WEBVTT

00:00:05.000 --> 00:00:09.000
Ready to go? (Audience: yeah) Alright then!

00:00:09.000 --> 00:00:15.960
Hi everyone, I'm Julian. I'm from the University of the Bundeswehr in Munich. I am here presenting

00:00:15.960 --> 00:00:22.400
the work that our collaborative team has done over the past few years, especially involving

00:00:22.400 --> 00:00:29.320
Michael Koch, my direct supervisor, as well as the other three co-authors are from the

00:00:29.320 --> 00:00:34.740
HAW Hamburg, where they work on pretty similar topics. And we got together a few years

00:00:34.740 --> 00:00:43.160
ago to do some exciting collaboration on topics that I'm about to show you. We'll go a little

00:00:43.160 --> 00:00:48.000
bit into the past as well in this presentation.

00:00:48.000 --> 00:00:52.200
So, we want to look at hybrid work contexts. What role does

00:00:52.200 --> 00:01:00.000
automation play in experiments in data analysis? Historically, it's been mostly the case that

00:01:00.000 --> 00:01:06.200
researchers into work processes have relied heavily on ethnographic methods. So observations,

00:01:06.200 --> 00:01:13.260
diary studies, interviews, etc., etc., mostly qualitative methods. Our question was, can

00:01:13.260 --> 00:01:17.380
we use body tracking sensors or other automated means

00:01:17.380 --> 00:01:22.480
to gather long-term insight, and meaningful long-term insight at that,

00:01:22.480 --> 00:01:26.180
from professional contexts without requiring continuous researcher presence?

00:01:26.180 --> 00:01:30.180
That's a big motivation. We've heard some presentations today on

00:01:30.180 --> 00:01:34.180
long-term studies, and of course it becomes a lot easier

00:01:34.180 --> 00:01:38.180
to gather long-term data if you don't have to be there on location as a researcher all the time.

00:01:38.180 --> 00:01:42.180
So that's a question that led us to our

00:01:42.180 --> 00:01:46.180
project “Investigation of the honeypot effect

00:01:46.180 --> 00:01:50.180
on semi-public interactive ambient displays in long-term field studies.”

00:01:50.180 --> 00:01:54.580
That was the official project name, it went from 2021 to 2024.

00:01:54.580 --> 00:01:58.180
In case you're not familiar with the honeypot effect,

00:01:58.180 --> 00:02:02.180
it describes this effect where

00:02:02.180 --> 00:02:06.180
we have some kind of interactive installation, for example an interactive display

00:02:06.180 --> 00:02:10.180
or some other public device, and there's someone already standing there interacting

00:02:10.180 --> 00:02:14.180
with it, and thereby signaling interest

00:02:14.180 --> 00:02:18.180
as well as social acceptability for interacting with the device. It makes it a bit

00:02:18.180 --> 00:02:22.180
easier for others to join in, compared to when no one is interacting

00:02:22.180 --> 00:02:26.180
with the device. And there are these almost comedic situations where a crowd of people

00:02:26.180 --> 00:02:30.180
is standing in a big circle around the display, and no one dares to be the first one

00:02:30.180 --> 00:02:34.180
to interact. So we wanted to look at the honeypot effect,

00:02:34.180 --> 00:02:38.180
basically find out how does the honeypot effect arise

00:02:38.180 --> 00:02:42.180
with public screens, interactive public screens, in practice.

00:02:42.180 --> 00:02:46.180
The initial project was not focused on hybrid work contexts.

00:02:46.180 --> 00:02:48.180
That came a bit later.

00:02:48.180 --> 00:02:52.180
As you can see, 2024 is in the past,

00:02:52.180 --> 00:02:56.180
so we were supposed to deliver answers for this project.

00:02:56.180 --> 00:03:00.180
And the main question, can we use special sensors to do these

00:03:00.180 --> 00:03:04.180
long-term analyses to gather meaningful insight.

00:03:04.180 --> 00:03:08.180
The answer is “kind of”. What we found out when we

00:03:08.180 --> 00:03:12.180
started on this journey was that we spent way more

00:03:12.180 --> 00:03:16.180
time than we expected developing methods, developing tools

00:03:16.180 --> 00:03:20.180
during the time that we thought we would be generating answers.

