WEBVTT

00:00:00.000 --> 00:00:00.894 align:middle line:90%


00:00:00.894 --> 00:00:01.788 align:middle line:90%
Hi, everyone.

00:00:01.788 --> 00:00:04.150 align:middle line:84%
Thank you so much,
Kelly, for the intro,

00:00:04.150 --> 00:00:07.792 align:middle line:84%
and thank you everyone for
attending today's panel.

00:00:07.792 --> 00:00:09.200 align:middle line:90%
Hi.

00:00:09.200 --> 00:00:11.570 align:middle line:84%
Today, my talk will be
on brains and thoughts

00:00:11.570 --> 00:00:13.390 align:middle line:84%
and making the
two work together.

00:00:13.390 --> 00:00:16.005 align:middle line:84%
So before I start, my
good friend, L-E, here,

00:00:16.005 --> 00:00:18.260 align:middle line:90%
would like to say a few words.

00:00:18.260 --> 00:00:18.926 align:middle line:90%
[VIDEO PLAYBACK]

00:00:18.926 --> 00:00:41.976 align:middle line:90%


00:00:41.976 --> 00:00:42.972 align:middle line:90%
[END VIDEO PLAYBACK]

00:00:42.972 --> 00:00:45.700 align:middle line:84%
[LAUGHS] So some
brief background.

00:00:45.700 --> 00:00:48.550 align:middle line:84%
I entered Mount Holyoke College
as a neuroscience and behavior

00:00:48.550 --> 00:00:51.670 align:middle line:84%
major and later realized towards
the end of my sophomore year

00:00:51.670 --> 00:00:55.930 align:middle line:84%
that I also wanted to pursue
a computer science major.

00:00:55.930 --> 00:00:58.130 align:middle line:84%
So since I was a
sophomore at that point,

00:00:58.130 --> 00:01:01.450 align:middle line:84%
I really had to scramble
to find an internship that

00:01:01.450 --> 00:01:05.260 align:middle line:84%
could incorporate some
knowledge from both fields.

00:01:05.260 --> 00:01:07.570 align:middle line:84%
And more importantly,
I really wanted

00:01:07.570 --> 00:01:10.180 align:middle line:84%
to gain new experience
in programming

00:01:10.180 --> 00:01:14.050 align:middle line:84%
and possibly learn new languages
to balance my skill sets.

00:01:14.050 --> 00:01:17.740 align:middle line:84%
So I worked previous summers
at the Yale Child Study Center,

00:01:17.740 --> 00:01:19.750 align:middle line:84%
and while researching
many of their labs,

00:01:19.750 --> 00:01:24.180 align:middle line:84%
I happened to stumble upon the
Technology and Innovation Lab.

00:01:24.180 --> 00:01:27.250 align:middle line:84%
And something that's really
common for STEM majors

00:01:27.250 --> 00:01:30.720 align:middle line:84%
is that if you're not a
client through a RISE program

00:01:30.720 --> 00:01:33.910 align:middle line:84%
or an undergraduate summer
program for research,

00:01:33.910 --> 00:01:37.650 align:middle line:84%
most of the time, you were
simply reaching out to a lab.

00:01:37.650 --> 00:01:41.210 align:middle line:84%
Either a lab
director, a post-doc,

00:01:41.210 --> 00:01:44.530 align:middle line:90%
or a principal investigator.

00:01:44.530 --> 00:01:48.940 align:middle line:84%
And usually you send an email,
send a resume, a cover letter,

00:01:48.940 --> 00:01:51.850 align:middle line:84%
and you link them to
your LinkedIn account.

00:01:51.850 --> 00:01:55.480 align:middle line:84%
And so before you
start applying to labs,

00:01:55.480 --> 00:01:57.960 align:middle line:84%
I think it's really
important to know clearly

00:01:57.960 --> 00:02:00.700 align:middle line:84%
what skill sets you wish
to stretch before you apply

00:02:00.700 --> 00:02:04.879 align:middle line:84%
to these labs, and
you let them know

00:02:04.879 --> 00:02:07.970 align:middle line:84%
what you're passionate about
and what you're eager to learn.

