Ricochet is the best place on the internet to discuss the issues of the day, either through commenting on posts or writing your own for our active and dynamic community in a fully moderated environment. In addition, the Ricochet Audio Network offers over 50 original podcasts with new episodes released every day.
Trump –> Armageddon: A Few Scenarios
Yesterday Genferei made a request:
Claire: Try to write a scenario where Trump causes Armageddon. Don’t leave out any steps or resort to hand-waving or amateur psychology or appeals to authority. Don’t forget there are a squillion hangers-on, advisers and career civil servants and/or soldiers involved. Perhaps you’ll convince us. The “I don’t know but it feels scary” isn’t convincing me.
Great question, and exactly why I love Ricochet: Sooner or later, someone’s going to point out exactly where your argument’s a little vague or flabby, and you’ll either tighten up your argument or change your mind, both or which are good outcomes.
It’s going to take me a couple of weeks to make this case, because I want to do this carefully. I don’t want to write a book in a single post, then come back to see all the tl;drs at the end. So here are the the argument I’ll make in the coming weeks, not necessarily exactly in this order:
- I’ll argue that a nuclear war, even a limited one, would be a catastrophe for the United States, as would a major global war like the First or Second World War. I’ll also note that we now face a number of other, very serious, national security threats.
- I’ll argue that the probability of the outbreak of such a war in the next president’s term or terms is greater than it has been since the end of the Cold War, and greater than at many points during the Cold War. I believe the probability will be unusually high no matter who’s elected president. My argument will be based on fairly standard and widely-accepted theories about why wars among great powers break out.
- To draw analogies that may be relevant, I’ll look at the origins of previous great-power conflicts, particularly the First and Second World Wars, but also at other unusually catastrophic and costly wars that broke out among powers akin to the United States and its present-day competitors. (I’ll also explain why I think they’re relevantly similar.)
- I may also consider the risk of civil war, and why it might be slightly higher under Trump than other presidents, although I still think it’s quite unlikely.
- I’ll describe in some detail the nuclear near-misses of the Cold War, some of which may still be unknown to all of you, and more of which, I’d assume, are unknown to all of us. I’ll see if we can draw relevant conclusions about why these near-misses didn’t become misses. (Tangential: Why do we call it a “near-miss?” Surely we mean a “near-launch?” Anyone know?) I’ll argue that because we’ve been very lucky from 1945 to the present, we tend to underestimate the risk and see such a war as impossible. I’ll argue that it’s not.
- I’ll make the argument that in matters of foreign policy and war, the US president is far less constrained by institutional checks and balances than he is in matters of domestic policy. Moreover, the powers of the executive during wartime were markedly enlarged after September 11, and few of these powers have been withdrawn.
- I’ll ask how many advisors, staffers, bureaucrats, hangers-on, advisers, or DoD officials truly have the power to interfere with the commander-in-chief should he make a decision they think unwise. I’ll ask, for example, “How many steps does it take to launch a nuclear weapon?” (fewer than you’d think), and ask as well what we know, historically speaking, about the willingness of soldiers to follow illegal or unwise orders. I’ll try to come up with an estimate — based on what we know of similar situations in the past — of the likelihood that his bad judgment would be questioned or his orders disobeyed in an ambiguous situation that’s widely and plausibly perceived as a great threat.
- I’ll walk you through several plausible scenarios in which the president would have to make very quick decisions in response to an emergency, scenarios in which the making the wrong decision would be catastrophic.
- I’ll sketch out what the president might do using several hypothetical versions of Donald Trump, all based on things he’s said during the campaign or my observations of him in “The Apprentice.” We don’t know which things he really means, and they often contradict each other, so I’ll try creating a number of plausible Donalds. They’ll range from “Secret-Churchill Donald” — someone who campaigns as a lying fool because he knows this is effective, but unknown to the public has an alter-ego who’s a highly-informed strategic genius surrounded by competent and experienced foreign policy advisors who challenge his assumptions ruthlessly. For this Donald I’ll assume he and his advisors share the goal of furthering American interests. On the other end of the spectrum might be “Psychopath Donald,” a man who would score a full 40 on Hare’s Psychopathy test (click the link to read what that is), and who would neither surround himself with competent advisers, nor take anyone’s advice, nor act toward any goal save that of keeping himself entertained and stimulated. I’ll then try to estimate the odds of his being or behaving as these different alters, and I’ll ask how these alters would be apt to handle the scenarios I’ve suggested in Step 7. I’ll try to sketch out a more rigorous way of calculating “odds of Armageddon” based on that.