00:03:20.180 --> 00:03:24.180
But we realized that we were one of very few teams

00:03:24.180 --> 00:03:28.180
worldwide doing this kind of work with body tracking sensors, and there basically were

00:03:28.180 --> 00:03:32.180
almost no tools, almost no methods to speak of, to look into

00:03:32.180 --> 00:03:36.180
the data. So we started basically from scratch and found

00:03:36.180 --> 00:03:40.180
ourselves making this face most of the time.

00:03:40.180 --> 00:03:44.180
To catch you up where you are,

00:03:44.180 --> 00:03:48.180
I should probably talk about what body tracking sensors are and what they do,

00:03:48.180 --> 00:03:52.180
because I assume not everyone is familiar with this topic.

00:03:52.180 --> 00:03:56.180
This will be boring for Thomas, because he has heard this three or four times... (Audience laughs)

00:03:56.180 --> 00:04:00.180
but for one, it's important to me to

00:04:00.180 --> 00:04:04.180
get everyone else caught up, to make sure you understand what we're talking about here.

00:04:04.180 --> 00:04:10.300
And also, he's the one who gave me 20 minutes for the presentation, so it's kind of his own fault. (Audience laughs)

00:04:10.300 --> 00:04:12.880
What do body tracking sensors do?

00:04:12.880 --> 00:04:17.180
You can imagine something like this. We have a photo here of a

00:04:17.180 --> 00:04:22.180
public interactive screen installation from one of our real deployments, in this case

00:04:22.180 --> 00:04:26.180
one from the Hamburg team, where an interactive screen showing

00:04:26.180 --> 00:04:30.180
some data relating to the teamwork processes

00:04:30.180 --> 00:04:34.180
in a software development company,

00:04:34.180 --> 00:04:40.180
I don't remember off the top of my head how many employees they have, I think it's in the triple digits,

00:04:40.180 --> 00:04:44.180
the screen sort of shows what's going on.

00:04:44.180 --> 00:04:46.180
One of our research focuses in

00:04:46.180 --> 00:04:48.180
Munich, we've been working on these screens for a

00:04:48.180 --> 00:04:50.180
long time, is

00:04:50.180 --> 00:04:52.180
fostering awareness

00:04:52.180 --> 00:04:54.180
and serendipity. So basically

00:04:54.180 --> 00:04:56.180
screens that

00:04:56.180 --> 00:04:58.180
passively show interesting things

00:04:58.180 --> 00:05:00.180
that are relevant to you as you walk past,

00:05:00.180 --> 00:05:02.180
that you can sort of pick up in passing

00:05:02.180 --> 00:05:06.180
without having to ask or to search for it, especially.

00:05:06.180 --> 00:05:08.180
This is something that

00:05:08.180 --> 00:05:10.180
matched with the interests of the Hamburg team,

00:05:10.180 --> 00:05:12.180
because they're trying to do something else.

00:05:12.180 --> 00:05:18.180
And what you see here on top of the screens is

00:05:18.180 --> 00:05:20.180
these Microsoft Kinect devices,

00:05:20.180 --> 00:05:22.180
that was the hardware in the

00:05:22.180 --> 00:05:24.180
Hamburg team's case,

00:05:24.180 --> 00:05:26.180
that are basically capable of

00:05:26.180 --> 00:05:28.180
doing a video recording, they're essentially

00:05:28.180 --> 00:05:30.180
optical cameras, and then on device

00:05:30.180 --> 00:05:32.180
do a transformation

00:05:32.180 --> 00:05:34.180
of the video signal into these

00:05:34.180 --> 00:05:38.180
sort of stylized, anonymized stick figures.

00:05:38.180 --> 00:05:40.180
They detect in the image

00:05:40.180 --> 00:05:42.180
where are humans,

00:05:42.180 --> 00:05:44.180
what are their current orientations, movements,

00:05:44.180 --> 00:05:46.180
and they get these

00:05:46.180 --> 00:05:48.180
movement data

00:05:48.180 --> 00:05:50.180
that you can then

00:05:50.180 --> 00:05:52.180
do other stuff with. For example,

00:05:52.180 --> 00:05:54.180
do live reactions,

00:05:54.180 --> 00:05:56.180
Kinect was originally a device used for gaming

00:05:56.180 --> 00:05:58.180
mostly, gesture detection in games.

00:05:58.180 --> 00:06:00.180
Or you can do what we did

00:06:00.180 --> 00:06:02.180
and record the data to look at it

00:06:02.180 --> 00:06:04.180
later, and analyze

00:06:04.180 --> 00:06:06.180
what sort of movements people do,

00:06:06.180 --> 00:06:08.180
when people enter or leave the scene,

00:06:08.180 --> 00:06:10.180
or how they interact in front of the screen.

