WEBVTT

00:00:00.000 --> 00:00:02.952 align:middle line:90%
[AUDIO LOGO]

00:00:02.952 --> 00:00:12.065 align:middle line:90%


00:00:12.065 --> 00:00:13.690 align:middle line:84%
RICHARD DE NEUFVILLE:
There's something

00:00:13.690 --> 00:00:16.149 align:middle line:90%
important about this title.

00:00:16.149 --> 00:00:21.700 align:middle line:84%
And it is that I'm talking
about preferred projects.

00:00:21.700 --> 00:00:24.490 align:middle line:84%
I'm not talking
about best projects.

00:00:24.490 --> 00:00:29.320 align:middle line:84%
And this is very
much on purpose.

00:00:29.320 --> 00:00:33.940 align:middle line:84%
Basically, when we
think about a system,

00:00:33.940 --> 00:00:36.820 align:middle line:84%
a system which has lots
of different performance

00:00:36.820 --> 00:00:40.030 align:middle line:84%
characteristics, which has
numbers of stakeholders who are

00:00:40.030 --> 00:00:41.980 align:middle line:84%
thinking about it--
there's a financial side,

00:00:41.980 --> 00:00:46.390 align:middle line:84%
there's a technological side,
there's the human relations

00:00:46.390 --> 00:00:49.090 align:middle line:84%
side, there's environmental
sides, et cetera,

00:00:49.090 --> 00:00:52.090 align:middle line:84%
different criteria
for performance--

00:00:52.090 --> 00:00:55.180 align:middle line:84%
it is basically
conceptually impossible

00:00:55.180 --> 00:00:57.070 align:middle line:90%
to talk about the best.

00:00:57.070 --> 00:01:01.870 align:middle line:84%
We may talk about what we think
is our preferred choice, what

00:01:01.870 --> 00:01:04.750 align:middle line:84%
is, in some ways, in
our minds, the best.

00:01:04.750 --> 00:01:08.330 align:middle line:84%
But in terms of an
optimization, it's difficult.

00:01:08.330 --> 00:01:14.290 align:middle line:84%
And I think that we ought
to recognize that and use

00:01:14.290 --> 00:01:18.190 align:middle line:84%
optimization techniques to sort
through dominated solutions,

00:01:18.190 --> 00:01:22.510 align:middle line:84%
a concept I'll come to
by the end of the class,

00:01:22.510 --> 00:01:25.000 align:middle line:84%
but really focus
on that we're not

00:01:25.000 --> 00:01:29.200 align:middle line:84%
going to be able to optimize
our systems overall.

00:01:29.200 --> 00:01:31.570 align:middle line:84%
We can think about
the alternatives.

00:01:31.570 --> 00:01:34.780 align:middle line:84%
And ultimately, we're going to
have what I like to think about

00:01:34.780 --> 00:01:36.670 align:middle line:90%
as a preferred project--

00:01:36.670 --> 00:01:39.670 align:middle line:84%
that is, one that has
a balance, a trade-off

00:01:39.670 --> 00:01:41.860 align:middle line:84%
between the different
elements, between

00:01:41.860 --> 00:01:44.950 align:middle line:84%
the different
stakeholders and we agree

00:01:44.950 --> 00:01:47.110 align:middle line:84%
is what we're going
to do and we think

00:01:47.110 --> 00:01:51.520 align:middle line:84%
it is the preferred
solution, although it is not,

00:01:51.520 --> 00:01:55.330 align:middle line:84%
mathematically
speaking, an optimal.

00:01:55.330 --> 00:01:57.310 align:middle line:84%
Let's go into that
just a little bit.

00:01:57.310 --> 00:02:01.370 align:middle line:90%


00:02:01.370 --> 00:02:05.710 align:middle line:84%
So the fundamental question I
can start off with is this--

00:02:05.710 --> 00:02:09.160 align:middle line:84%
is it meaningful to
look for the best?

00:02:09.160 --> 00:02:11.140 align:middle line:90%
what can we expect to do--

00:02:11.140 --> 00:02:15.820 align:middle line:84%
and then to think
about, all right,

00:02:15.820 --> 00:02:18.940 align:middle line:84%
if we're going to think about
a goodness of some sort,

00:02:18.940 --> 00:02:21.640 align:middle line:90%
we have to talk about value.

00:02:21.640 --> 00:02:23.530 align:middle line:84%
And we have to
think about what it

00:02:23.530 --> 00:02:28.823 align:middle line:84%
means to valuate things where we
have different elements of it.

00:02:28.823 --> 00:02:30.490 align:middle line:84%
We have different
characteristics of it.

00:02:30.490 --> 00:02:32.500 align:middle line:90%
We have performance technically.

00:02:32.500 --> 00:02:34.870 align:middle line:84%
We have performance in
terms of environmentally.

00:02:34.870 --> 00:02:37.120 align:middle line:84%
We have performance in
terms of financially.