- I’ll also sketch out what the world might look like if he followed through with various things he’s said he’ll do, using the most common-sensical, plain-English interpretation of his words, and argue that some of these things would be likely to raise the risk of global or thermonuclear war even higher than it already is. Some of the things he’s said are contradictory, so I’ll sketch out both or all three or four scenarios. I’ll predict the effect these actions would be apt to have based on the best historical analogies I can find. I’ll offer some evidence of how these statements, even if he has no intention of acting on them, have already changed the perception of America among its allies, enemies, non-aligned states, and terrorist entities, making us less secure. I’ll outline how they would change even more dramatically if in office he acts on his campaign promises (as best I understand them), and what the implications of this would be. I’ll make the case that even if judgements such as these are incorrect or unfair, other states will be obliged, out of an abundance of caution, to prepare for worst-case scenarios, and thus their fear of him will tend to be self-fulfilling.
- I’ll look at two ways he could end up making decisions, good or bad, unconstrained by the usual checks on a president’s power. The first is the “sudden shock” scenario — something like September 11, or another highly traumatic event, after which the checks on his power might literally be gone (a plane hits Congress while it’s in session, or a bomb takes out the Supreme Court), or easily overridden (think of Trump’s gift for demagoguery, of our known willingness to accede to all kinds of liberty-killing legislation in the wake of a terrorist attack, our generally poor understanding of how our government is supposed to work and the importance of checks and balances, and how easily that combination could be exploited if Americans were even more frightened than they were on September 11.).
- The second is the “slow accretion of untrammelled power” scenario to which I alluded in a comment yesterday. As I wrote, “His personality reminds me [not so much as Hitler but of] Recep Tayyip Erdoğan. … If you look at the last line [of a piece I wrote for City Journal about Istanbul in 2010], you’ll see I wrote it when we didn’t yet know what would happen next. That what happened next has been catastrophic makes me all the more uneasy about Trump’s personality. We may have checks and balances sufficient to contain him for a while, but over the course of two terms, even enormously secure restraints can wear thin. Tayyip managed seriatum to discredit the military and imprison the top brass, stack the courts, stack the bureaucracy, quash the press, transform the Constitution, and ultimately make it impossible to get rid of him. It’s easier to do than you’d think.” In other words, I’ll sketch out the way I’ve personally seen a charismatic, shrewd and power-hungry leader undermine checks on his power that were widely believed to be nearly-failsafe.
That’s twelve posts, which I’ll work through over the coming two weeks. Genferei, would you consider the challenge met if by this line of reasoning I derive, “Armageddon is a higher risk with Donald Trump in office than it is with any other plausible aspirant to the presidency?” If not, does this mean that no argument would convince you, or does it mean you’re looking for a different kind of argument? If so, what kind of argument would that be?
(A closing thought: I’ll make the argument more fully and seriously in days to come, but how do you reckon Donald would score on the Hare Psychopathy test? Genferei asked me to eschew amateur psychology — as would Hare himself — but I reckon Trump won’t be sitting down with a professional psychologist anytime soon, so what choice have we but to practice amateur psychology? Run Trump through the scale as dispassionately as you can. What number do you get, roughly?)
Published in Foreign Policy, General, History
There is no reason to make this artificial distinction between subjective and objective probability. We’re talking about good reasoning under uncertainty. Since the axioms of probability have been shown to be ideal to this task (See Jaynes, chapter 2), they should be used in all cases of uncertainty and partial information if you want reliable conclusions.