00:06:10.180 --> 00:06:12.180
That was sort of our idea, if we could put

00:06:12.180 --> 00:06:14.180
these sensors on our screens,

00:06:14.180 --> 00:06:16.180
to get an idea of how people interact

00:06:16.180 --> 00:06:18.180
with them, how, for example, the

00:06:18.180 --> 00:06:20.180
honeypot effect arises, or how other things

00:06:20.180 --> 00:06:22.180
happen in these

00:06:22.180 --> 00:06:24.180
sorts of constellations.

00:06:24.180 --> 00:06:26.180
Because we're

00:06:26.180 --> 00:06:28.180
modern and digital

00:06:28.180 --> 00:06:30.180
and technology-based, I'm going to show you this

00:06:30.180 --> 00:06:32.180
live, here, on the

00:06:32.180 --> 00:06:34.180
PoseViz website.

00:06:34.180 --> 00:06:36.180
This is one of those

00:06:36.180 --> 00:06:38.180
3D recordings, an example

00:06:38.180 --> 00:06:40.180
that I recorded myself, with one of

00:06:40.180 --> 00:06:42.180
our sensors, so you can see me

00:06:42.180 --> 00:06:44.180
jump around, dance around,

00:06:44.180 --> 00:06:46.180
a little bit. And basically,

00:06:46.180 --> 00:06:48.180
this software for visualization

00:06:48.180 --> 00:06:50.180
is something that we had to

00:06:50.180 --> 00:06:52.180
implement from scratch. So this takes

00:06:52.180 --> 00:06:54.180
the recorded body tracking data,

00:06:54.180 --> 00:06:56.180
and you can play it back,

00:06:56.180 --> 00:06:58.180
see the movements that the person

00:06:58.180 --> 00:07:00.180
did, you can

00:07:00.180 --> 00:07:02.180
play it back from any perspective, you can

00:07:02.180 --> 00:07:04.180
zoom, you can

00:07:04.180 --> 00:07:06.180
have a variety of

00:07:06.180 --> 00:07:08.180
toggles for rendering that you can

00:07:08.180 --> 00:07:10.180
turn on and off.

00:07:10.180 --> 00:07:12.180
And this is where we had a technical outage

00:07:12.180 --> 00:07:14.180
at the conference, so it falls on

00:07:14.180 --> 00:07:16.180
post-conference Julian

00:07:16.180 --> 00:07:18.180
to show you some of the

00:07:18.180 --> 00:07:20.180
visualization toggles in

00:07:20.180 --> 00:07:22.180
the software that we have

00:07:22.180 --> 00:07:24.180
written. For example, the gaze estimation

00:07:24.180 --> 00:07:26.180
is one that I find

00:07:26.180 --> 00:07:28.180
particularly interesting.

00:07:28.180 --> 00:07:30.180
What you can see here is, based on

00:07:30.180 --> 00:07:32.180
the orientation of the head and the

00:07:32.180 --> 00:07:34.180
body, we perform a rough

00:07:34.180 --> 00:07:36.180
estimation of the

00:07:36.180 --> 00:07:38.180
person's eye direction,

00:07:38.180 --> 00:07:40.180
based purely on the body data.

00:07:40.180 --> 00:07:42.180
The second one that's interesting is

00:07:42.180 --> 00:07:44.180
the walking trajectories.

00:07:44.180 --> 00:07:46.180
You can see, especially if I scroll back

00:07:46.180 --> 00:07:48.180
to somewhere near the beginning,

00:07:48.180 --> 00:07:50.180
on the floor there is a two-dimensional

00:07:50.180 --> 00:07:52.180
projection of

00:07:52.180 --> 00:07:54.180
this recording's walking trajectory,

00:07:54.180 --> 00:07:56.180
where the position

00:07:56.180 --> 00:07:58.180
of the person is projected down

00:07:58.180 --> 00:08:00.180
and recorded, especially

00:08:00.180 --> 00:08:02.180
if we look at it from the top-down camera.

00:08:02.180 --> 00:08:04.180
We can

00:08:04.180 --> 00:08:06.180
see how a

00:08:06.180 --> 00:08:08.180
walking path

00:08:08.180 --> 00:08:10.180
for this specific recording

00:08:10.180 --> 00:08:12.180
emerges, can be extracted

00:08:12.180 --> 00:08:14.180
and can be quantitatively

00:08:14.180 --> 00:08:16.180
analyzed as

00:08:16.180 --> 00:08:18.180
one recording of many.