00:02:07.970 --> 00:02:10.860 align:middle line:84%
And I think that really makes
up for any lack of experience

00:02:10.860 --> 00:02:12.080 align:middle line:90%
in the field.

00:02:12.080 --> 00:02:16.570 align:middle line:84%
I believe that while I did lack
a lot of technical experience,

00:02:16.570 --> 00:02:21.460 align:middle line:84%
my willingness to learn
a lot of new languages

00:02:21.460 --> 00:02:28.712 align:middle line:84%
and understand how to program
really got labs to consider me.

00:02:28.712 --> 00:02:30.322 align:middle line:84%
And I think, while
it's certainly

00:02:30.322 --> 00:02:32.530 align:middle line:84%
important to build upon
already existing connections,

00:02:32.530 --> 00:02:35.214 align:middle line:84%
it's always an
option to consider.

00:02:35.214 --> 00:02:38.980 align:middle line:84%
Yale Child Study Center
was a really good entity

00:02:38.980 --> 00:02:41.125 align:middle line:84%
with a strong community
of researchers,

00:02:41.125 --> 00:02:44.080 align:middle line:84%
and they are all more or less
knew one another in the Center.

00:02:44.080 --> 00:02:47.440 align:middle line:84%
So I was really able to build
upon the various connections

00:02:47.440 --> 00:02:52.370 align:middle line:84%
and managed to find Technology
Innovation Lab through that.

00:02:52.370 --> 00:02:56.410 align:middle line:84%
So overall, there's no official
application or interview,

00:02:56.410 --> 00:02:59.910 align:middle line:84%
except for maybe a formal
psych interview sometimes.

00:02:59.910 --> 00:03:03.370 align:middle line:84%
So it's important to
maintain flexibility

00:03:03.370 --> 00:03:07.500 align:middle line:84%
when you're applying to
a research experience

00:03:07.500 --> 00:03:13.630 align:middle line:84%
at a lab that isn't offered
through a RISE program.

00:03:13.630 --> 00:03:16.155 align:middle line:84%
And so it was also
really hard to emphasize

00:03:16.155 --> 00:03:19.330 align:middle line:90%
various aspects of your resume.

00:03:19.330 --> 00:03:21.640 align:middle line:84%
So for example,
while I didn't have

00:03:21.640 --> 00:03:23.950 align:middle line:84%
a lot of experience
in technology,

00:03:23.950 --> 00:03:28.670 align:middle line:84%
I had a lot of experience
working with children.

00:03:28.670 --> 00:03:30.220 align:middle line:84%
I had a lot of
tutoring experience,

00:03:30.220 --> 00:03:31.870 align:middle line:84%
and I also had a
lot of knowledge

00:03:31.870 --> 00:03:33.400 align:middle line:90%
in scientific writing.

00:03:33.400 --> 00:03:36.160 align:middle line:84%
And so it was all about making
certain skill sets appeal

00:03:36.160 --> 00:03:38.770 align:middle line:84%
to the particular
lab's mission statement

00:03:38.770 --> 00:03:40.910 align:middle line:90%
and their overall goals.

00:03:40.910 --> 00:03:44.050 align:middle line:84%
So that was how I
found my internship.

00:03:44.050 --> 00:03:49.630 align:middle line:84%
My projects, well,
L-E, what you saw,

00:03:49.630 --> 00:03:51.630 align:middle line:90%
was my main project this summer.

00:03:51.630 --> 00:03:54.820 align:middle line:84%
I also worked briefly with
ongoing projects working with

00:03:54.820 --> 00:03:58.360 align:middle line:84%
[? Keypon ?] and [? Sparrow. ?]
I can talk to people about that

00:03:58.360 --> 00:04:00.510 align:middle line:90%
later if they're interested.

