00:24:28
and one other technique which is a bit dangerous to use its human we not coding for
00:24:33
that because it's a little bit difficult to detect um um at this point in time
00:24:38
um so we wrote the speeches i'm gonna show you the speeches in a minute um and then we
00:24:44
have trained code is and this is way i hope the the apple able to to help because
00:24:49
this takes a long time i have a coding manual which i should by the way make available to you which we used to
00:24:54
train humans you take several weeks to train a human to be able to extract these features were lightly and correlate highly
00:25:00
with um me and and and the future car and and the benjamin to have
00:25:05
who developed the score this manual so we use what's called the
00:25:08
cup the statistic it's a way to quantify agreement between writers
00:25:12
so what once you wrote the speech we gave it to the strain code is to see
00:25:16
how much we could um um um um explain the difference between the two speeches
00:25:21
so for example the non charismatic speech had forty one sentences um we had a pretty
00:25:26
good agreement um there is very speech would yeah it's a pretty good agreement
00:25:30
and we had pretty high kappa statistic what's important to show here is that non charismatic speech
00:25:36
had only sixty number forty one sentences that were charismatic about thirty nine percent
00:25:40
and the charismatic speech was about it ninety one percent this difference in proportions was
00:25:44
highly significant we also looked at nonverbal behaviour facial expression and body language
00:25:51
um and again we had very very significant meaning you'll a
00:25:55
signalling um from the body language so um we
00:25:59
also did a check way we asked people to watch the videos that were not part of the experiment
00:26:03
to see whether they sold differences in the speech is now these two things measure moral conviction
00:26:09
and notice there's no difference between those two speeches because we help the constant so that's important
00:26:14
however the charismatic speech was more inspiring and and the person
00:26:19
he out was much was seen as a mall
00:26:21
i'm i'm i'm a prototypical either nice very important i'm running
00:26:25
an experiment yeah weep that's costing us thousands of pounds
00:26:29
and we don't expose the workers who have the non charismatic speech to really cracked speech you gonna see it now it's not
00:26:35
a bad speech at all in fact it's a very good speech so what we do is we comparing are really good
00:26:40
non charismatic speech with a with a charismatic speech so let me let me just show you what it looks like this is the actor
00:26:48
user name is richard activity is great grandfather was actually
00:26:52
clement aptly prime minister in england just before the
00:26:58
second will will
00:27:00
so i'm gonna show you that the introduction the middle part and the conclusion so i'm just
00:27:06
taking a snippet of about one minute from each speech and the marking up the features
00:27:11
that the act to uses what he's saying them only the rhetorical features the
00:27:15
body language is quite obvious you gonna see when he's doing it okay
00:27:19
um and then we again i'll show you what happened in experiment and i'll show you some of experiments that we've run
00:27:26
okay so here's the introduction from the non charismatic version of the speech
00:27:32
you know because let's yeah all the time all the oh yeah
00:27:40
sure yeah oh because it will release was was now
00:27:49
this is on this list which ah it's a ah well okay for you you know
00:28:01
okay so the charismatic version of this
00:28:07
ah okay
00:28:10
well let's see ah over the uh oh yeah
00:28:15
and where there is that's for sure and from
00:28:21
the chewed her lip i grew so
00:28:27
huh huh huh huh huh huh huh what are your goals
00:28:38
okay from the middle part now which trying to get them to stop as many on the lips as possible
00:28:44
well well well well well oh yes or no
00:28:52
this is for real or as well
00:28:58
uh it's where the first one or or or or was it those are great
00:29:09
yeah oh yeah yeah yeah maybe also one is ah oh
00:29:19
so who has to go well that's yeah that's what it was a real or not
00:29:29
oh this was to so would you agree this is not a bad speech or
00:29:37
not a bad speech at all how can we put some charisma into that
00:29:42
well well well one room
00:29:48
yes
00:29:51
it was a raw raw oh or are you still for c. uh huh huh
00:30:02
you should ah ah style
00:30:08
or as uh huh uh huh uh huh ah ah ah ah bee house or
00:30:21
or a ah
00:30:25
ooh this and uh well
00:30:29
uh_huh ah
00:30:33
'kay the finale
00:30:38
just before the works out well for that so they work and what has okay the last
00:30:49
please do this yeah 'cause this one oh oh sure
00:30:56
mm or or or we will he walks out so we don't have any other interaction of
00:31:03
contamination between him and the group so it disappears is i gotta go now and
00:31:09
uh that is uh
00:31:14
you should yeah
00:31:17
so er works for it