00:08:18.180 --> 00:08:20.180
OK, returning

00:08:20.180 --> 00:08:22.180
to my slides, you can see here

00:08:22.180 --> 00:08:24.180
once again the walking path

00:08:24.180 --> 00:08:26.180
recording

00:08:26.180 --> 00:08:28.180
that I just showed off,

00:08:28.180 --> 00:08:30.180
a 3D version as well as a

00:08:30.180 --> 00:08:32.180
2D projected down version

00:08:32.180 --> 00:08:34.180
that is suitable for

00:08:34.180 --> 00:08:36.180
data extraction and large-scale

00:08:36.180 --> 00:08:38.180
quantitative analysis.

00:08:38.180 --> 00:08:40.180
Indeed, this

00:08:40.180 --> 00:08:42.180
large-scale analysis of walking paths

00:08:42.180 --> 00:08:44.180
is something that we did as

00:08:44.180 --> 00:08:46.180
part of our larger project and which

00:08:46.180 --> 00:08:48.180
led to a separate publication.

00:08:48.180 --> 00:08:50.180
You can see in

00:08:50.180 --> 00:08:52.180
this visualization a

00:08:52.180 --> 00:08:54.180
large collection of walking paths

00:08:54.180 --> 00:08:56.180
in front of one

00:08:56.180 --> 00:08:58.180
real-world interactive screen

00:08:58.180 --> 00:09:00.180
deployment.

00:09:00.180 --> 00:09:02.180
In this view, the screen

00:09:02.180 --> 00:09:04.180
is at the bottom of

00:09:04.180 --> 00:09:06.180
the map, of the projection,

00:09:06.180 --> 00:09:08.180
and you can see this sort of cone

00:09:08.180 --> 00:09:10.180
shape showing the field of view

00:09:10.180 --> 00:09:12.180
of the sensor.

00:09:12.180 --> 00:09:14.180
What's obvious in this

00:09:14.180 --> 00:09:16.180
visualization is that

00:09:16.180 --> 00:09:18.180
there is a sort of

00:09:18.180 --> 00:09:20.180
cluster in the top

00:09:20.180 --> 00:09:22.180
left of the recording area,

00:09:22.180 --> 00:09:24.180
in this case where people just

00:09:24.180 --> 00:09:26.180
walked by the screen

00:09:26.180 --> 00:09:28.180
at some distance without ever

00:09:28.180 --> 00:09:30.180
interacting with it at all.

00:09:30.180 --> 00:09:32.180
What we did in this context was

00:09:32.180 --> 00:09:34.180
to combine similarity measures

00:09:34.180 --> 00:09:36.180
with clustering algorithms

00:09:36.180 --> 00:09:38.180
to investigate how

00:09:38.180 --> 00:09:40.180
we could extract these sort of patterns.

00:09:40.180 --> 00:09:42.180
For example, as

00:09:42.180 --> 00:09:44.180
we reduce the cluster

00:09:44.180 --> 00:09:46.180
size and increase the cluster

00:09:46.180 --> 00:09:48.180
count, you begin to

00:09:48.180 --> 00:09:50.180
see a reduction of

00:09:50.180 --> 00:09:52.180
the overall

00:09:52.180 --> 00:09:54.180
collection of walking paths

00:09:54.180 --> 00:09:56.180
down to a concentrated

00:09:56.180 --> 00:09:58.180
collection of

00:09:58.180 --> 00:10:00.180
similar paths

00:10:00.180 --> 00:10:02.180
that allow us to

00:10:02.180 --> 00:10:04.180
take a quantitative look

00:10:04.180 --> 00:10:06.180
at what

00:10:06.180 --> 00:10:08.180
walking paths in the vicinity of the screen

00:10:08.180 --> 00:10:10.180
are actually common, how common are they,

00:10:10.180 --> 00:10:12.180
what are

00:10:12.180 --> 00:10:14.180
similarities, what are interesting differences.

00:10:16.180 --> 00:10:18.180
So all in all,

00:10:18.180 --> 00:10:20.180
what are things that we managed to

00:10:20.180 --> 00:10:22.180
successfully measure through body

00:10:22.180 --> 00:10:24.180
tracking data? The walking paths

00:10:24.180 --> 00:10:26.180
were definitely a success story.

