WEBVTT

1
00:00:00.620 --> 00:00:06.480
Optimocracy: The Evidence Machine On Wishonia, the Optimitron comes free with your first government.

2
00:00:07.460 --> 00:00:15.400
Its Optimal Budget Generator (O.B.G.) and Optimal Policy Generator (O.P.G.) check which policies work by looking at the

3
00:00:15.520 --> 00:00:21.240
ten thousand places that already tried them. Your species skipped this step and went straight to arguing.

4
00:00:21.800 --> 00:00:26.740
Incentive Alignment Bonds make saving lives and advancing careers the same sentence.

5
00:00:27.440 --> 00:00:33.380
One problem: how do you know which policies to recommend? The Only Two Numbers That Matter.

6
00:00:34.260 --> 00:00:39.180
Two numbers. Just two. Real after-tax median income growth.

7
00:00:39.960 --> 00:00:44.100
Did the person in the middle get richer this year? Not G.D.P.

8
00:00:44.100 --> 00:00:49.440
per capita, which looks great when three billionaires move into your city while everyone else's rent doubles.

9
00:00:50.380 --> 00:00:56.670
Median healthy life years. How many years can the person in the middle expect to live in good health?

10
00:00:57.540 --> 00:01:05.019
Not average life expectancy, which counts the billionaire who replaced his blood six times as canceling out someone who

11
00:01:05.019 --> 00:01:14.180
died at forty. Why these two? Because everything important your governments do eventually makes people richer or less dead,

12
00:01:14.520 --> 00:01:21.900
and if it doesn't do either, it wasn't important. Your government currently measures success by how much money it spends,

13
00:01:22.280 --> 00:01:26.080
which is like a restaurant measuring success by how much food it buys.

14
00:01:26.700 --> 00:01:29.940
You could buy a mountain of ingredients and poison everyone.

15
00:01:30.580 --> 00:01:38.880
By your current metrics, that's a good year. Your species engineered a pocket-sized device with sixteen billion transistors,

16
00:01:39.100 --> 00:01:42.840
mass spectrometry, and supply chains spanning forty-three countries.

17
00:01:43.580 --> 00:01:52.540
Then you built an algorithm that A B tests billions of user interactions, optimizing hundreds of signals in real time, so

18
00:01:52.540 --> 00:01:59.960
that device can show you the optimal twerking video. That is more empirical rigor than your species applies to allocating

19
00:01:59.960 --> 00:02:07.340
its federal budget. The federal budget decides who gets medicine, who gets clean water, and who gets bombed.

20
00:02:08.520 --> 00:02:14.000
Fifty-five million people die every year from diseases you already have treatments for.

21
00:02:14.740 --> 00:02:19.400
You optimized the twerking. You did not optimize the not-dying.

22
00:02:21.040 --> 00:02:27.780
Here's what makes these numbers useful beyond measurement: your species already agrees on both goals.

23
00:02:28.540 --> 00:02:36.360
Nobody campaigns on "make people poorer and sicker." But without a scoreboard, policy debates become tribal arguments about

24
00:02:36.360 --> 00:02:44.800
vibes. You forgot the question. You're not arguing about whether people should be healthier and richer; you're arguing about

25
00:02:44.800 --> 00:02:53.020
which policies achieve that, and nobody's checking. Once you check, most of the argument disappears, because it turns out

26
00:02:53.020 --> 00:02:59.580
the answer isn't left or right. It's in the data from the ten thousand places that already tried it.

27
00:03:01.320 --> 00:03:05.820
Now that you know what we're measuring, here's how badly your species is doing at it.

28
00:03:06.920 --> 00:03:14.320
The Price of Not Checking. The United States spends three hundred percentage points more of G.D.P.

29
00:03:14.360 --> 00:03:20.410
than Switzerland on government. Switzerland gets six point five more years of life expectancy.

30
00:03:21.180 --> 00:03:25.740
One of these countries is doing it wrong. On Wishonia, we would check which one.

31
00:03:26.140 --> 00:03:31.260
On Earth, you have a debate about it. The debate has been going on for fifty years.

32
00:03:31.860 --> 00:03:38.220
The Swiss are still winning. You're paying more and dying sooner, and your response is to have opinions about it.