00:04:00.510 --> 00:04:04.720 align:middle line:84%
But L-E was my main
project, and it was just

00:04:04.720 --> 00:04:07.590 align:middle line:84%
getting started when I entered
the lab, so it was really

00:04:07.590 --> 00:04:10.450 align:middle line:84%
a great experience to see
L-E change and evolve,

00:04:10.450 --> 00:04:13.780 align:middle line:84%
working out the bugs,
developing the protocol

00:04:13.780 --> 00:04:16.430 align:middle line:84%
and the experimental design
throughout the summer.

00:04:16.430 --> 00:04:20.560 align:middle line:84%
So to go over, briefly, what
I did, both of my coding

00:04:20.560 --> 00:04:23.602 align:middle line:84%
was in Visual Studio, which I
had no prior experience working

00:04:23.602 --> 00:04:25.370 align:middle line:90%
in.

00:04:25.370 --> 00:04:28.830 align:middle line:84%
And then it was also
working with an Arduino

00:04:28.830 --> 00:04:35.300 align:middle line:84%
board, which is a pretty
standard microcontroller.

00:04:35.300 --> 00:04:39.260 align:middle line:84%
We interfaced it with an
Arduino servosensor here,

00:04:39.260 --> 00:04:43.260 align:middle line:84%
and that was able to attach
through multiple servomotors

00:04:43.260 --> 00:04:46.017 align:middle line:84%
that would allow us to
control the facial movements

00:04:46.017 --> 00:04:47.330 align:middle line:90%
of the robot.

00:04:47.330 --> 00:04:51.340 align:middle line:84%
So Visual Basic was mostly to
call the servomotor functions

00:04:51.340 --> 00:04:53.305 align:middle line:84%
and coordinate the
movements of the robot's

00:04:53.305 --> 00:04:57.470 align:middle line:90%
face and other behaviors.

00:04:57.470 --> 00:04:59.980 align:middle line:84%
So this is what it overall
looked like from the back

00:04:59.980 --> 00:05:03.041 align:middle line:90%
after we put it together.

00:05:03.041 --> 00:05:05.230 align:middle line:84%
And we also created
a GUI from which

00:05:05.230 --> 00:05:08.080 align:middle line:84%
we could tele-operate
the robot or remotely

00:05:08.080 --> 00:05:11.540 align:middle line:90%
control from an operating room.

00:05:11.540 --> 00:05:17.810 align:middle line:84%
And so as you can see, there
are faces, various actions

00:05:17.810 --> 00:05:21.995 align:middle line:84%
and body positions,
conversational basics, scripts,

00:05:21.995 --> 00:05:22.900 align:middle line:90%
and so forth.

00:05:22.900 --> 00:05:25.585 align:middle line:84%
We also wanted to personalize
it as much as we could.

00:05:25.585 --> 00:05:26.940 align:middle line:90%
So we would have a name box.

00:05:26.940 --> 00:05:29.666 align:middle line:84%
We'd enter the child's name,
so when the child came in,

00:05:29.666 --> 00:05:31.780 align:middle line:84%
the robot would be
like, hi, Jeffrey,

00:05:31.780 --> 00:05:35.070 align:middle line:84%
and the child would be like,
whoa, how do you know me?

00:05:35.070 --> 00:05:38.540 align:middle line:84%
So that was really exciting
to see the child react.

00:05:38.540 --> 00:05:41.190 align:middle line:84%
So after implementing all
these behaviors and gestures,

00:05:41.190 --> 00:05:44.310 align:middle line:84%
giving it a voice,
giving it a personality.

00:05:44.310 --> 00:05:47.182 align:middle line:84%
Getting to work
with it, we moved on

00:05:47.182 --> 00:05:48.640 align:middle line:84%
to the more behavioral
neuroscience

00:05:48.640 --> 00:05:51.660 align:middle line:84%
aspect of the project,
which was introducing them

00:05:51.660 --> 00:05:54.185 align:middle line:90%
to ASD therapies.