00:10:26.180 --> 00:10:28.180
I will say that we definitely

00:10:28.180 --> 00:10:30.180
did not exhaust the topic.

00:10:30.180 --> 00:10:32.180
There is potential, more

00:10:32.180 --> 00:10:34.180
work to be done there as well.

00:10:34.180 --> 00:10:38.180
But we made some significant inroads.

00:10:38.180 --> 00:10:42.180
Another thing that we successfully managed to

00:10:42.180 --> 00:10:44.180
detect was when we

00:10:44.180 --> 00:10:46.180
noticed in recordings

00:10:46.180 --> 00:10:48.180
that there was a

00:10:48.180 --> 00:10:50.180
recurring pattern

00:10:50.180 --> 00:10:52.180
of someone stopping in

00:10:52.180 --> 00:10:54.180
front of the screen and

00:10:54.180 --> 00:10:56.180
then in a certain rhythm

00:10:56.180 --> 00:10:58.180
moving their

00:10:58.180 --> 00:11:00.180
forearm, their hand, towards their face.

00:11:00.180 --> 00:11:02.180
And for a while that puzzled us

00:11:02.180 --> 00:11:04.180
what exactly was happening there

00:11:04.180 --> 00:11:06.180
until we realized that

00:11:06.180 --> 00:11:08.180
that must have been people who

00:11:08.180 --> 00:11:10.180
were visiting the nearby cafeteria and

00:11:10.180 --> 00:11:12.180
just stopped by the screen,

00:11:12.180 --> 00:11:14.180
in front of the screen, to use

00:11:14.180 --> 00:11:16.180
some of their break time to

00:11:16.180 --> 00:11:18.180
soak up information while

00:11:18.180 --> 00:11:20.180
sipping at the mug

00:11:20.180 --> 00:11:22.180
of their drink of choice.

00:11:22.180 --> 00:11:24.180
And to talk about how we approached this

00:11:24.180 --> 00:11:26.180
effect, I will hand it back to

00:11:26.180 --> 00:11:28.180
conference Julian.

00:11:28.180 --> 00:11:30.180
Yeah, what I wanted to add about the

00:11:30.180 --> 00:11:32.180
coffee mug effect, as we dubbed it, is

00:11:32.180 --> 00:11:34.180
that we had a student

00:11:34.180 --> 00:11:36.180
who was investigating this as a sort of project

00:11:36.180 --> 00:11:38.180
and he actually

00:11:38.180 --> 00:11:40.180
used a deep learning classifier

00:11:40.180 --> 00:11:42.180
for that. So that might be relevant

00:11:42.180 --> 00:11:45.180
to your work, Frederik. (Audience member: m-hm)

00:11:45.180 --> 00:11:48.180
To basically produce machine learning tools

00:11:48.180 --> 00:11:50.180
to identify these

00:11:50.180 --> 00:11:52.180
patterns in large quantities of data

00:11:52.180 --> 00:11:54.180
and not just body tracking data

00:11:54.180 --> 00:11:56.180
but video data. I think the

00:11:56.180 --> 00:11:58.180
challenges might be somewhat similar

00:11:58.180 --> 00:12:00.180
and it's definitely

00:12:00.180 --> 00:12:02.180
interesting work that happened there.

00:12:03.180 --> 00:12:06.180
So that's one of our success stories and

00:12:06.180 --> 00:12:09.180
the Honeypot effect was, uh,

00:12:09.180 --> 00:12:12.180
we investigated kind of, we were able to

00:12:12.180 --> 00:12:14.180
sort of put again

00:12:14.180 --> 00:12:16.180
patterns against the problem. We were able to see,

00:12:16.180 --> 00:12:18.180
to detect, for example,

00:12:18.180 --> 00:12:20.180
when someone stops at the screen and then

00:12:20.180 --> 00:12:22.180
within the next 10-20 seconds

00:12:22.180 --> 00:12:24.180
someone else stops also.

00:12:24.180 --> 00:12:26.180
We were able to see how often

00:12:26.180 --> 00:12:28.180
it does happen, how often it doesn't happen,

00:12:28.180 --> 00:12:30.180
how often someone just walks past without stopping,

00:12:30.180 --> 00:12:32.180
how often someone walks past and

00:12:32.180 --> 00:12:34.180
takes a look but doesn't stop

00:12:34.180 --> 00:12:36.180
and so on and so forth. We were able to do

00:12:36.180 --> 00:12:38.180
these quantitative analyses pretty well.

