This came up as a tangent from @habryka and me discussing whether The Hidden Complexity of Wishes was correct.
Is the result of a US presidential election 4 years from now >0.1% contingent on quantum randomness (i.e. is an otherwise omniscient observer forecasting the 2028 election today capable of >99.9% confidence, or is there >0.1% irreducible uncertainty due to quantum mechanics observer-effects)?
I think the answer is yes, because chaotic systems will quickly amplify this randomness to change many facts about the world on election day.
Quantum randomness causes different radioactive decays, which slightly perturb positions of particles around the world by a few nanometers.
Chaotic systems will quickly amplify these tiny perturbations into macro-scale perturbations:
Weather doubles perturbations every 4 days or so
The genes of ~all babies less than 3 years old will be different
Many events relevant to the election are contingent on these differences
Weather-related natural disasters, other circumstances like pandemics (either mutation or lab leak), political gaffes by candidates, assassinations (historically >0.1% and seem pretty random), cancer deaths, etc.
If even a small proportion of election variance is random, you get more than 0.1% election randomness.
Say humanity’s best estimates for the vote margin of the 2028 election have a standard deviation of 76 electoral votes centered on 0. Even if 90% of variance is in theory predictable and only 10% is true randomness (aleatoric), then the nonrandom factors have s.d. 70 and random factors have s.d. 24. If nonrandom factors have 1sd influence, random factors will flip the election with probability well over 0.1%.
In reality it’s much worse than this, because we haven’t even identified 2 leading candidates.
Oliver thinks the answer is no, because in a system as large and complicated as the world, there should be some macro-scale patterns that survive, and an omniscient observer will pick up on all such patterns. Humans are limited to obvious patterns like economic trends and extrapolating polls, but there are likely way more patterns than this, which a forecaster could use to accurately get over well 99.9% confidence.
Who is right?
I think the answer pretty much has to be “yes”, for the following reasons.
As noted in the above post, weather is chaotic.
Elections are sometimes close. For example, the winner of the 2000 presidential election came down to a margin of 537 votes in Florida.
Geographic location correlates reasonably strongly with party preference.
Weather affects specific geographic areas.
Weather influences voter turnout[1] --
During the 2000 election, in Okaloosa County, Florida (at the western tip of the panhandle), 71k of the county’s 171k residents voted, with 52186 votes going to Bush and 16989 votes going to Gore, for a 42% turnout rate.
On the day of November 7, 2000, there was no significant rainfall in Pensacola (which is the closest weather station I could find with records going back that far). A storm which dropped 2 inches of rain on the tip of the Florida panhandle that day would have reduced voter turnout by 1.8%,[1] which would have resulted in a margin that leaned 634 votes closer to Gore. Which would have tipped Florida, which would in turn have tipped the election.
Now, November is the “dry” season in Florida, so heavy rains like that are not incredibly common. Still, they can happen. For example, on 2015-11-02, 2.34 inches of rain fell.[2] That was only one day, out of the 140 days I looked at, which would have flipped the 2000 election, and the 2000 election was, to my knowledge, the closest of the 59 US presidential elections so far. Still, there are a number of other tracks that a storm could have taken, which would also have flipped the 2000 election.[3] And in the 1976 election, somewhat worse weather in the great lakes region would likely have flipped Ohio and Wisconsin, where Carter beat Ford by narrow margins.[4]
So I think “weather, on election day specifically, flips the 2028 election in a way that cannot be foreseen now” is already well over 0.1%. And that’s not even getting into other weather stuff like “how many hurricanes hit the gulf coast in 2028, and where exactly do they land?”.
Gomez, B. T., Hansford, T. G., & Krause, G. A. (2007). The Republicans should pray for rain: Weather, turnout, and voting in US presidential elections.:
I pulled the weather for the week before and after November 7 for the past 10 years from the weather.gov api and that was the highest rainfall date.
Looking at the 2000 election map in Florida, any good thunderstorm in the panhandle, in the northeast corner of the state, or on the west-middle-south of the peninsula would have done the trick.
https://en.wikipedia.org/wiki/1976_United_States_presidential_election—Carter won Ohio and Wisconsin by 11k and 35k votes, respectively.
I think this argument is not sufficient. Turnout effects of weather can flip elections that are already close, and from our limited perspective, more than 0.1% of elections are close. But the question is asking about the 2028 election in particular, which will probably not be so close.
do we have any reason to believe that particular election won’t be close
well, as a conditional argument against it being close: if trump wins in 2024 and enacts project 2025, I expect trump’s successor’s win margin to be an unprecedented-in-the-usa landslide
Do you have greater than 99.9% confidence that it will not be close?
This seems to be focussing on one specific means by which quantum randomness might affect a result.
Another means may be via personal health of a candidate. For example, everyone has pre-cancerous cells that just need the right trigger to form a malignancy, especially in the older people that tend to be candidates in US presidential elections, or for an undetected existing cancer to progress to become serious.
