WEBVTT

00:00:00.000 --> 00:00:06.020 align:middle line:90%


00:00:06.020 --> 00:00:11.410 align:middle line:84%
So to assess the
significance in my data,

00:00:11.410 --> 00:00:16.610 align:middle line:84%
it's a very important point
because I am using data

00:00:16.610 --> 00:00:20.380 align:middle line:84%
from major accidents and
they are fairly here,

00:00:20.380 --> 00:00:22.430 align:middle line:90%
where situations.

00:00:22.430 --> 00:00:26.760 align:middle line:84%
So some points
for some problems,

00:00:26.760 --> 00:00:29.880 align:middle line:84%
some root causes in
their accident reports,

00:00:29.880 --> 00:00:33.090 align:middle line:84%
we have lots of
important descriptions

00:00:33.090 --> 00:00:34.260 align:middle line:90%
from the regulators.

00:00:34.260 --> 00:00:37.740 align:middle line:84%
But for other points,
other problems and causes,

00:00:37.740 --> 00:00:39.290 align:middle line:90%
we have very few.

00:00:39.290 --> 00:00:42.060 align:middle line:84%
So it's very difficult to
assess the significance

00:00:42.060 --> 00:00:43.860 align:middle line:90%
of this few data.

00:00:43.860 --> 00:00:46.660 align:middle line:90%
We call it sparse data.

00:00:46.660 --> 00:00:50.790 align:middle line:84%
And our institute are all
explained with sparse data,

00:00:50.790 --> 00:00:52.720 align:middle line:90%
not only with big data.

00:00:52.720 --> 00:00:57.750 align:middle line:84%
So again, I am using
some methodologies

00:00:57.750 --> 00:01:00.900 align:middle line:90%
to assess this problem.

00:01:00.900 --> 00:01:08.550 align:middle line:84%
And about the findings, I will
never drop any kind of point.

00:01:08.550 --> 00:01:11.010 align:middle line:90%
We try to use all of--

00:01:11.010 --> 00:01:12.450 align:middle line:90%
we have to use all of them.

00:01:12.450 --> 00:01:18.190 align:middle line:90%


00:01:18.190 --> 00:01:20.110 align:middle line:84%
I have to consider
different ways

00:01:20.110 --> 00:01:24.840 align:middle line:84%
to interpret my findings because
at the very beginning when

00:01:24.840 --> 00:01:28.750 align:middle line:84%
I started to do verification
and validation to other people's

00:01:28.750 --> 00:01:33.310 align:middle line:84%
results, my results,
my findings were

00:01:33.310 --> 00:01:35.640 align:middle line:90%
behaving a little bit weird.

00:01:35.640 --> 00:01:38.050 align:middle line:84%
And I started to
change the model

00:01:38.050 --> 00:01:44.430 align:middle line:84%
and trying to see how it goes,
how it impacted my results.

00:01:44.430 --> 00:01:48.250 align:middle line:84%
And I could see that I started
to speak of this problem,

00:01:48.250 --> 00:01:52.510 align:middle line:84%
to open this problem in
conferences in my first two

00:01:52.510 --> 00:01:53.170 align:middle line:90%
conferences.

00:01:53.170 --> 00:01:57.700 align:middle line:84%
And I could see that everyone in
the community, in my community,

00:01:57.700 --> 00:02:00.190 align:middle line:84%
that is human
reliability analysts,

00:02:00.190 --> 00:02:02.860 align:middle line:84%
they were also having
the same problems.

00:02:02.860 --> 00:02:04.420 align:middle line:84%
But they were not
exactly discussing

00:02:04.420 --> 00:02:07.360 align:middle line:84%
that on the papers
because everyone thought

00:02:07.360 --> 00:02:09.610 align:middle line:84%
that was such a
big problem that it

00:02:09.610 --> 00:02:12.610 align:middle line:84%
was a problem of their research,
and not a problem of the data

00:02:12.610 --> 00:02:13.610 align:middle line:90%
itself.

00:02:13.610 --> 00:02:18.280 align:middle line:84%
So when we are
tracking human errors,

00:02:18.280 --> 00:02:22.780 align:middle line:84%
we started to see that we cannot
see as it was an engineering

00:02:22.780 --> 00:02:23.830 align:middle line:90%
system.

00:02:23.830 --> 00:02:28.090 align:middle line:84%
I cannot compare my data and
my findings with my colleagues'

00:02:28.090 --> 00:02:34.300 align:middle line:84%
data and findings because
they are analysing various

00:02:34.300 --> 00:02:38.680 align:middle line:84%
engineering
systems-based findings,

00:02:38.680 --> 00:02:42.290 align:middle line:84%
and I'm looking for human
errors, probabilities.

00:02:42.290 --> 00:02:46.870 align:middle line:84%
So we started to see that
data interpretation had

00:02:46.870 --> 00:02:48.160 align:middle line:90%
to be somehow different.

00:02:48.160 --> 00:02:51.040 align:middle line:84%
We had to involve
some psychologists

00:02:51.040 --> 00:02:54.574 align:middle line:84%
and some philosophers
together with engineering.

00:02:54.574 --> 00:02:58.840 align:middle line:90%


00:02:58.840 --> 00:03:03.190 align:middle line:84%
Usually in my research topic,
that is human reliability,

00:03:03.190 --> 00:03:06.370 align:middle line:84%
when there are some
holes in the grades,

00:03:06.370 --> 00:03:10.150 align:middle line:84%
when there is some holes with
data, when we have sparse data,

00:03:10.150 --> 00:03:13.720 align:middle line:90%
people use expert judgment.

00:03:13.720 --> 00:03:17.020 align:middle line:84%
So they ask some experts
to give some judgments

00:03:17.020 --> 00:03:19.390 align:middle line:90%
of what that point should be.

00:03:19.390 --> 00:03:23.980 align:middle line:84%
And that's not
considered very credible

00:03:23.980 --> 00:03:26.120 align:middle line:90%
in my area of research.

00:03:26.120 --> 00:03:30.160 align:middle line:84%
So we started to use what we
call imprecise probability.

00:03:30.160 --> 00:03:34.030 align:middle line:84%
So we could quite
relax and be honest

00:03:34.030 --> 00:03:37.840 align:middle line:84%
that at that point,
in that interval, we

00:03:37.840 --> 00:03:40.180 align:middle line:84%
don't have any
points, and that's

00:03:40.180 --> 00:03:42.970 align:middle line:84%
going to be explicit
in the result.

00:03:42.970 --> 00:03:46.990 align:middle line:84%
So the result's not anymore
a point, it's an interval.

00:03:46.990 --> 00:03:50.650 align:middle line:84%
So in this sense, we believe
that we are being transparent

00:03:50.650 --> 00:03:54.910 align:middle line:84%
with society, that when
they see that interval

00:03:54.910 --> 00:03:58.990 align:middle line:84%
they know that we
still need more data,

00:03:58.990 --> 00:04:01.440 align:middle line:90%
we don't have all the data.