00:12:38.180 --> 00:12:40.180
However, it's a

00:12:40.180 --> 00:12:42.180
pretty big however,

00:12:42.180 --> 00:12:44.180
the main problem of this methodical

00:12:44.180 --> 00:12:46.180
approach is the wall of intent.

00:12:46.180 --> 00:12:48.180
The body tracking data does not show why

00:12:48.180 --> 00:12:50.180
people do what they do. We are not in

00:12:50.180 --> 00:12:52.180
people's heads. So

00:12:52.180 --> 00:12:55.180
we only have the observation.

00:12:58.180 --> 00:13:00.180
So for our

00:13:00.180 --> 00:13:02.180
current research, and this is basically the main

00:13:02.180 --> 00:13:04.180
thing that was in the position paper

00:13:04.180 --> 00:13:06.180
that I submitted, we tried

00:13:06.180 --> 00:13:08.180
to come up with a new framework

00:13:08.180 --> 00:13:10.180
to reintegrate the

00:13:10.180 --> 00:13:12.180
qualitative methods that we sort of

00:13:12.180 --> 00:13:14.180
in the COVID-induced

00:13:14.180 --> 00:13:16.180
haze sort of let fall by the wayside.

00:13:16.180 --> 00:13:18.180
And what

00:13:18.180 --> 00:13:20.180
you can see here on the right side

00:13:20.180 --> 00:13:22.180
is the body tracking and

00:13:22.180 --> 00:13:24.180
touch interaction data that we collected.

00:13:24.180 --> 00:13:26.180
The process for that, which

00:13:26.180 --> 00:13:28.180
goes sort of iteratively

00:13:28.180 --> 00:13:30.180
through data collection

00:13:30.180 --> 00:13:32.180
and data preparation. We have to do a lot

00:13:32.180 --> 00:13:34.180
of discarding of noise when

00:13:34.180 --> 00:13:36.180
actually no one is in front of the screen, for example.

00:13:36.180 --> 00:13:38.180
Exploration

00:13:38.180 --> 00:13:40.180
of the data. We did a lot of

00:13:40.180 --> 00:13:42.180
experimental visualization, checking of individual

00:13:42.180 --> 00:13:44.180
recordings to see what's going on.

00:13:44.180 --> 00:13:46.180
Feature extraction.

00:13:46.180 --> 00:13:48.180
Whenever we decided

00:13:48.180 --> 00:13:50.180
on specific elements that were interesting,

00:13:50.180 --> 00:13:52.180
we had to come up with a way, usually

00:13:52.180 --> 00:13:54.180
a bespoke way, to find

00:13:54.180 --> 00:13:56.180
applicable recordings and

00:13:56.180 --> 00:13:58.180
get the interesting segments.

00:13:58.180 --> 00:14:00.180
And then finally the data analysis

00:14:00.180 --> 00:14:02.180
to actually extract

00:14:02.180 --> 00:14:04.180
the insight we were able to

00:14:04.180 --> 00:14:06.180
gather. So the rough process

00:14:06.180 --> 00:14:08.180
just for the body tracking

00:14:08.180 --> 00:14:10.180
and other quantitative data.

00:14:10.180 --> 00:14:13.180
And to bring the qualitative data

00:14:13.180 --> 00:14:16.180
back in, our vision is

00:14:16.180 --> 00:14:18.180
to incorporate diary studies,

00:14:18.180 --> 00:14:20.180
especially for long-term

00:14:20.180 --> 00:14:22.180
examinations of professional environments,

00:14:22.180 --> 00:14:24.180
we think they are a very valuable tool,

00:14:24.180 --> 00:14:26.180
and also one where the

00:14:26.180 --> 00:14:28.180
researcher doesn't have to be continuously present.

00:14:28.180 --> 00:14:30.180
And so they pair fairly well

00:14:30.180 --> 00:14:32.180
in this long-term approach with

00:14:32.180 --> 00:14:34.180
our sensor-based

00:14:34.180 --> 00:14:36.180
data gathering methods.

00:14:36.180 --> 00:14:38.180
But also other context information.

00:14:38.180 --> 00:14:40.180
Stuff that was already on the

00:14:40.180 --> 00:14:42.180
display. We have logs of that.

00:14:42.180 --> 00:14:44.180
We have setting characteristics. What's going on

00:14:44.180 --> 00:14:46.180
on location?