Is there comparable with 0.1% chance that due to a cosmic ray or any other event, that a candidate will have something happen that is serious enough that it affects their ability to run for the 2028 election? It seems likely that the result of an election depends at least moderately strongly upon who is running.
I think I’m with Thomas Kwa on this one, mainly because I see 0.1% as a really low bar. E.g.
let’s say ≥10% chance that the election is close enough to turn on “random-ish things” rather than “structural factors” (random-ish things would include [as discussed in other answers] weather on election day, decisions of particular individuals that “could have gone either way” [to run or not, to endorse or not, to assassinate or not, to release incriminating information on a candidate right before election day or not, whatever Patient Zero of COVID-19 was doing at the time], etc.);
and out of those ≥10%, it seems reasonable to me to guess ≥1% chance that a re-roll of quantum randomness would flip the result.
So that gets us above 0.1%.
I think my guess would be more like, I dunno, 1%–10%?
I think “one of the potential candidates might quantum-randomly die in the timeframe” is a pretty strong argument that there’s at least ~0.1% quantum uncertainty.
ETA: For some stats on this, see this table from the government of Canada. Annual death rate ranges from 0.1% for 35-year olds, up to 0.5% for 55-year-olds, and 3% for 75-year-olds. Multiply those by 4 to get the death rate in the relevant window. Obviously only a small fraction of those deaths will be quantum-randomness-influenced. Also note the relatively high rate of presidential assassinations -- 4 of 45(!) presidents were asssassinated in office(although I assume the “true” probability is lower now)
I don’t think most people die for quantum-randomness reasons. I expect very little probability of someone dying is related to quantum randomness (though my guess is someone might disagree, but then we are just kind of back to the OP question about how much quantum randomness influences macro-level events).
You don’t think so? I think this is clearly the case over someone’s entire life. Starting to condition on a 4-year timescale, I think accidental deaths, assassinations, and viral infections are certainly quantum-randomness-affected at greater than 0.1% probability. Maybe also things like cancer and its progression(dependent on mutations which may or may not happen on a short time scale?) but I don’t really know much about it.
The base rate is too low for all but the oldest segment of the population.
A quick google search says roughly 0.8% death rate of 64-75 year olds over four years[EDIT: see correction downthread], and most of that will be from health problems which show signs far in advance. For younger people, the death rate itself would be below 0.1%.Also, how are you geting 0.8%? This website says that the mortality rate of 65-75 year olds is 2%. So over 4 years that should be 8%, which I think makes it much more plausible that quantum death probability is 0.1%(although clearly 8% isn’t the real probability, any presidential candidate is probably way less likely to die than the background population)
You’re right, I misread. I retract that part of my claim.
Yyyyeah, I’m not totally sure if you get 0.1% just from people dying, hence the ~. But I think it’s at least within a factor of 10, which makes me think the total quantum randomness factor is at least 0.1%. And to defend the “people dying” factor, (a) many of the candidates are in fact pretty old these days (b) presidents have a relatively high rate of being assassinated − 4 of 45(!), although I assume the actual probability is lower now than the historical average (c) randomness within the 4-year window could affect how quickly a pre-existing health problem progresses, although this might result in them dropping out early/not rather than actually dying.
For younger people(in presidential-candidate age ranges) the annual death rate ranges from 0.15% to 0.5%, see here (So the 4-year death rate ranges from 0.6% to 2%)
How about cancer deaths? From the point of view of 2012, was Beau Biden’s death in 2015 after diagnosis in 2013 due to quantum randomness? That sure had a big effect on the Democratic primary, if not the general election.
Ok, so in the box of gas case, you have individual gas molecules following newtonian physics with no process to give any kind of structure at all. However, a human brain of a voter is more like an optimizing process. Most small perturbances to the “house” example, say you go nudge a 2x4 in the materials pile a tiny amount, which is much larger than quantum randomness, and the workers building the house will still put that piece of wood where it goes in the structure, or select an equivalent piece of wood which ends up forming the structure still.
Inside a human brain of a voter, including the primary voters who will determine the candidates for the 2028 election, there is an overall function the brain circuitry tries to satisfy, and many synaptic weights that encode policies that will attempt to satisfy that function. Small fluctuations from quantum randomness are unlikely to change voter preferences, since these are encoded by many weights that reflect life experiences etc. And you can empirically check this—look at experiments where people are presented with new evidence that challenges their currently held beliefs. Generally this just doesn’t work, people interpret the new evidence to support their beliefs, even if a rational agent would not.
What about bigger trends, such as the price of gasoline, or economic activity and employment rates and relative spending power per voter? Same thing here, small quantum randomness is drowned out by larger macroscale trends and inertia, many of the fluctuations cancel out. As in, if you can “reroll” history, and the fluctuations have the same distribution as before, but a different random seed, many of these larger trends would happen identically to the original run.