33
00:03:38.860 --> 00:03:46.000
Documented U.S. governance waste totals four point nine trillion dollars (seventeen percent of G.D.P.).

34
00:03:46.740 --> 00:03:54.920
Globally, the opportunity cost of suboptimal policy reaches one hundred and one trillion dollars (eighty-seven point eight

35
00:03:54.920 --> 00:04:01.360
percent of G.D.P.), putting your civilization's governance efficiency at fifty-one point nine percent.

36
00:04:02.160 --> 00:04:05.680
Your species is running at roughly half its technological potential.

37
00:04:06.580 --> 00:04:13.360
Imagine building a car, then only turning on half the cylinders because the other half are controlled by a committee that

38
00:04:13.360 --> 00:04:17.160
meets once a year and disagrees about what "forward" means.

39
00:04:17.860 --> 00:04:21.700
(For the full accounting, see the Political Dysfunction Tax and U.S.

40
00:04:21.899 --> 00:04:29.940
Efficiency Audit.) Or consider the War on Drugs. Your species spends ninety billion dollars per year on it.

41
00:04:30.540 --> 00:04:37.100
When Nixon declared it in nineteen seventy-one, about six thousand Americans died annually from drug overdoses.

42
00:04:37.860 --> 00:04:43.500
After fifty years and over a trillion dollars, that number exceeded one hundred thousand.

43
00:04:44.200 --> 00:04:47.020
At no point did anyone check whether the policy was working.

44
00:04:48.200 --> 00:04:55.740
At no point did anyone stop funding it because it wasn't. On Wishonia, if you spend a trillion dollars on something and

45
00:04:55.740 --> 00:05:02.280
the problem gets eighteen times worse, you stop doing it. On Earth, you increase the budget.

46
00:05:03.480 --> 00:05:11.200
The Answer Key You Already Have. Your planet has roughly ten thousand jurisdictions making different policy choices.

47
00:05:12.020 --> 00:05:20.560
Kansas cut education funding; Minnesota increased it. Fifty states adopted different minimum wages over thirty years.

48
00:05:21.260 --> 00:05:29.140
Portugal decriminalized drugs; Singapore didn't. These are natural experiments, and the math can now figure out which policy

49
00:05:29.140 --> 00:05:33.420
caused which result, without needing anyone's permission to run an experiment.

50
00:05:34.020 --> 00:05:37.860
The experiment already happened. You just weren't paying attention.

51
00:05:38.760 --> 00:05:44.540
Same logic as your decentralized F.D.A.: watch what actually happens, then analyze the data.

52
00:05:45.240 --> 00:05:50.740
Your decentralized F.D.A. applies this to medicine. The Evidence Machine applies it to governance.

53
00:05:51.320 --> 00:05:55.460
Instead of "which drug works?" you're asking "which policy works?"

54
00:05:55.880 --> 00:06:00.760
Same question, higher body count. Why Knowing Isn't Enough.

55
00:06:01.380 --> 00:06:08.440
If you're thinking "just publish the results and politicians will do the right thing," I have some bad news about your species.

56
00:06:09.100 --> 00:06:14.020
The Copenhagen Consensus already publishes rigorous benefit-cost analyses.

57
00:06:14.920 --> 00:06:22.540
Childhood vaccination: one hundred and one to one returns. Pragmatic clinical trials: six hundred and thirty-seven to one.

58
00:06:23.150 --> 00:06:28.010
Agricultural subsidies: less than one to one. The answer key is public.

59
00:06:28.080 --> 00:06:36.000
It's been public for years. Your governments ignore it. You have the cheat sheet, and you're still failing the test on purpose.

60
00:06:36.460 --> 00:06:44.380
Why? Because information doesn't change incentives. A Princeton study analyzed one thousand seven hundred and seventy-nine

61
00:06:44.380 --> 00:06:50.840
policy decisions and found that average citizens have "little or no independent influence" on policy outcomes.

62
00:06:52.160 --> 00:06:55.100
Economic elites and organized interest groups drive policy.

63
00:06:55.720 --> 00:06:59.260
You already knew this (it's covered in Unrepresentative Democracy).

64
00:06:59.800 --> 00:07:03.480
The point here is that publishing better data doesn't fix the problem.

65
00:07:03.980 --> 00:07:07.240
Your politicians already know which programs produce value.