00:31:23
or or or or or or or
00:31:29
or was it
00:31:38
mm uh
00:31:41
or
00:31:43
so
00:31:45
or
00:31:48
how to roller group right
00:31:53
when mine condoms friends all this they're like walked whispering normal
00:31:57
ups flying starvation plastic it's crazy no way you know
00:32:02
i to fly out there to train the act the way to pay a deco we had
00:32:05
to pay a you know where to pay the bill for printing because the work is
00:32:10
anyway well managed to convince them that we run this um so everything was done as you say this was the font
00:32:15
the studio what he did we also folding doing it live with a with a camera in
00:32:19
front with a there was a on on the computer um so we ran the study
00:32:25
and are low and behold the baseline group actually got more than two hundred and the pilot which
00:32:32
means that really we didn't give a bad speech was not a bad speech piece right
00:32:38
coincidentally really i'm incredibly of followed very closely economic
00:32:43
theory this is a paper from it would
00:32:45
lead lessee uh a very well known and they they converse to actually was uh
00:32:50
sharing a bush's economic council we've got twenty percent exactly way
00:32:54
economic theory would predict but this we were very surprise
00:32:57
charisma was able to um uh uh um i get this statistically equal results
00:33:05
and and around but yeah we don't pay the bonuses at these workers on
00:33:08
a lot more money um and yeah we don't pay anything um extra
00:33:13
this is the interesting thing so for the baseline group the cost was about fourteen pen spawned a lot
00:33:18
slightly higher for the peach right the reason why this was size because the baseline group did more
00:33:22
than what we uh expected him to do we think about two hundred two hundred ten
00:33:27
but here's the cool part we reached the cost per units by about nineteen percent
00:33:32
why we got the same performs as speech right but we didn't pay the bonuses these workers and i had a lot more money ten
00:33:37
twelve twenty pounds more on average and all the other workers okay um so you can see some us the the more um uh
00:33:45
more information yeah so you know about ten pounds some and a lot more 'cause they went all the way up to three fifty um so you can see
00:33:53
how many we have maximum number of let's completed not remember somebody's letters uh had had mistakes in them right
00:34:00
not is the interesting thing the average quality index was no different they
00:34:04
all did the job very seriously which is very important to show
00:34:07
that this was not a failed experiment in that the workers were not
00:34:10
just doing whatever what we kind of captures intrinsic motivation okay
00:34:15
and cost the lead a um on average relative quality of of of perfect quality okay
00:34:20
um you can see this is the distribution of the works performance and charisma pushes
00:34:25
a bunch of people right right to maximum performance that that that um um as compared to the other treatments
00:34:32
so what we try to do then is replicate these results in a public good game where we have another trained actor
00:34:38
it's gonna brief foreplay as well playing anonymously that or no
00:34:42
who they play is i will will either have lintel computer the computers are talking to each other
00:34:48
you're gonna see the same video which is gonna make the demand effect for you
00:34:51
to contribute money to public good so what does this mean in economics
00:34:55
we wanna capture whether you are gonna be identifying with the the because what the the the the sense to you but also
00:35:01
what your belief is about what you think about people will do when you have a coordination problem okay so what happens
00:35:07
we give you money twenty points twenty points that we convert to money and then you have the choice
00:35:12
you can keep this money in your pocket we can put it into a public a cow
00:35:16
if you can't we will multiply it by certain multiple kinda and then give it back to everybody
00:35:23
now the practical solution here are the best possible so i'll come for everyone is that we all put our money back
00:35:29
this man is multiplied and install divided but free is the dominant strategy is to hope that
00:35:34
these three of the dice are dumb enough to put that money in the pot
00:35:38
you keep your money the money gets multiplied and then you give a shit back
00:35:41
of the multiplication and the money okay so this is the dominant strategy
00:35:46
everyone is not so stupid 'cause we've windows domain strategy and what happens is as soon
00:35:51
as i see someone defecting 'cause we're gonna play this game over twenty rounds
00:35:54
you know people are pretty much good well they're gonna put some money in the public account
00:35:59
but if they see that the arsenal putting it straight away i think i'm not
00:36:02
a needed i'm not gonna get taken for right and they immediately defect okay so what
00:36:06
we're gonna do nice compared to standard probably goods again we we have no
00:36:10