Given our priors (and usually data too), we do want our reasoning to be fully objective in a sense Jaynes discusses on p. 373:
(A) The prior probabilities represent our prior information, and are to be determined, not by introspection, but by logical analysis of that information.
(B) Since the final conclusions depend necessarily on both the prior information and the data, it follows that, in formulating a problem, one must specify the prior information to be used just as fully as one specifies the data.
(C) Our goal is that inferences are to be completely ‘objective’ in the sense that two persons with the same prior information must assign the same prior probabilities.
Sometimes aspects of our priors can have a logical basis that seems almost objective, such as when we use frequency estimates, ignorance priors, transformation groups, or maximum entropy, but when dealing with the uncertainty of nontrivial human interactions, I expect there will always be a large subjective component. No way to avoid that.
No one requires your priors to match someone else’s.
Mr. Walla,
Can you not have the good grace to concede on this matter? Priors can vary greatly, yes. But that is just the beginning of the difficulties. That is a microscopic droplet compared to the vast ocean of understanding that says history and human conduct in general is totally unsuited to probabilistic analysis. Even Economics has a severe sense of its limitations, and that discipline is considerably more predictable than the affairs of nations projected into the future.
This Gilboa et al paper is a meta article arguing that some method would be objective if used. Thanks for the reference, but I was hoping for an actual paper with actual probabilistic conclusions that you personally find objective.
But the paper makes the classic mistake of confusing majority subjective opinion with objectivity:
No, what I said was that, though some aspects of the calculation are likely objective, the final conclusion always has a subjective component. I offered to find that subjective component in any paper you wish to put forth as having no subjectivity.
If my argument has a false premise but includes the objective logic that 2+2=4, my final conclusion was still subjective.
Actually, it often is. Researchers may choose how far back to go in their time-series sample, they choose how to classify things in their sample, they may choose to censure outliers, they may have to decide how to handle missing data. All of these subjective choices affect the final answer, making the final answer also subjective.
It is a subjective choice that affects the final stated probability, rendering it subjective.
I noticed that you didn’t answer this question, which would clarify your point? How are you not dismissing all empirical knowledge as subjective?
Why aren’t you answering?
I think that I bulleted the circular argument to clarify my position in my first two replies to him, but he continued to misread the original post as a statement of my beliefs on probability in general instead of a point about what would be necessary to rectify an outline of arguments that are fundamentally circular.
This assumes a particular functional form for the failure density, which is not always terrific. For example, it is a terrible model for redundant systems. Also, with incomplete information, estimates of failure rates usually incorporate subjective information. For example, what if no failure has yet been observed?—what is your MTBF in that case?
The formula is therefore an assumption, an opinion, and people can differ on its appropriateness in individual cases. It is therefore not objective.
However, if you *assume* it is true, and if you agree with all the numbers plugged into the formula, then the answer you get (if you don’t make a mistake) is objective.
You’re either contradicting yourself or you’re saying what I’ve said from the beginning. Given premises, a probability can be given an objective interpretation. Given your hard stance on objectivity, you shouldn’t be writing something like the first bold sentence.
Surely if something is incorrect, it cannot also be objective, therefore it must be subjective.
I not equating the two, but both are subjective. One would hope we always try to be as objective as we can, but some subjective element always remains in most subjects, including engineering, weather, and human interactions.
No, not as good in general, though arguments from assumptions may be fatally hurt by wrong or useless assumptions.
Wasn’t it NASA engineers who designed rockets that exploded and space telescopes that failed? That proves that their reasoning had a subjective component.
Since this is what I’ve been saying all along, I think that this represents a concession.
You’re still ducking my question about the NASA engineer and calculations of trajectories. Are such results subjective? You should answer the question so that people can evaluate the consistency of your position and how you’re using words like objective and subjective. I say that you’re playing word games. Answer the question so that I can see if you’re just striking a convenient radical skeptic’s pose.
In common usage, subjective does not equal incorrect. If a result based on the modeling is correct, is it subjective? Yes or no. Can a modeling choice be completely a matter of opinion, or are then constrained by reality? Answer the question instead of ducking it.