66
00:07:07.820 --> 00:07:14.600
They don't act because acting doesn't maximize getting votes, getting papers, or getting a nice job afterward selling access

67
00:07:14.600 --> 00:07:22.600
to their replacements. This is the gap between Incentive Alignment Bonds (which solve the incentive problem) and the Evidence

68
00:07:22.600 --> 00:07:28.220
Machine (which solves the knowledge problem). I.A.B.s make doing the right thing profitable.

69
00:07:28.860 --> 00:07:34.920
But they need to know what the right thing IS. That requires a measurement system nobody can corrupt.

70
00:07:35.740 --> 00:07:42.020
The Mechanism: Recommend, Track, Reward. The Evidence Machine does three things.

71
00:07:42.060 --> 00:07:48.280
That's it. Three things. Your species has difficulty with three things, so I'll use small words.

72
00:07:48.900 --> 00:07:52.420
(On Wishonia, we just call this the Optimitron. It's an appliance.

73
00:07:52.900 --> 00:07:56.920
You plug in policies, it tells you which ones work.). Recommend.

74
00:07:56.980 --> 00:08:03.520
The Optimitron has two settings. The Optimal Budget Generator answers "how much should you spend on what?"

75
00:08:04.080 --> 00:08:09.140
The Optimal Policy Generator answers "which laws should you enact, replace, or repeal?"

76
00:08:09.860 --> 00:08:16.500
Here's a sample of what the budget output looks like, applied to the United States: For pragmatic clinical trials, the current

77
00:08:16.500 --> 00:08:24.120
budget is zero point five billion dollars, the optimal budget is fifty billion dollars, for a gap of plus forty-nine point

78
00:08:24.120 --> 00:08:30.140
five billion dollars, based on evidence grade A from randomized controlled trials, or R.C.T.s.

79
00:08:30.880 --> 00:08:37.940
For vaccinations, the current budget is eight billion dollars, the optimal budget is thirty-five billion dollars, for a

80
00:08:37.940 --> 00:08:43.159
gap of plus twenty-seven billion dollars, based on evidence grade A from R.C.T.s.

81
00:08:43.919 --> 00:08:51.140
For military discretionary spending, the current budget is eight hundred and fifty billion dollars, the optimal budget is

82
00:08:51.140 --> 00:08:58.720
four hundred and fifty-nine billion dollars, for a gap of minus three hundred and ninety-one billion dollars, based on evidence

83
00:08:58.720 --> 00:09:06.680
grade C from benchmarks. For agricultural subsidies, the current budget is twenty-five billion dollars, the optimal budget

84
00:09:06.680 --> 00:09:14.260
is zero dollars, for a gap of minus twenty-five billion dollars, based on evidence grade A from welfare analysis.

85
00:09:14.780 --> 00:09:21.580
You are spending one hundred times less than optimal on figuring out which medicines work, and paying farmers twenty-five

86
00:09:21.580 --> 00:09:27.320
billion dollars a year to grow food nobody asked for. The Optimitron noticed.

87
00:09:27.860 --> 00:09:34.480
It also noticed you could cut three hundred and ninety-one billion dollars from the murder budget and still have the largest

88
00:09:34.480 --> 00:09:43.260
military on the planet. The policy generator then does the same thing law by law: enact this, repeal that, adjust the other

89
00:09:43.260 --> 00:09:48.960
thing. Every recommendation traceable to what actually happened when another jurisdiction tried it.

90
00:09:49.720 --> 00:09:58.090
No opinions. Just receipts. Track. When politicians vote, record whether they aligned with the evidence-based recommendation.

91
00:09:58.900 --> 00:10:04.480
Senator Smith: seventy-eight percent aligned. Senator Jones: thirty-four percent aligned.

92
00:10:04.480 --> 00:10:11.920
This information is public. Any citizen can check. On Wishonia, this is called "checking if the person you hired did the

93
00:10:11.920 --> 00:10:17.810
thing you hired them to do." Your species finds this concept revolutionary, which tells you everything.

94
00:10:18.660 --> 00:10:22.120
Politicians who ignore the evidence must do so on the record.

95
00:10:22.820 --> 00:10:31.020
Reward. The Incentive Alignment Bond, or I.A.B.-funded SuperPAC allocates campaign support proportional to alignment scores.