two moms on them with the charismatic speech to one with is a charismatic speech and the control speech to see what
00:36:16
happens to contributions over time so we have three different groups
00:36:20
so here's how we convert the points here the readings
00:36:25
place observe the decisions of group members after every round we also asked them
00:36:29
what do you think your group members are gonna do next raw
00:36:34
okay so we measure the beliefs and they preferences may reveal preferences by what they do
00:36:39
same acted delivering um there's a similar speech uh we have exactly the same
00:36:42
thing as before but now the speeches about contributing to public good
00:36:46
um same number of words should be the same amount of facts we go through all the grim statistics and all that stuff soul that's okay
00:36:52
i'm just so you can see we have twenty four tactics used in the charismatic speech and non in the
00:36:58
non charismatic speech so this is what we manipulated um so just give you money to read this
00:37:24
ah
00:37:37
so the way this is written is typically how you would hear a micro economist explaining the
00:37:42
tragedy of the commons okay and how can we be checked is a little bit more
00:37:46
to talk about in terms of the tragedy that comes of your the tragedy of the commons opposes a common grazing wrong
00:37:51
for farmers to share a common interest is to protect the grazing got so it
00:37:54
would covers a new grass can draw on it it is in the common
00:37:57
interest of the farmers to cooperate in finnish underground because with cooperation they can all
00:38:01
benefit however each farmer may have a selfish interest rate is uh how
00:38:05
cookouts grace as much as possible but if all the farmers do that the common good will break down the field will turn from a screen
00:38:11
cross in too much so much more visual but given exactly the same information
00:38:16
we have the same actor i'm a young guy non charismatic version charismatic version
00:38:21
and we observe that the contributions over time uh as a function
00:38:25
of the treatments so you see we increase by about ten percent these
00:38:29
countries now this is not the individual effort this is trusting whether
00:38:34
you believe that when lisa started it was gonna charge at the same time
00:38:37
so this is a very very strong test now the difference is between
00:38:42
be a these three means is not significant so we can understand why and we running up some follow up a
00:38:47
experiments to see but just to show you how quickly the public good breaks down in the control group
00:38:54
so you see when you put the non charismatic speech it does bring it up so this is significant difference between these two and the controls speech
00:39:00
but this time there's no significant difference between the two speeches so it does break down also but that
00:39:05
but we do sustain a lot higher yeah it is the initial around a lot higher contributions um
00:39:12
we didn't have the experiment way we matched what contributions then make it probably good to charity
00:39:19
just to compare again the two speeches and again the charismatic speech testing to do something
00:39:25
but again here we're not measuring individual effort we measuring beliefs
00:39:29
and preferences and actually selfish uh whether we can affect they um they on um
00:39:36
utility economic utility okay so um we're running a now another study where we will make sure
00:39:43
that everyone has access to the same video because in private booths analyse with the
00:39:47
other c. perhaps case they may not think that everyone is for the video so one
00:39:51
treatment be doing is that they're gonna see the video together so for people
00:39:55
not for people everyone in the treatment before we randomised them there was you all the videos together and then
00:40:00
we randomised exclude does everyone knows is all the videos so that's the first thing we gonna do
00:40:04
and the second thing we're gonna do is we're gonna do an individual effort tossed in the lead we can perfect observe what they doing
00:40:10
um to see whether we can affect this significantly but anyway this to experience i don't consider them completely fail but they
00:40:16
still line forming us on how we can better test this so in terms of i'm studying i'm naturally occurring charisma
00:40:23
and this is what the data now that we shared with phil um we took a random sample of two hundred forty tight tops now
00:40:29
um why ted talks they done in a similar platform they have um a similar
00:40:34
structure similar timeframe um all these people highly coach so i you know it
00:40:40
still but we have huge variance in the features in the talks so what we're gonna do detect random sample the talks
00:40:46
we just take the transcript of what they say and i give it to three code is to extract the features
00:40:52
so we know how well the code is code with each other we have those mistakes and all the little works very nicely
00:40:57
we control for how famous the person is by looking appealing from that we keep your page a few months before they gave the talk