00:02:37.120 --> 00:02:41.680 align:middle line:84%
And we need to have a way
to translate each of those

00:02:41.680 --> 00:02:46.630 align:middle line:84%
performance indices into
some value-functional--

00:02:46.630 --> 00:02:49.210 align:middle line:90%
I'll refer to it generally.

00:02:49.210 --> 00:02:50.930 align:middle line:84%
And a value
function, by the way,

00:02:50.930 --> 00:02:54.460 align:middle line:84%
from a technical point of view,
is not the same as utility

00:02:54.460 --> 00:02:58.360 align:middle line:84%
function, but it is
the same kind of idea

00:02:58.360 --> 00:03:00.160 align:middle line:90%
in a general language form.

00:03:00.160 --> 00:03:02.050 align:middle line:84%
But technically, there's
a great difference

00:03:02.050 --> 00:03:05.210 align:middle line:84%
between value and
utility of the sort

00:03:05.210 --> 00:03:09.020 align:middle line:84%
that utility requires a lot
of different assumptions

00:03:09.020 --> 00:03:12.920 align:middle line:90%
and is very difficult to design.

00:03:12.920 --> 00:03:16.370 align:middle line:84%
Then, having set the
scene in this way,

00:03:16.370 --> 00:03:18.200 align:middle line:84%
I want to talk about
the several methods

00:03:18.200 --> 00:03:21.650 align:middle line:84%
I'd like to suggest
to you that we

00:03:21.650 --> 00:03:26.600 align:middle line:84%
can use to evaluate and choose
between different solutions.

00:03:26.600 --> 00:03:29.760 align:middle line:90%


00:03:29.760 --> 00:03:32.810 align:middle line:84%
So let's go down to
think about, what

00:03:32.810 --> 00:03:35.360 align:middle line:84%
is it meaningful to
optimize in the sense

00:03:35.360 --> 00:03:38.750 align:middle line:84%
to have a mathematical
search for the best?

00:03:38.750 --> 00:03:41.520 align:middle line:90%
Well, what defines best?

00:03:41.520 --> 00:03:43.740 align:middle line:84%
Well, as we know,
it's the extreme--

00:03:43.740 --> 00:03:46.780 align:middle line:84%
the high or the low
of all possibilities.

00:03:46.780 --> 00:03:50.010 align:middle line:84%
So to be able to know whether
you're the highest or lowest,

00:03:50.010 --> 00:03:52.150 align:middle line:84%
you have to have a
metric of performance.

00:03:52.150 --> 00:03:54.840 align:middle line:84%
If I'm looking at
altitude, I can

00:03:54.840 --> 00:03:59.280 align:middle line:84%
calculate something is at 1,000
feet, 2,000 feet, 3,000 feet.

00:03:59.280 --> 00:04:02.250 align:middle line:84%
I can know which
one is the highest.

00:04:02.250 --> 00:04:05.060 align:middle line:84%
Well, let's say
it's Mount Everest.

00:04:05.060 --> 00:04:07.640 align:middle line:84%
Or if we can't have
a single measure,

00:04:07.640 --> 00:04:10.580 align:middle line:84%
we have some way of
saying that, all right,

00:04:10.580 --> 00:04:13.700 align:middle line:84%
we have different
levels of performance.

00:04:13.700 --> 00:04:17.149 align:middle line:84%
Take a hospital,
say, that it is I

00:04:17.149 --> 00:04:20.630 align:middle line:84%
want to be able to
reduce fatalities.

00:04:20.630 --> 00:04:27.260 align:middle line:84%
I want to provide the best
service to cancer patients,

00:04:27.260 --> 00:04:31.562 align:middle line:84%
that they have less
after-consequences.

00:04:31.562 --> 00:04:33.020 align:middle line:84%
I want to make sure
that et cetera.

00:04:33.020 --> 00:04:35.630 align:middle line:84%
There are various measures
of performance of it.

00:04:35.630 --> 00:04:39.330 align:middle line:84%
And it might have economic
efficiency put in there.

00:04:39.330 --> 00:04:42.440 align:middle line:84%
And to be able to
talk about the optimum

00:04:42.440 --> 00:04:45.170 align:middle line:84%
from a mathematical way, we
have to put these metrics

00:04:45.170 --> 00:04:49.370 align:middle line:84%
into a single scale,
like so many deaths

00:04:49.370 --> 00:04:52.020 align:middle line:84%
saved are equivalent
to so many dollars are

00:04:52.020 --> 00:04:56.480 align:middle line:84%
equivalent to so many reduced
cancer treatments-- whatever.

00:04:56.480 --> 00:05:01.880 align:middle line:84%
So the basic question is,
is either one realistic

00:05:01.880 --> 00:05:03.200 align:middle line:90%
or two realistic?

00:05:03.200 --> 00:05:04.940 align:middle line:84%
Is it really possible
that we're going

00:05:04.940 --> 00:05:07.370 align:middle line:84%
to have a single
metric of performance

00:05:07.370 --> 00:05:09.770 align:middle line:90%
for any of our systems?