00:05:54.185 --> 00:05:58.120 align:middle line:84%
And the most prevalent therapy
is pivotal response treatment,

00:05:58.120 --> 00:05:59.940 align:middle line:84%
and it's considered
an effective therapy

00:05:59.940 --> 00:06:02.685 align:middle line:84%
that's more personalized
for the child.

00:06:02.685 --> 00:06:05.630 align:middle line:90%


00:06:05.630 --> 00:06:09.090 align:middle line:84%
It focuses on the child's
interests and works

00:06:09.090 --> 00:06:14.430 align:middle line:84%
with the child in a way
that's much less stressful.

00:06:14.430 --> 00:06:19.116 align:middle line:84%
A lot of ASD treatments,
known behavioral therapies,

00:06:19.116 --> 00:06:20.490 align:middle line:84%
are known to be
stressful and not

00:06:20.490 --> 00:06:23.710 align:middle line:84%
necessarily effective
for children with ASD.

00:06:23.710 --> 00:06:28.830 align:middle line:84%
And so we try to build upon
pivotal response training.

00:06:28.830 --> 00:06:32.340 align:middle line:84%
There's also a lot of ongoing
research regarding the learning

00:06:32.340 --> 00:06:34.650 align:middle line:84%
by teaching paradigm,
which was established

00:06:34.650 --> 00:06:37.200 align:middle line:84%
by a psychiatrist
named William Glasser.

00:06:37.200 --> 00:06:42.030 align:middle line:84%
And he established that 90%
of what we retain the most

00:06:42.030 --> 00:06:43.740 align:middle line:90%
is through what we teach.

00:06:43.740 --> 00:06:47.040 align:middle line:84%
And so there's a learning
period associated with that.

00:06:47.040 --> 00:06:49.800 align:middle line:84%
There's also a lot of
emerging research based

00:06:49.800 --> 00:06:52.890 align:middle line:84%
on the benefits of
socially-assistive robots used

00:06:52.890 --> 00:06:55.140 align:middle line:90%
in ASD therapy.

00:06:55.140 --> 00:06:58.830 align:middle line:84%
They're known to reduce stress
and pressure often associated

00:06:58.830 --> 00:07:00.820 align:middle line:90%
with interacting with people.

00:07:00.820 --> 00:07:03.220 align:middle line:84%
So we believe that
socially-assitive robots

00:07:03.220 --> 00:07:05.685 align:middle line:84%
would be an effective
way to help children

00:07:05.685 --> 00:07:09.645 align:middle line:84%
with ASD feel more comfortable
learning in academic settings,

00:07:09.645 --> 00:07:13.050 align:middle line:84%
if they were to engage
with a robot beforehand

00:07:13.050 --> 00:07:17.650 align:middle line:84%
or to learn how to
interact with them.

00:07:17.650 --> 00:07:19.890 align:middle line:84%
So I'm just going to
go over this really

00:07:19.890 --> 00:07:21.500 align:middle line:90%
quickly because of lack of time.

00:07:21.500 --> 00:07:23.970 align:middle line:84%
We recruited about
six to 10 participants

00:07:23.970 --> 00:07:25.145 align:middle line:90%
for our pilot study.

00:07:25.145 --> 00:07:27.330 align:middle line:84%
They were children ranged
from five to eight,

00:07:27.330 --> 00:07:33.070 align:middle line:84%
all varying intensities on
the autism spectrum disorder.

00:07:33.070 --> 00:07:37.510 align:middle line:84%
And we allotted about a
20-minute session for the child

00:07:37.510 --> 00:07:39.720 align:middle line:84%
to interact freely
with the robot,

00:07:39.720 --> 00:07:42.590 align:middle line:84%
to get a general idea of what
the child was interested in.