00:14:46.180 --> 00:14:48.180
What's the weather like? Is there a vacation day

00:14:48.180 --> 00:14:50.180
currently? All these things sort of

00:14:50.180 --> 00:14:52.180
go into the behavior

00:14:52.180 --> 00:14:54.180
of people in front of the screen.

00:14:54.180 --> 00:14:56.180
The audience

00:14:56.180 --> 00:14:58.180
as well. And we have other qualitative

00:14:58.180 --> 00:15:00.180
data sources. Field notes from observations

00:15:00.180 --> 00:15:02.180
which would hopefully happen

00:15:02.180 --> 00:15:04.180
more again. As well as

00:15:04.180 --> 00:15:06.180
interviews and focus groups to

00:15:06.180 --> 00:15:08.180
get talking with

00:15:08.180 --> 00:15:10.180
people more again. We did some

00:15:10.180 --> 00:15:12.180
summative talks, discussions

00:15:12.180 --> 00:15:14.180
with employees of the software company.

00:15:14.180 --> 00:15:16.180
But we noticed pretty quickly

00:15:16.180 --> 00:15:18.180
that we need to be talking more.

00:15:18.180 --> 00:15:20.180
To actually learn more

00:15:20.180 --> 00:15:22.180
about things like the honeypot effect or other

00:15:22.180 --> 00:15:24.180
aspects of interaction

00:15:24.180 --> 00:15:26.180
with our tools.

00:15:26.180 --> 00:15:28.180
How people use these kinds of devices

00:15:28.180 --> 00:15:30.180
to for example interact with one another,

00:15:30.180 --> 00:15:32.180
with colleagues,

00:15:32.180 --> 00:15:34.180
and this company has an offshoot in Portugal

00:15:34.180 --> 00:15:36.180
that they collaborate with.

00:15:36.180 --> 00:15:38.180
And the way

00:15:38.180 --> 00:15:40.180
they sort of use these different tools

00:15:40.180 --> 00:15:42.180
to facilitate

00:15:42.180 --> 00:15:44.180
these interactions is something that we can

00:15:44.180 --> 00:15:46.180
only barely get at

00:15:46.180 --> 00:15:48.180
purely based on body tracking data.

00:15:48.180 --> 00:15:50.180
And crucially, by themselves

00:15:50.180 --> 00:15:52.180
neither of these halves are completely

00:15:52.180 --> 00:15:54.180
new. But we want to establish

00:15:54.180 --> 00:15:56.180
a synchronization mechanism

00:15:56.180 --> 00:15:58.180
in between them to

00:15:58.180 --> 00:16:00.180
get at the moments

00:16:00.180 --> 00:16:02.180
where these two ways

00:16:02.180 --> 00:16:04.180
of data gathering and interpretation

00:16:04.180 --> 00:16:06.180
can inform each other.

00:16:06.180 --> 00:16:08.180
For example we

00:16:08.180 --> 00:16:11.180
pretty often had the case that we

00:16:11.180 --> 00:16:14.180
had to travel either from an

00:16:14.180 --> 00:16:16.180
observation or some knowledge

00:16:16.180 --> 00:16:18.180
of some external event that we had

00:16:18.180 --> 00:16:20.180
to a specific recording.

00:16:20.180 --> 00:16:22.180
And we were able to look at that

00:16:22.180 --> 00:16:24.180
in our visualization software and see

00:16:24.180 --> 00:16:26.180
if anything interesting is happening there.

00:16:26.180 --> 00:16:28.180
Or the other way around

00:16:28.180 --> 00:16:30.180
we noticed a specific pattern,

00:16:30.180 --> 00:16:32.180
a specific anomaly in a specific recording,

00:16:32.180 --> 00:16:34.180
were able to do

00:16:34.180 --> 00:16:36.180
a check via timestamp

00:16:36.180 --> 00:16:38.180
or via other circumstances,

00:16:38.180 --> 00:16:40.180
is this something interesting? For example,

00:16:40.180 --> 00:16:42.180
I don't know, every

00:16:42.180 --> 00:16:44.180
Thursday at 11 there's suddenly a big

00:16:44.180 --> 00:16:46.180
group of people in front of the screen,

00:16:46.180 --> 00:16:48.180
but nothing in the official schedule

00:16:48.180 --> 00:16:50.180
says anything about Thursday at 11.

