WEBVTT 00:00:00.000 --> 00:00:02.470 align:middle line:90% [SQUEAKING] 00:00:02.470 --> 00:00:04.446 align:middle line:90% [RUSTLING] 00:00:04.446 --> 00:00:05.434 align:middle line:90% [CLICKING] 00:00:05.434 --> 00:00:12.285 align:middle line:90% 00:00:12.285 --> 00:00:13.910 align:middle line:84% RICHARD DE NEUFVILLE: Decision analysis 00:00:13.910 --> 00:00:17.000 align:middle line:84% provide a structured, efficient way 00:00:17.000 --> 00:00:20.390 align:middle line:84% to help people recognize this complexity, 00:00:20.390 --> 00:00:23.030 align:middle line:84% to recognize the probabilities, and to deal with it. 00:00:23.030 --> 00:00:26.310 align:middle line:90% 00:00:26.310 --> 00:00:31.640 align:middle line:84% So what are the general features of this method? 00:00:31.640 --> 00:00:36.100 align:middle line:84% First of all, it offers a simple way of defining the choices. 00:00:36.100 --> 00:00:40.030 align:middle line:84% Indeed, it assumes that everything 00:00:40.030 --> 00:00:44.470 align:middle line:84% is a discrete choice that is either yes, no, 00:00:44.470 --> 00:00:47.770 align:middle line:84% or yes, somewhat, a lot, maybe, not at all. 00:00:47.770 --> 00:00:49.600 align:middle line:90% But it has discrete-- 00:00:49.600 --> 00:00:53.470 align:middle line:84% it forces you to think in terms of discrete choices, not 00:00:53.470 --> 00:00:56.150 align:middle line:90% a continuum. 00:00:56.150 --> 00:01:01.240 align:middle line:84% It also enables you, but generally compels you, 00:01:01.240 --> 00:01:02.860 align:middle line:90% to look over several periods. 00:01:02.860 --> 00:01:05.590 align:middle line:84% That is, it's not just what you do in the first period, 00:01:05.590 --> 00:01:07.510 align:middle line:90% but then what you do. 00:01:07.510 --> 00:01:11.590 align:middle line:84% Just like going to a course at MIT, 00:01:11.590 --> 00:01:14.350 align:middle line:84% you have your choice of first semester courses. 00:01:14.350 --> 00:01:18.220 align:middle line:84% And the choices you make then enable you to do certain things 00:01:18.220 --> 00:01:21.430 align:middle line:84% or prevent you from doing other things because you didn't take 00:01:21.430 --> 00:01:25.570 align:middle line:84% the prerequisite course and so forth so that you want to think 00:01:25.570 --> 00:01:29.000 align:middle line:90% about it over several periods. 00:01:29.000 --> 00:01:31.460 align:middle line:84% The third element, which is key to the whole reason we 00:01:31.460 --> 00:01:37.130 align:middle line:84% even discussing it, is because it deals with uncertainties. 00:01:37.130 --> 00:01:39.750 align:middle line:84% And it is also a standardized method. 00:01:39.750 --> 00:01:43.970 align:middle line:84% So there are tools available to do this, 00:01:43.970 --> 00:01:45.870 align:middle line:90% to do decision analysis. 00:01:45.870 --> 00:01:51.908 align:middle line:84% And so if you are into decision analysis, 00:01:51.908 --> 00:01:54.200 align:middle line:84% there are a lot of other people who do it the same way, 00:01:54.200 --> 00:01:56.310 align:middle line:90% they do it in a standard way. 00:01:56.310 --> 00:01:57.740 align:middle line:90% So it's a structured process. 00:01:57.740 --> 00:01:59.840 align:middle line:90% It's like Excel. 00:01:59.840 --> 00:02:02.960 align:middle line:84% It provides a widespread language 00:02:02.960 --> 00:02:05.870 align:middle line:90% to think about certain things. 00:02:05.870 --> 00:02:11.840 align:middle line:84% And finally, which I think is relevant here, 00:02:11.840 --> 00:02:17.420 align:middle line:84% is that it can include not just a numerical result in terms 00:02:17.420 --> 00:02:21.800 align:middle line:84% of the income or the present value, 00:02:21.800 --> 00:02:26.210 align:middle line:84% but also can deal with risk aversions, your consumer 00:02:26.210 --> 00:02:29.930 align:middle line:84% satisfaction, some kind of notion of utility. 00:02:29.930 --> 00:02:34.220 align:middle line:84% We'll talk more about that in a later session 00:02:34.220 --> 00:02:36.350 align:middle line:90% in the half semester course. 