00:07:42.590 --> 00:07:45.340 align:middle line:84%
And from that, after
that 20-minute session,

00:07:45.340 --> 00:07:47.610 align:middle line:84%
we would develop a
simple script in which

00:07:47.610 --> 00:07:53.720 align:middle line:84%
the child could ask
questions and encouraged

00:07:53.720 --> 00:07:56.300 align:middle line:84%
the child to teach the
robot a concept that they

00:07:56.300 --> 00:07:57.870 align:middle line:90%
were passionate about.

00:07:57.870 --> 00:08:00.620 align:middle line:84%
And we formulated
the same script

00:08:00.620 --> 00:08:03.350 align:middle line:84%
that we have with the robot
with a human confederate,

00:08:03.350 --> 00:08:06.440 align:middle line:84%
and after that, we
compared and measured,

00:08:06.440 --> 00:08:09.710 align:middle line:84%
through video coding,
the following points.

00:08:09.710 --> 00:08:11.240 align:middle line:84%
We looked at
percentage and duration

00:08:11.240 --> 00:08:14.410 align:middle line:84%
of responses from the child
when the robot initiated

00:08:14.410 --> 00:08:17.785 align:middle line:84%
a communication versus when
the human confederate initiates

00:08:17.785 --> 00:08:20.860 align:middle line:84%
communication and when
the facilitator initiates

00:08:20.860 --> 00:08:22.250 align:middle line:90%
communication.

00:08:22.250 --> 00:08:25.085 align:middle line:84%
The lag time of the child's
responses to the robot.

00:08:25.085 --> 00:08:29.010 align:middle line:84%
The percentage and duration of
child-initiated communication,

00:08:29.010 --> 00:08:32.470 align:middle line:84%
number of conversational turns,
and quality of conversation.

00:08:32.470 --> 00:08:35.850 align:middle line:84%
And this was done
through video coding.

00:08:35.850 --> 00:08:39.299 align:middle line:84%
Unfortunately, video coding is
a very long and tedious process,

00:08:39.299 --> 00:08:42.360 align:middle line:84%
and I did not intern at the
Technology Innovation Lab

00:08:42.360 --> 00:08:45.940 align:middle line:84%
long enough to obtain
statistical data.

00:08:45.940 --> 00:08:48.588 align:middle line:90%
But we did receive--

00:08:48.588 --> 00:08:50.340 align:middle line:84%
well, L-E received
a lot of love.

00:08:50.340 --> 00:08:53.280 align:middle line:84%
There were a lot
of parent comments.

00:08:53.280 --> 00:08:55.880 align:middle line:84%
Parents who sat in
the observation room

00:08:55.880 --> 00:08:57.690 align:middle line:90%
said they were really excited.

00:08:57.690 --> 00:09:00.860 align:middle line:84%
Their child had never been
so engaged and passionate

00:09:00.860 --> 00:09:06.410 align:middle line:84%
and have not been
so talkative before.

00:09:06.410 --> 00:09:10.480 align:middle line:84%
With L-E, a lot of children
demanded revisits with L-E

00:09:10.480 --> 00:09:14.430 align:middle line:84%
after meeting him and
having the two sessions.

00:09:14.430 --> 00:09:16.465 align:middle line:90%
So those were all good--

00:09:16.465 --> 00:09:18.690 align:middle line:84%
I think that's the
best data of all,

00:09:18.690 --> 00:09:22.230 align:middle line:90%
seeing people react well to L-E.

00:09:22.230 --> 00:09:25.500 align:middle line:84%
And then for future
directions, we

00:09:25.500 --> 00:09:27.950 align:middle line:84%
plan on making L-E
more autonomous.

00:09:27.950 --> 00:09:30.900 align:middle line:84%
Before I left, I added a little
bit of speech recognition,

00:09:30.900 --> 00:09:32.330 align:middle line:90%
but it was very buggy.

00:09:32.330 --> 00:09:36.510 align:middle line:84%
So getting rid of the
bugs, adding a camera

00:09:36.510 --> 00:09:38.486 align:middle line:90%
for face detection, et cetera.