00:16:50.180 --> 00:16:52.180
It might be a motivation to ask

00:16:52.180 --> 00:16:54.180
the people what is going on at that time

00:16:54.180 --> 00:16:56.180
or is something interesting happening.

00:16:56.180 --> 00:16:58.180
So that's the sort of thing that we want to do.

00:16:58.180 --> 00:17:00.180
This is a future plan

00:17:00.180 --> 00:17:02.180
more or less,

00:17:02.180 --> 00:17:04.180
which I'm going to get at a little bit more

00:17:04.180 --> 00:17:06.180
here in my conclusions.

00:17:06.180 --> 00:17:08.180
The body tracking data on its own

00:17:08.180 --> 00:17:10.180
can be a rich source of descriptive

00:17:10.180 --> 00:17:12.180
observational insight.

00:17:12.180 --> 00:17:14.180
It doesn't take you very much

00:17:14.180 --> 00:17:16.180
further than that however.

00:17:16.180 --> 00:17:18.180
Observation of movement leaves intentions completely

00:17:18.180 --> 00:17:20.180
in the dark, and for what we've

00:17:20.180 --> 00:17:22.180
been able to do I think

00:17:22.180 --> 00:17:24.180
there's some interesting room

00:17:24.180 --> 00:17:26.180
for exploration

00:17:26.180 --> 00:17:28.180
of this quantitative way of looking at things.

00:17:28.180 --> 00:17:30.180
But we are currently more

00:17:30.180 --> 00:17:32.180
excited about the integration

00:17:32.180 --> 00:17:35.780
with qualitative methods for long term studies.

00:17:38.180 --> 00:17:40.180
Yeah, just said that. Future

00:17:40.180 --> 00:17:42.180
progress however will depend on

00:17:42.180 --> 00:17:44.180
pragmatic factors. What do I mean by that?

00:17:44.180 --> 00:17:46.180
I have a small confession to make to you.

00:17:46.180 --> 00:17:48.180
We kind of tricked you.

00:17:49.180 --> 00:17:52.180
When we were approached and asked

00:17:52.180 --> 00:17:54.180
if we wanted to submit

00:17:54.180 --> 00:17:56.180
something to this workshop,

00:17:56.180 --> 00:17:58.180
we didn't really have any original ideas,

00:17:58.180 --> 00:18:00.180
so we basically

00:18:00.180 --> 00:18:02.180
took a bunch of text from our

00:18:02.180 --> 00:18:04.180
rejected grant proposal

00:18:04.180 --> 00:18:06.180
and submitted it as a position paper. (Audience laughs)

00:18:06.180 --> 00:18:08.180
All this that I just said is something

00:18:08.180 --> 00:18:10.180
that we wanted

00:18:10.180 --> 00:18:12.180
to work on.

00:18:12.180 --> 00:18:14.180
Sadly the grant was rejected

00:18:14.180 --> 00:18:16.180
so this is currently up in the air

00:18:16.180 --> 00:18:18.180
if it's going to happen.

00:18:18.180 --> 00:18:20.180
We have said we are probably going to try again

00:18:20.180 --> 00:18:22.180
some time this year as time permits.

00:18:22.180 --> 00:18:24.180
If the DFG says yes

00:18:24.180 --> 00:18:26.180
then maybe this will happen

00:18:26.180 --> 00:18:28.180
over the next few years.

00:18:28.180 --> 00:18:31.180
If not, then I don't know. We'll see.

00:18:32.680 --> 00:18:36.680
Here are my email and social links if you want to get in touch.

00:18:36.680 --> 00:18:40.180
And this is a link to my website where you can find

00:18:40.180 --> 00:18:42.180
my position paper as a webpage,

00:18:42.180 --> 00:18:44.180
that's a little easier to read than the

00:18:44.180 --> 00:18:46.180
PDF version.

00:18:46.180 --> 00:18:48.180
You can also find these slides there.

00:18:48.180 --> 00:18:52.180
We are currently basically on that site.

00:18:52.180 --> 00:18:54.180
Here. It looks like this.

00:18:54.180 --> 00:18:56.180
And you will also hopefully,

00:18:56.180 --> 00:18:58.180
if the recording is salvageable,

00:18:58.180 --> 00:19:02.180
find an audio recording to connect to the slides

00:19:02.180 --> 00:19:04.180
there some time next week

00:19:04.180 --> 00:19:06.180
when I get around to it.

00:19:06.180 --> 00:19:08.180
Thank you!