00:05:09.770 --> 00:05:14.480 align:middle line:84%
And is it even possible that if
we don't have a single metric

00:05:14.480 --> 00:05:16.940 align:middle line:84%
that it's realistic
to put everything

00:05:16.940 --> 00:05:24.980 align:middle line:84%
on a single scale in a
way that's acceptable?

00:05:24.980 --> 00:05:29.870 align:middle line:84%
Are we going to put one
life is equal to $3,822

00:05:29.870 --> 00:05:32.750 align:middle line:84%
or 10 times as much,
whatever it may be?

00:05:32.750 --> 00:05:38.102 align:middle line:84%
Is that kind of analysis
or weighting possible?

00:05:38.102 --> 00:05:39.560 align:middle line:84%
Just think about
that for a moment.

00:05:39.560 --> 00:05:44.310 align:middle line:90%


00:05:44.310 --> 00:05:47.910 align:middle line:84%
In some ways, you'd say,
well, yeah, of course not.

00:05:47.910 --> 00:05:49.740 align:middle line:84%
We don't really know
what a life is worth

00:05:49.740 --> 00:05:53.760 align:middle line:84%
or how we have years
of schooling translated

00:05:53.760 --> 00:05:58.230 align:middle line:90%
into dollars or whatever.

00:05:58.230 --> 00:06:00.650 align:middle line:84%
But the fact is if you look
through the literature,

00:06:00.650 --> 00:06:03.230 align:middle line:84%
there are all kinds of ways
of saying, yes, we can do it.

00:06:03.230 --> 00:06:05.880 align:middle line:84%
We can have multi-objective
optimization,

00:06:05.880 --> 00:06:10.190 align:middle line:84%
which means that we have some
kind of weight on factor one

00:06:10.190 --> 00:06:12.410 align:middle line:84%
and another kind of
weight on factor two

00:06:12.410 --> 00:06:14.450 align:middle line:84%
and another kind of
weight on factor three.

00:06:14.450 --> 00:06:16.580 align:middle line:84%
And if we accept
that those weights

00:06:16.580 --> 00:06:20.750 align:middle line:84%
for those different factors,
which are basically relative

00:06:20.750 --> 00:06:24.170 align:middle line:84%
values between one or the
other, if we accept those,

00:06:24.170 --> 00:06:27.380 align:middle line:84%
then, yes, we can think about
multi-objective evaluation.

00:06:27.380 --> 00:06:31.680 align:middle line:84%
But it isn't a thing that
we can go through and say,

00:06:31.680 --> 00:06:33.920 align:middle line:84%
well, those weights are
sacred, and we really

00:06:33.920 --> 00:06:34.970 align:middle line:90%
know what they are.

00:06:34.970 --> 00:06:36.470 align:middle line:90%
They are assumptions.

00:06:36.470 --> 00:06:40.730 align:middle line:84%
And even if they are locally
reasonably acceptable--

00:06:40.730 --> 00:06:43.850 align:middle line:84%
locally in terms of
over a certain range--

00:06:43.850 --> 00:06:47.790 align:middle line:84%
at summing, they're probably
nonlinear weighting functions

00:06:47.790 --> 00:06:54.170 align:middle line:84%
so that a person's function in
terms of having enough to eat

00:06:54.170 --> 00:06:57.620 align:middle line:90%
and enough shelter on them--

00:06:57.620 --> 00:06:59.780 align:middle line:84%
presumably, a
first-order consideration

00:06:59.780 --> 00:07:04.700 align:middle line:84%
is do they have enough eat and
then when they have enough eat,

00:07:04.700 --> 00:07:06.260 align:middle line:84%
it's not so important
to have more,

00:07:06.260 --> 00:07:08.840 align:middle line:84%
but it's more important
to have better shelter.

00:07:08.840 --> 00:07:11.150 align:middle line:84%
So there's a
continuum of changes

00:07:11.150 --> 00:07:12.830 align:middle line:90%
in what these weights are.

00:07:12.830 --> 00:07:16.460 align:middle line:84%
So what I'd like you
to consider and I'm

00:07:16.460 --> 00:07:20.630 align:middle line:84%
happy to discuss with
you, but the issue

00:07:20.630 --> 00:07:24.170 align:middle line:84%
is that it's
unrealistic to suppose

00:07:24.170 --> 00:07:27.740 align:middle line:84%
that we can have either a
single metric of performance

00:07:27.740 --> 00:07:32.510 align:middle line:84%
or that we can put all the other
metrics onto a single scale

00:07:32.510 --> 00:07:34.830 align:middle line:84%
that's acceptable
and, therefore,

00:07:34.830 --> 00:07:37.940 align:middle line:84%
that we can talk about a
mathematical optimization

00:07:37.940 --> 00:07:41.870 align:middle line:84%
to find the extreme high or
low of all the possibilities.

00:07:41.870 --> 00:07:47.710 align:middle line:84%
That is a fundamental thought
I'd like you to keep in mind.

00:07:47.710 --> 00:07:53.000 align:middle line:90%