00:41:02
because you know it by the to talk it was gonna want to take pop up because that feels good but because it's been
00:41:07
so we also take photographs of the person's face and i run a separate study on how attractive they are
00:41:14
because we know you look more tractable sexy people are gonna just click
00:41:18
on it just to see a more statically pleasing human being
00:41:21
just that by itself predicts highly we the people are gonna be a gonna go and watch
00:41:25
the table so we control for everything in can imagine where the top was given
00:41:29
um at the g. d. p. of the country the population of the country in which it was
00:41:32
given because i could also drive of used interaction do the things the colours they wearing
00:41:37
the job they do i mean is it intrigue and things because of course i haven't manipulated the so they could be indoctrinated
00:41:43
but here's the interesting thing you put twenty more of these charismatic tactics in speech
00:41:48
um you increase used by forty four percent which is a huge effect
00:41:51
so the average tech talk is about here in terms of charismatic tactics which is weird i mean this huge variance
00:41:58
in these things and you want to know what what the hell are these people doing um
00:42:01
you know when when they put them up in the train them to give each tool
00:42:05
um so this is not what we have allies into c.
00:42:08
can the computer um correlated highly with the three humans
00:42:13
you did this and also predict the views independently of the human so this is what we're doing right now
00:42:18
um we replicate this whole saying true that we follow thirty politicians and thirty c. e. o.'s
00:42:24
over three months we talk about three thousand treats randomly from with a treat it
00:42:29
and then we have a panel model all the time so we control for the number follow was
00:42:34
they have because the number follows is for sure gonna to determine if you get retreated
00:42:38
we control the number follows the have whether the politician or was yeah man of female hold
00:42:43
the ah the job they do what's in the tree to have stacked the video
00:42:46
link whatever a menu is a billion things we control for glass the fixed effects that
00:42:51
are common to each person to see the marginal effect of putting one treat
00:42:56
uh in a in one treat a charismatic tactic normally hundred forty characters it's like two three sentences
00:43:01
but still you can put a metaphor in a rhetorical question um you can
00:43:05
still show you more conviction so this is another data set that we
00:43:08
have um i'm probably a bit more tricky because there's less context in
00:43:12
it but that would be really cool test with the neural networks
00:43:16
again you put a treat 'em a a charismatic tactic in a tree to increase retreats by about thirty two
00:43:22
sandwiches uh it's something which we believe is pretty consequential so now um we're gonna
00:43:27
we're right now um by setting up an experimental put a study where um
00:43:35
well i could maybe say is anyone onto the yeah so it's and the one that you okay well
00:43:40
anyway you want no but we setting up accounts read accounts where we're gonna start getting forest
00:43:45
and and then one account will be treating all the time charismatic he and in a separate account the same content
00:43:51
but not charismatic lee and then we get it set up another to cause we were gonna switch on
00:43:55
a daily basis and then was gonna take a real accounts of real people in intervene once in a while
00:43:59
slipping a charismatic treat uh to see that that increase the repeat rate so we're gonna does experimentally
00:44:05
we've set up a whole bunch of field experiments that we gonna study um how to do this
00:44:10
um um how to um um to study how to better train charisma in people so
00:44:15
we we can see its effect on outcomes um so raise many questions um
00:44:22
does it assessed et cetera et cetera so well these are
00:44:25
very interesting backpack fat for feed most interesting thing is
00:44:29
from a cat archival data can we get a computer to reliably do it because it is
00:44:35
error prone when humans do it it takes a lot of time gotta train them um so you know if we
00:44:40
can find a way to do it in the idea is that we develop something that we can also
00:44:44
i'm sure papal on and and have a an online platform so people can upload the
00:44:48
speech and then the computer would be we immediately return how charismatic it is
00:44:53
okay so this is now um with the ground that was starting to do it
00:44:56
so until we meet again this is my new slogan inference by the birds
00:45:04
oh yeah that's it thank you
00:45:13
so apparently this uh i should finish five minutes ago but so we've got ten minutes for culinary and
00:45:19
i'll be there for lunch as well so i'm happy to take any questions which you might have
00:45:28
so must uh got to compare s. and you know how you
00:45:34
and then gentlemen
00:45:41
so you know
00:45:44
or these are known in yeah yeah remote t. v.