00:02:36.350 --> 00:02:40.670 align:middle line:84% But for the moment, it allows you to say, well, yeah. 00:02:40.670 --> 00:02:45.620 align:middle line:84% We could have on average our outcome is good, 00:02:45.620 --> 00:02:47.330 align:middle line:90% but the bad outcome is-- 00:02:47.330 --> 00:02:50.450 align:middle line:90% I really don't like it at all. 00:02:50.450 --> 00:02:53.990 align:middle line:84% If, for example, if I'm saving for my pension, 00:02:53.990 --> 00:02:57.260 align:middle line:84% I don't just think on average I might have $1 million 00:02:57.260 --> 00:03:01.010 align:middle line:84% to retire on, but it makes a great deal of difference for me 00:03:01.010 --> 00:03:04.610 align:middle line:90% if that million is-- 00:03:04.610 --> 00:03:06.920 align:middle line:84% for an extreme, it's either $2 million or nothing. 00:03:06.920 --> 00:03:09.290 align:middle line:84% That would be an average of a million. 00:03:09.290 --> 00:03:11.600 align:middle line:84% But if I had nothing, I'd be feeling very unhappy 00:03:11.600 --> 00:03:16.490 align:middle line:84% about my retirement and so that I would weight 00:03:16.490 --> 00:03:17.930 align:middle line:90% against that possibility. 00:03:17.930 --> 00:03:24.330 align:middle line:84% So it allows you to think about utility assessment. 00:03:24.330 --> 00:03:25.750 align:middle line:90% Now, all right. 00:03:25.750 --> 00:03:30.870 align:middle line:84% The central element of the decision analysis 00:03:30.870 --> 00:03:32.960 align:middle line:90% is the decision tree. 00:03:32.960 --> 00:03:37.400 align:middle line:84% It shows the choices over several period. 00:03:37.400 --> 00:03:44.190 align:middle line:84% And it has basically four outcomes, four elements. 00:03:44.190 --> 00:03:47.310 align:middle line:84% It has the choices, the structure, the way 00:03:47.310 --> 00:03:49.110 align:middle line:90% you draw it-- 00:03:49.110 --> 00:03:52.260 align:middle line:84% it has the choices and the possible outcomes. 00:03:52.260 --> 00:03:54.220 align:middle line:90% I'll show you those in a moment. 00:03:54.220 --> 00:03:56.760 align:middle line:84% And secondly, it has data, the data 00:03:56.760 --> 00:03:59.760 align:middle line:84% both about uncertainties that may occur 00:03:59.760 --> 00:04:02.680 align:middle line:90% and the value of the outcomes. 00:04:02.680 --> 00:04:10.190 align:middle line:84% So four elements-- choices of the structure and the data. 00:04:10.190 --> 00:04:12.370 align:middle line:90% So here is the structure. 00:04:12.370 --> 00:04:16.790 align:middle line:84% So basically, it's a disciplined way of present alternatives. 00:04:16.790 --> 00:04:19.779 align:middle line:90% And it has two elements. 00:04:19.779 --> 00:04:21.560 align:middle line:90% It has the decision points. 00:04:21.560 --> 00:04:26.500 align:middle line:84% So over on the left-hand side here, I'm starting, 00:04:26.500 --> 00:04:28.190 align:middle line:90% and I have a choice to make. 00:04:28.190 --> 00:04:31.390 align:middle line:84% And in this, I'm going to have a very simple case just 00:04:31.390 --> 00:04:33.540 align:middle line:90% to illustrate it. 00:04:33.540 --> 00:04:37.150 align:middle line:84% I can either do choice one or choice two. 00:04:37.150 --> 00:04:38.470 align:middle line:90% I have these two branches. 00:04:38.470 --> 00:04:41.260 align:middle line:84% Of course, I could have five or six, but or-- 00:04:41.260 --> 00:04:42.250 align:middle line:90% I have a point. 00:04:42.250 --> 00:04:48.880 align:middle line:84% Then, having done something, say I decided to take-- 00:04:48.880 --> 00:04:52.690 align:middle line:84% whatever, I decide to build a plant somewhere, 00:04:52.690 --> 00:04:58.570 align:middle line:84% then I have a result, a chance outcome of it performed well, 00:04:58.570 --> 00:05:01.690 align:middle line:84% not so well, it was a disaster, whatever it might be. 00:05:01.690 --> 00:05:04.510 align:middle line:90% I have possible outcomes. 00:05:04.510 --> 00:05:12.770 align:middle line:84% And for each of the choices, typically known as a branch, 00:05:12.770 --> 00:05:15.550 align:middle line:84% you will have a possible outcomes. 00:05:15.550 --> 00:05:19.970 align:middle line:84% And they will lead you to another place where you say, 00:05:19.970 --> 00:05:22.060 align:middle line:90% all right, I build this plant. 