This all boils down to you, in violation of the principle of charity, cherry picking a self-serving definition of subjective. You appear to be resting your argument on definition (1) or (4),
when it suits you. You know that everyone including the adherents of users of the descriptions subjective and objective in relation to the interpretation of probability are using the definitions (2) and (3),
Your chosen definition isn’t privileged in argument when you attempt to claim that someone else is wrong. When refuting a statement, you have to work with the terms as used, not redefine them to suit you.
Is this really the style you wish to use? In my replies, I’ll be deleting any of your comments in this style, for they don’t further discussion.
We can say that MTTF derived probabilities of failure are more objective than estimates of the probability of rain tomorrow at noon. Sound okay to you? But still, as I have mentioned in #156, using these MTTF numbers is not an exercise in pure objectivity.
I disagree with this assessment. Most of economics is uninterested in modeling human opinion per se.
Ditto my above.
If this is your opinion now, terrific, but that’s not what you originally said in #88:
In your #88, you’re telling Claire that she must
You see? You were asking for formal objective proof, or “legal or scientific proof “that are “reproducible […] and therefore not subjective”.
True. Some probabilities, even those claimed to be objective, fail to be correct reflections of reality.
Goodness, I did answer it. Have you failed to notice the numbers I put at the top, like [#154] above? I am multitasking lately, but my answers to you are in order of your writing.
You omitted the phrase, something like, which critically affects the meaning of my request, and if you didn’t understand my point, that premises cannot be used to justify premises, I clarified it in the next point.
Mathematically, “equal” and “implies” are different. I did not say or think that subjective “equals” incorrect.
Objective beliefs are contacts with reality, so they should be correct. An incorrect belief therefore cannot be objective, therefore it is subjective.
If you are asserting that some beliefs are objective, objectively true, but also incorrect, I am surprised. Example please?
Do you suppose that the NASA scientists watching Columbia explode imagined that their computations were perfectly objective, and no self analysis or rethinking was needed for next time?
You’re cherry-picking what can be objective. Even in reasoning choices are made. The axiomatic foundations of a deductive system are chosen, and the choices have consequence for the set of provable statements. Whether you embed a formalization of deductive systems in a set theory that you take to be primary or whether you embed set theory in a formal system of deduction whose set of entities you take to be primary is a choice. (See the book by Peter Andrews, Introduction to Mathematical Logic and Type Theory: To Truth Through Proof.)
You’re just declaring what you want to be objective and excluding what you want without presenting any definition that can draw a stable distinction between your categories. You’re a radical skeptic; I get it. I don’t agree, and I’m more of an empiricist, and I’m not interested in engaging you in a pointless debate about skepticism. I’ve been very clear on the point. If you want to talk allow any subjectivity, give a stable definition, one that doesn’t ultimately consume everything.
No, I’m asserting that a mistake is not a judgment or consequence of reasoning any more than the failure of a complex system to perform its function is the result of any mind.
Your post is just a long-winded way of drawing a distinction between objects in reality and objects in the mind without getting to the point, answering if the mind can know anything at all. Your argument is the hard skeptics position and the one that eliminates the possibility of all reasoning.
Personally, I consider myself far less skeptical than you, Joseph. You were the one, in #88, telling Claire what she has to do to back up her probabilities to make them objective:
I was always interested in her subject and proposed topics. I said all along that probabilities can be useful even in situations that are obviously lacking in the kind of objectivity you insisted on.
Further, I don’t think humans can get through an average day without using the kind of logical probabilistic reasoning that we have been naturally selected for. This reasoning, called extended logic by E.T. Jaynes, can operate from information of many kinds, including subjective judgement. Mostly, it is not objective.
Further still, businesses use this kind of reasoning from subjective information all the time. Their information when new possibilities emerge is limited and not objective. How else but through subjective probability can they make major decisions? I’m thinking now of “How to Measure Anything: Finding the Value of Intangibles in Business.”
I think analysis of the desirability of a potential president is often much like this—measuring intangibles. I think it is possible to do this quite skillfully, and such methods are essentially Bayesian, and subjective.