00:45:50
like the speech recognition
00:45:54
really
00:45:56
yes
00:45:58
how we
00:46:00
well actually we account for it fun enough by using the
00:46:03
comp model so we're watching the dependent variable as account
00:46:07
which isn't on the new models so that's the first thing i don't know if that's what you mean
00:46:11
the second thing is we don't we don't care about the audience effect we don't even look at them with only the model them
00:46:16
and and our i just don't even look at the videos because that would be highly biased by what
00:46:20
they see is going on in the audience clapping laughing whatever so we only had the transcript
00:46:26
and then you look at the transcript so we're looking at the basis of the communication which is in
00:46:30
the rhetorical structure now we do know that and you're right about that that what you say
00:46:35
and how it's delivered is actually just correlated in does have a can have an impact
00:46:40
on others and by contagion effect you know that you like and for all the nickel wow you know it you know that's gonna
00:46:46
making the bingo yell a lot you know and it it does but by contagion up but we don't care about that stuff
00:46:52
what we care about is the number of use and those number
00:46:55
of use theoretically should move the mall charismatic the speech it's
00:47:00
and the more sexy person wrote the most six of the tightly so you know we
00:47:04
we we control for these things weights being a classified and all that stuff
00:47:09
um so yeah i mean we don't have all these things and especially because we don't
00:47:13
manipulating we can control for that but i don't know if that answer the question
00:47:21
yeah
00:47:23
the way they are more it's for every
00:47:28
the north because it's gonna be highly by seeing humans were coding these things
00:47:33
okay now it's possible someone gave a great talk okay and everyone talked about and then and treated and all that stuff no
00:47:39
i'm not i don't measure that at all and it's it's likely that's gonna drive use i'm just measuring the basis that
00:47:45
the the the basis of the top which is this on this the syntax the
00:47:49
syntactic construct of the sentences and just isn't productive use which is crazy
00:47:54
controlling for the face and and all the other thing the fame of the person so we don't even care about that and
00:47:59
the fact that we predicted is crazy just yeah we could relate better if we did what you're saying h. t.