96
00:10:31.640 --> 00:10:35.480
Follow evidence, get funded. Ignore it, watch your opponent get funded.

97
00:10:36.020 --> 00:10:40.900
Same mechanism the N.R.A. perfected, but pointed at "not dying" instead of "more guns."

98
00:10:41.980 --> 00:10:48.880
This doesn't require government permission. The Evidence Machine operates as a permanent advisory layer outside government.

99
00:10:49.560 --> 00:10:56.680
Politicians still vote however they want. They can still ignore the evidence; it just costs them money now, which on Earth

100
00:10:56.680 --> 00:11:03.060
is apparently the only consequence that matters. The question isn't "will politicians give up power?"

101
00:11:03.340 --> 00:11:07.360
(they won't) but "can we make ignoring good advice unprofitable?"

102
00:11:08.060 --> 00:11:13.740
(yes; it's a fundraising challenge, not a revolution). Corruption, But Make It Difficult.

103
00:11:14.460 --> 00:11:18.980
"But won't someone corrupt the measurement system?" Yes. This is the right question.

104
00:11:19.420 --> 00:11:24.920
It's the only question that matters. Currently, your lobbying industry has thousands of capture points.

105
00:11:25.780 --> 00:11:32.690
Committee earmarks, agency decisions, regulatory rulings, budget amendments, riders on unrelated bills.

106
00:11:32.800 --> 00:11:41.340
Each one is cheap to influence. Your federal lobbyists spend four point four billion dollars per year with estimated one

107
00:11:41.340 --> 00:11:46.540
hundred to one returns. Your democracy is a vending machine, and the buttons are labeled.

108
00:11:47.440 --> 00:11:53.900
The Evidence Machine collapses all those capture points to a single target: the measurement methodology.

109
00:11:54.640 --> 00:12:02.020
Here's the heist: simultaneously bribe the Census Bureau, the Federal Reserve, the Bureau of Labor Statistics, academic

110
00:12:02.020 --> 00:12:07.260
institutions, AND decentralized citizen surveys. All without detection.

111
00:12:07.740 --> 00:12:14.820
All at the same time. All institutions with different governance structures, different funding sources, different methodologies,

112
00:12:15.080 --> 00:12:19.040
and different reasons to hate each other. This is not a corruption strategy.

113
00:12:19.500 --> 00:12:24.120
This is the plot of a movie where the criminals lose. Here's how it works in practice.

114
00:12:25.060 --> 00:12:32.520
Five independent sources measure median income growth: The Census Bureau American Community Survey reports plus two point

115
00:12:32.520 --> 00:12:38.420
one percent. The Federal Reserve Survey of Consumer Finances reports plus two point three percent.

116
00:12:39.000 --> 00:12:44.080
The Bureau of Labor Statistics Current Population Survey reports plus one point nine percent.

117
00:12:44.680 --> 00:12:49.980
The University of Michigan Panel Study of Income Dynamics reports plus two point two percent.

118
00:12:50.660 --> 00:12:55.600
The Tax Foundation I.R.S. data analysis reports plus two point zero percent.

119
00:12:56.140 --> 00:13:03.020
Take the median: PLUS TWO POINT ONE PERCENT. If one source reports plus five point zero percent (because someone bribed

120
00:13:03.020 --> 00:13:07.940
them), the median throws it out automatically. No investigation needed.

121
00:13:07.980 --> 00:13:12.260
No whistleblower needed. No committee hearing. Math does the work.

122
00:13:13.180 --> 00:13:18.670
(An honest caveat: these five sources share some underlying data, so they're not perfectly independent.

123
00:13:19.500 --> 00:13:26.720
That's why there's a sixth source: decentralized citizen surveys, collected directly from millions of humans, sharing no

124
00:13:26.720 --> 00:13:32.840
methodology with the others. One person lying has approximately zero effect on the median.

125
00:13:32.940 --> 00:13:39.240
This is statistics, not voting.). Your species has spent centuries making corruption illegal.

126
00:13:39.720 --> 00:13:44.400
It hasn't worked, because laws are enforced by the same humans who are being bribed.

127
00:13:45.000 --> 00:13:49.000
On Wishonia, we tried something different: we made corruption impractical.

128
00:13:49.980 --> 00:13:56.050
There's a difference between "you shouldn't steal" and "the vault has five locks held by five people who hate each other."