00:09:38.486 --> 00:09:43.280 align:middle line:84%
So moving from
tele-operating to autonomy.

00:09:43.280 --> 00:09:45.990 align:middle line:84%
And then we also
want to introduce L-E

00:09:45.990 --> 00:09:47.620 align:middle line:90%
to a lot of local schools.

00:09:47.620 --> 00:09:50.790 align:middle line:84%
They planned on piloting
them in in-school tutoring

00:09:50.790 --> 00:09:54.300 align:middle line:90%
programs at local schools.

00:09:54.300 --> 00:09:55.680 align:middle line:90%
OK.

00:09:55.680 --> 00:09:57.680 align:middle line:84%
Before I end, I just
wanted to show you

00:09:57.680 --> 00:10:00.470 align:middle line:84%
that when we first
started, L-E was not

00:10:00.470 --> 00:10:02.800 align:middle line:90%
a very attractive robot.

00:10:02.800 --> 00:10:05.030 align:middle line:90%
L-E was [INAUDIBLE].

00:10:05.030 --> 00:10:09.336 align:middle line:84%
And so there was a lot
of growing and evolving.

00:10:09.336 --> 00:10:11.520 align:middle line:90%
So while L-E grew, I also grew.

00:10:11.520 --> 00:10:12.640 align:middle line:90%
I learned lot of things.

00:10:12.640 --> 00:10:14.894 align:middle line:90%
I made a lot of mistakes.

00:10:14.894 --> 00:10:17.060 align:middle line:84%
So if you don't get a
mechanical engineering student

00:10:17.060 --> 00:10:20.504 align:middle line:84%
to help you with the 3D
printer, the robot's face

00:10:20.504 --> 00:10:21.234 align:middle line:90%
will be orange.

00:10:21.234 --> 00:10:24.450 align:middle line:90%
[LAUGHS] But we worked with it.

00:10:24.450 --> 00:10:25.465 align:middle line:90%
The kids were fine.

00:10:25.465 --> 00:10:27.400 align:middle line:90%
They liked orange.

00:10:27.400 --> 00:10:32.810 align:middle line:84%
We just told them L-E was
a little under the weather.

00:10:32.810 --> 00:10:34.460 align:middle line:90%
So we worked with that.

00:10:34.460 --> 00:10:36.950 align:middle line:84%
Other times, things
just don't work out,

00:10:36.950 --> 00:10:39.990 align:middle line:90%
and the robot looks creepy.

00:10:39.990 --> 00:10:42.025 align:middle line:84%
Like uncanny valley,
there are no eyebrows,

00:10:42.025 --> 00:10:43.640 align:middle line:90%
it just looks really creepy.

00:10:43.640 --> 00:10:47.730 align:middle line:84%
But ultimately, L-E ended up
being quite a cute, adorable,

00:10:47.730 --> 00:10:48.830 align:middle line:90%
and charming robot.

00:10:48.830 --> 00:10:51.350 align:middle line:84%
The tie, the little
bowtie, we also

00:10:51.350 --> 00:10:52.886 align:middle line:90%
got a bunch of different kinds.

00:10:52.886 --> 00:10:54.635 align:middle line:90%
We made a lot of little bowties.

00:10:54.635 --> 00:10:56.860 align:middle line:84%
So L-E's got a lot of
different personalities

00:10:56.860 --> 00:11:01.550 align:middle line:84%
to understand what
the child likes.

00:11:01.550 --> 00:11:03.990 align:middle line:84%
For this project, I'd like
to thank my mentor, the PI,

00:11:03.990 --> 00:11:07.820 align:middle line:84%
and I met a lot of awesome
and brilliant professors

00:11:07.820 --> 00:11:10.910 align:middle line:84%
and researchers that
I had the pleasure

00:11:10.910 --> 00:11:12.710 align:middle line:90%
of working with this summer.

00:11:12.710 --> 00:11:15.760 align:middle line:90%
So questions?

00:11:15.760 --> 00:11:17.910 align:middle line:90%