00:48:08
for up these things yeah true true
00:48:14
yeah yes yes it's incredible animal sex interfaces the move use you're gonna get
00:48:23
yeah
00:48:25
what
00:48:28
yes sure yeah
00:48:31
yeah
00:48:35
uh the gentleman here from him
00:48:41
uh_huh 'cause charismatic
00:48:45
so right
00:48:49
but
00:48:50
oh yes oh yes or no or
00:48:59
yeah there's a there's a good reason for that stuff on the extremes we have few
00:49:05
observations of the conference it was a higher it's not because it might backfire
00:49:09
yeah it's not that it's a it's goes like this it's just that on the extreme extreme extreme so i i mean i just put it would be
00:49:16
two standard deviations um to show and three send you the actions show what happens it expects yeah
00:49:27
well the question
00:49:30
and you know another question i think i could use a few more words about removable party but you you have focused yes yes
00:49:43
so i'm i'm so as i was saying about my marion should witness that we're talking early on is is working with that danielle on this
00:49:49
um and we actually we can work independently was with mari and
00:49:53
um to try to documented typology of nonverbal gestures which
00:49:57
he does may use to signal certain things okay like you know what i just did now we're you know
00:50:03
i know you can do it um you know what does that mean um so what we're trying to do
00:50:08
is try to get a typology of things that you just do that goes with what they say
00:50:13
um to see whether this helps in communicating the message and wife active
00:50:19
man i'm so in the speech that i have with that young
00:50:22
guy i actually have got eight uh of a four versions of
00:50:26
the speech i have the charismatic speech with charismatic nonverbal behaviour
00:50:32
non charismatic speech with charismatic verbal behaviour so you know you saying he's
00:50:37
exactly the same thing but is doing um and he's saying
00:50:40
exactly same thing charismatic the in non charismatic the but using either body language or not to to try to get it
00:50:47
the quantification of how much actually the nonverbal signalling max's irrespective of the speech and does it need
00:50:53
to go with the particular you know when i say i know you can do it
00:50:58
because very well with that but you know if if i say something else like you know the
00:51:02
pool networks are really cool okay i mean maybe does go with that a little bit
00:51:07
um you are saying that something someone evil to certain gestures they're not
00:51:12
because being given that uh rica's beaches so performance to the politicians find there's
00:51:18
no recorded we think of the right of trouble or possible yeah trough
00:51:24
just an example i think that's great you you do good you could imagine having these
00:51:31
vocabulary of gestures a very specific busy oh yeah and and i'm gonna behave based charisma
00:51:38
measuring method that you can what i'm i don't know it doesn't but i mean it does that all the time but that's a bit annoying the way he does it uh
00:51:44
and it's a bit repetitive and when you listen to me it's kind of i feel very
00:51:48
uncomfortable after while because it keeps on repeating very similar gestures and i keep something there
00:51:52
so he i think that's an example if we were to measure him in some kind of way with your
00:51:56
knowledge infected do little but that's cool but you did too much it's probably gonna not uh not well
00:52:02
and and one thing just to mention daniel for the us elections we went back to
00:52:06
nineteen sixteen and got the transcripts of the to pretend this to the white house
00:52:11
democrat republican and and the transcript we we couldn't use the transcript for all the elections because in the nineteen sixty we
00:52:18
had no this you there's no voice recording there's no um
00:52:22
video recording i started actually in the kennedy d.
00:52:26
nixon election line him sixty ones and i guess you somewhere there um
00:52:32
so so to hold everything constant we only use the verbal
00:52:36
and still with available when economic signals all week so we interact exxon take
00:52:41
the economic production with the charisma when economic production signal is is
00:52:47
you schooling the election fifty fifty this when charisma really matches so like we we
00:52:52
we we uh so even if we don't even look at the nonverbal
00:52:57
we still do a very good job predicting why so we think we not
00:53:00
sure but um that is research suggesting that when you communicating symbolic ways
00:53:05
that triggers something in your new rule real you you brain and and and kind of comes
00:53:11
out in the body so if i start telling a story very visual story you know
00:53:15
and the little girl in the beach and sounds coming up and threw it you know
00:53:18
it's much more amenable to using body gestures then if i say something very sterile
00:53:23
and then in the lead we we actually train people to use charisma