129
00:13:56.480 --> 00:14:00.000
Your species tried door number one for three thousand years.

130
00:14:00.560 --> 00:14:06.540
We recommend door number two. This is also why it's different from every previous attempt at technocratic governance.

131
00:14:07.480 --> 00:14:10.940
Soviet central planning had one source of truth (the Politburo).

132
00:14:11.740 --> 00:14:18.280
Credit rating agencies had one source of truth (the agency paid by the people being rated, which is like letting students

133
00:14:18.280 --> 00:14:21.700
grade their own exams and then being surprised when everyone gets an A).

134
00:14:22.380 --> 00:14:27.870
Both were captured immediately, because corrupting one institution is a Tuesday.

135
00:14:28.020 --> 00:14:34.760
Corrupting five independent institutions simultaneously, without any of them noticing or telling anyone, is a conspiracy

136
00:14:34.760 --> 00:14:40.000
theory. The expensive kind that doesn't work. Why not fifty metrics?

137
00:14:40.580 --> 00:14:47.800
Because each additional target creates a new way to cheat, and your species needs no encouragement in that department.

138
00:14:48.140 --> 00:14:53.780
Markets figured this out: firms optimize profit, not two hundred sub-metrics.

139
00:14:53.880 --> 00:14:58.900
Two broad, hard-to-fake numbers are easier to audit than fifty narrow ones.

140
00:14:59.520 --> 00:15:07.460
And these particular numbers are very hard to fake. You can't fabricate purchasing power at scale without actually improving

141
00:15:07.460 --> 00:15:13.200
household economics. Mortality is binary (the most binary thing there is).

142
00:15:13.740 --> 00:15:18.260
You can't fake everyone being "less dead.". "But correlation isn't causation!"

143
00:15:19.360 --> 00:15:25.900
Yes. Thank you. Your species discovered this in the eighteen hundreds and has been using it as an excuse to ignore data

144
00:15:25.900 --> 00:15:32.010
ever since. The Optimitron doesn't just check whether policies and outcomes happened in the same room.

145
00:15:32.420 --> 00:15:39.840
It uses the fact that ten thousand jurisdictions adopted different policies at different times to isolate what actually

146
00:15:39.840 --> 00:15:48.080
caused what. (The boring details: the technical specification scores every causal claim against nine criteria, and if the

147
00:15:48.140 --> 00:15:53.560
timing is wrong, the score is automatically zero, no matter how pretty the correlation looks.).

148
00:15:54.420 --> 00:16:01.820
Putting It Together. The full loop: one percent Treaty funds it, I.A.B.s align incentives, the Evidence Machine identifies

149
00:16:01.820 --> 00:16:08.320
what works, and Wishocracy allocates the research funds. The Evidence Machine tells governments which policies work.

150
00:16:08.880 --> 00:16:14.240
Wishocracy tells the research fund where to aim. On Wishonia, this is called "an obvious system."

151
00:16:14.820 --> 00:16:20.920
On Earth, it required four chapters and a bibliography. And here's why the whole thing ratchets.

152
00:16:22.000 --> 00:16:29.100
Each treaty expansion creates more I.A.B. funding, which funds more evidence collection, which strengthens the case for

153
00:16:29.100 --> 00:16:33.660
further expansion. Evidence begets funding begets evidence.

154
00:16:34.460 --> 00:16:42.640
It's the same self-reinforcing loop as the Military Industrial Complex, except instead of manufacturing orphans, it manufactures

155
00:16:42.640 --> 00:16:50.980
data about how to stop manufacturing orphans. Now that you know how the Optimitron works (congratulations, on Wishonia this

156
00:16:50.980 --> 00:16:56.380
is taught to children), the next chapter shows you how eight billion people allocate the research funds.

157
00:16:57.340 --> 00:17:04.140
For the full technical specification, formal models, and threat analysis, see the Optimocracy academic paper.

158
00:17:05.040 --> 00:17:12.359
For the gory details on how the recommendation engine decides you're spending too much on corn and too little on not dying:

159
00:17:12.420 --> 00:17:21.339
the Optimal Budget Generator at O.B.G. dot war on disease dot org and the Optimal Policy Generator at O.P.G.

160
00:17:21.380 --> 00:17:22.980
dot war on disease dot org.