00:05:22.060 --> 00:05:23.600 align:middle line:90% It was a success. 00:05:23.600 --> 00:05:27.490 align:middle line:84% Now, do I build up some more or stay put? 00:05:27.490 --> 00:05:30.700 align:middle line:84% Or I build this plant, it was a failure. 00:05:30.700 --> 00:05:37.250 align:middle line:84% I can now decide to sell it or renovate it, whatever. 00:05:37.250 --> 00:05:42.830 align:middle line:84% So we got sequences-- decision, chance, decision, chance 00:05:42.830 --> 00:05:44.780 align:middle line:90% is the basic structure of this. 00:05:44.780 --> 00:05:48.700 align:middle line:90% 00:05:48.700 --> 00:05:51.640 align:middle line:90% Now, the data, then, has-- 00:05:51.640 --> 00:05:52.180 align:middle line:90% all right. 00:05:52.180 --> 00:05:56.720 align:middle line:84% So I have elements about the outcomes. 00:05:56.720 --> 00:05:59.950 align:middle line:84% So given all this, I've made a decision. 00:05:59.950 --> 00:06:01.540 align:middle line:90% I observe what happened. 00:06:01.540 --> 00:06:02.770 align:middle line:90% I made another decision. 00:06:02.770 --> 00:06:04.160 align:middle line:90% I observe what happens. 00:06:04.160 --> 00:06:07.030 align:middle line:84% And I have an outcome one, outcome two. 00:06:07.030 --> 00:06:08.740 align:middle line:84% And in this particular case where 00:06:08.740 --> 00:06:12.500 align:middle line:84% I have two choices, two outcomes, at each stage, 00:06:12.500 --> 00:06:16.370 align:middle line:84% I have, in this very simple two-stage process, 00:06:16.370 --> 00:06:20.340 align:middle line:90% I have 16 outcomes. 00:06:20.340 --> 00:06:25.040 align:middle line:84% Now, you might ask, well, where is the decision tree? 00:06:25.040 --> 00:06:27.020 align:middle line:84% Well, basically, you've got to think 00:06:27.020 --> 00:06:29.540 align:middle line:84% that this tree is lying on its side 00:06:29.540 --> 00:06:33.830 align:middle line:84% and that this is the root here, and the branches go 00:06:33.830 --> 00:06:37.200 align:middle line:90% up, and leaves and so forth. 00:06:37.200 --> 00:06:39.830 align:middle line:84% And what I invite you to consider, first of all here, 00:06:39.830 --> 00:06:44.060 align:middle line:84% is when does this decision tree become a messy bush? 00:06:44.060 --> 00:06:48.140 align:middle line:84% Because realize that I've only had two choices, 00:06:48.140 --> 00:06:49.860 align:middle line:90% two decisions here. 00:06:49.860 --> 00:06:50.990 align:middle line:90% I could have had five. 00:06:50.990 --> 00:06:52.370 align:middle line:90% And I could have five outcomes. 00:06:52.370 --> 00:06:55.130 align:middle line:90% And now, I'm going to have 25. 00:06:55.130 --> 00:06:57.650 align:middle line:84% And I repeat this, and I'm going to have 25 squared. 00:06:57.650 --> 00:07:00.620 align:middle line:84% And I have 600-some outcomes, et cetera. 00:07:00.620 --> 00:07:04.100 align:middle line:84% It can really easily get very complicated 00:07:04.100 --> 00:07:08.280 align:middle line:84% when I haven't looked at a particularly difficult problem. 00:07:08.280 --> 00:07:11.720 align:middle line:84% So for example, if I were trying to represent a game of chess 00:07:11.720 --> 00:07:19.520 align:middle line:84% here and I could move basically something like eight, 00:07:19.520 --> 00:07:23.540 align:middle line:84% 10 moves at the beginning with my pawns, 00:07:23.540 --> 00:07:30.950 align:middle line:84% and my castle, and my rooks, I would have 10 possible choices 00:07:30.950 --> 00:07:33.950 align:middle line:84% or 12 possible choices here, and then I 00:07:33.950 --> 00:07:35.850 align:middle line:84% could have what my opponent does. 00:07:35.850 --> 00:07:39.290 align:middle line:84% And a realistic problem can explode on. 00:07:39.290 --> 00:07:44.030 align:middle line:84% So one of the issues here, again, back to my emphasis 00:07:44.030 --> 00:07:46.520 align:middle line:84% at the beginning or my remark at the beginning 00:07:46.520 --> 00:07:49.370 align:middle line:84% is that decision analysis is good for simple problems. 00:07:49.370 --> 00:07:54.340 align:middle line:84% For complicated problems, it just gets too messy. 00:07:54.340 --> 00:08:00.000 align:middle line:90%