00:53:27
what we did notices separate people looked at the nonverbal and available
00:53:31
the more retarded the rich the speech was in in the text objectively coded the
00:53:38
mall if they use gestures the correlation was point five which is pretty high
00:53:42
if i correct formation mentor in the criterion and and and and in the two variables
00:53:47
to if we corrected correlation point five divided by say the square root of
00:53:51
for daytime spun eighty which is reliability minute correlations point six something it's even right now
00:53:56
so it's really funny that the text is a good proxy for the convertible
00:54:01
crazy i mean uh
00:54:03
to trends then why do we have another question
00:54:08
or it's good you guys make a good tech team
00:54:13
um so so but let's say if you take one mosque away must yeah but i think that we i think so
00:54:22
well it is but that's not that's not it's nothing do with charisma
00:54:28
but like she's like a yeah it's not so sure that sure what's so he's not a smooth
00:54:35
talker let's separate two things so um in my study we actually measured that maybe just a
00:54:41
yeah yeah i know he's charismatic in the way i'm injectors but they just didn't show your answer your questions came in it
00:54:47
so this was the um basic research that we did way we um showed it charisma can be taught
00:54:54
um and we did a field experiment with risk on and then we did
00:54:57
a lab experiment um with my uh u. m. b. a. students
00:55:03
um so this is just the prototypical guy so what i wanna show use this yeah this is the important table
00:55:10
oops
00:55:12
use me
00:55:14
see yeah communication skills we had to separate code is code for that
00:55:19
was the speech well structured hi my name is john did i'm gonna talk about this bill
00:55:24
was the person speaking clearly hi my name is john not yeah i'm i'm on whatever
00:55:29
how many times that used non lexical utterances like on um um um um we
00:55:35
counted all the stuff we count how loudly speaking how fast it was speaking
00:55:40
so we control for communication skills and you see when we put that in and it varied between
00:55:44
the for speech they gave in the second speech it does actually nothing's completely dominated by christmas
00:55:50
and a lot than the the nice case like to use is a guy called up to allow i don't know
00:55:55
if anyone in the audience knows of a book oh yeah so book i was a present of india
00:56:00
um uh he was a physicist uh use the guy who put up into into space with is that we three nineteen eighty four
00:56:05
the speech of colour the guys that fixed camel accent ever
00:56:09
he speaks very bad grammatically incorrect status it like crazy
00:56:14
back to use is excellent storytelling and metaphors and i mean i can show you if you want is just one minute
00:56:20
sure sure you have a minute to show that's right and then we could just finish with this to show you that it's not the
00:56:25
communication skills that matt is is really the the the visuals images
00:56:30
show you have very quickly so um uh to allow um
00:56:39
mm or a or a or a line
00:56:49
yeah
00:56:53
yeah well or yeah
00:56:59
hey i or h. l. a.
00:57:05
uh_huh he or or or or or
00:57:12
yeah
00:57:14
uh oh
00:57:17
wow well i oh or
00:57:21
or or or or
00:57:26
oh yeah he are now he uh well
00:57:35
and it's in uh the p. l. o. e. yeah
00:57:48
yeah well he uh or or yeah
00:57:57
uh_huh well
00:58:01
oh of course i was who he
00:58:05
or i or or uh_huh or
00:58:14
well well well i uh i i i uh our huh uh_huh
00:58:28
uh_huh uh huh huh i i ah ah ah yeah
00:58:37
oh or or or or
00:58:44
oh yeah i mean i know how are
00:58:52
um oh uh oh yeah well i yeah
00:59:03
uh_huh ah ah yeah i uh yeah yeah
00:59:15
oh or or or uh huh oh yeah ah well he
00:59:27
oh
00:59:29
uh_huh yeah oh yeah or maybe go or mm kay
00:59:41
or
00:59:43
huh huh huh huh
00:59:50
oh
00:59:52
oh well uh huh well i i i i i. e.
01:00:05
ah so it's just you know so typically we had many cases of people were really
01:00:09
bad communicates per se but you put a metaphor you put a story animated
01:00:15
people completely all that and they it it doesn't matter in fact sometimes if a person is choose who they actually are
01:00:23
does look authentic enough and i even come across as being manipulative but you know these these kind of things i mean now you know them
01:00:29
we don't really before i made this very salient you it's not something that we actually
01:00:33
pay attention to 'em unless someone over does it in attract too much attention anyway
01:00:39
maybe which and this country discussion over lunch for those of you wanting
01:00:44
should be ordered

Share this talk: 


Recommended talks

Non-Stationary Signal Processing and its Application in Speech Recognition
Zoltán Tüske, RWTH Aachen University
Sept. 7, 2012 · 2:29 p.m.
130 views
Storage and management of sensitive speech data
Henk van den Heuvel
Sept. 4, 2019 · 11:51 a.m.