What about historical events where critical decisions came down to the whim of one man? One of the famous ones being Hitler’s choice of when to attack, allegedly years before his advisors recommended. Maybe. One theory is that Hitler personally had a terminal illness, and so this man would always order an early attack. You could not convince him not to attack with rational evidence or arguments on the probability of winning, because he would personally be dead and unable to see the outcome if the war were delayed.
The existence of the terminal illness would be known to an omniscient observer, and so this would be a case where a seeming “black swan”—an irrational early attack—would be predictable.
WW2 is a specific case where we know that the material advantages the winning side had were enormous, we know the ultimate winners, but tiny perturbations could have changed the course of entire campaigns.
I wonder what the margin of infection was for patient 0 in Wuhan...
agreed on most points, but 1. we’re asking about 0.1% impact, and 2. optimizing processes can have multiple dynamical sinks, so the question is how often we cross decision boundaries in the aggregate dynamical system due to quantum randomness, and I expect the answer is often enough that it has more than 0.1% impact, because any time a chaotic system has the opportunity to make a difference in a person’s behavior, the chaotic system will depend on a significant portion of the bits of the quantum system, and in particular that means the weather over a long enough timescale is a quantum rng, since the mostly-newtonian dynamics have sensitive dependence on initial conditions, and so the slightest perturbation will accumulate. the question is how long it takes, I imagine folks who do work on weather simulation know that pretty well.
So then who is right would depend on whether 1 in 1000 decisions are close enough to the boundary a small disturbance will change it.
Once we develop neural implants this sounds like a testable hypothesis, you could actually manipulate someones brain, changing single neuron activations, and see how often this matters on some repeated decision task.
I think the main point is there is still a structure here, human agents resist disturbances. Atoms of gas don’t, any change above any granularity limits from physics will change the behavior at a collision.
for it to influence an election, 1 in 1k people need to have their election decision change, which depends on a lot more than one decision being locally influenced by quantum randomness. I still think that, pending whatever oracle resolves this question, the biggest impact path is going to look like weather ⇒ economic fortunes of those who are near their political policy decision boundary. It’s possible there are other ways for their fortunes to change from chaos, and the “maybe candidate death depends on quantum randomness” take might be it. But I still think the main thing that is chaotic enough to have a significant impact is the weather.
Correct. I also had a bit to think and are you aware of ternary logic for radiation resistance? The idea is, every circuit and memory cell in an IC has 3 parallel gates, and 3 parallel memory cells for every binary decision or memory read.
Frequently in the chip there are majority gates, where the majority input determines the output.
What this does is random disturbances from radiation must disturb 2 inputs during the same clock cycle, or the output will be the same.
If “quantum fluctuations” are kind of like radiation, and some human synapses are acting like majority gates, then for most decisions they will have no effect at all. In chip designs with n-way redundancy using the above, choosing n, it is possible to design a chip that will ignore radiation to a target pError that can be much smaller than 1 in 10,000. It can be 1 in billions of bit flip events or more cause an output bit to change.
I do not know if human neurology is this robust. I kinda suspect it might not be, but for a “deep belief” like politics, it could be that quantum fluctuations don’t flip a single vote across the US electorate.
I agree that it would only flip votes if those votes were the person basically not giving a crap. again, I only expect significant impact from upstream effects like a different distribution of extreme weather events resulting in a different economic outcome for enough people to matter.
Wouldn’t this random distribution of quantum events (gaussian?) flip an equal number of marginal voters in each direction? Meaning if it changes 1 in 1000 voters, or 220,000 people, I think the law of large numbers kicks in here. It would cause approximately as many (A->B) transitions as (B->A)
The totals would be within a few votes, but smaller counties might have larger shifts because their set sizes are small.
on average across worlds, yes; or if the distribution in your weather simulator says most expected weather events in that timeframe are small, yes. but for a given timeline, extreme weather events would be expected to be biased in one direction in terms of which areas were impacted by which events. if you change the random seed for the weather 5 times, then based on my current knowledge and in the current time period I expect you’ll get at least 5 different natural disasters, which are each region-correlated in their impact, and so can have a significant bias in which votes they flip on an economic recovery basis. this is faul_sname’s argument for election day weather events, but generalized to any weather event extreme enough to impact economic fortunes. if the weather events are small enough to impact people’s economic fortunes individually, then yeah, the expected impact goes down due to law of large numbers, but I also don’t expect small weather events to significantly impact votes, due to the same argument you made about humans being optimizers. (election day weather also might be enough, but I expect a significant number of natural disasters before 2028.)
I strongly feel Habryka is right here. Things are not that contingent. In particular, the invocation of chaos theory feels misleading here. The weather is chaotic on relevant timescales but most of our world and society is very much not.
Interested to hear different intuitions
The weather significantly impacts society, especially lately. Eg, to pick one kind of event, via wikipedia via IPCC report 2021:
this has practical effects for a variety of parts of the economic network, notably shipping and natural disasters, both of which affect prices in ways that can affect what policies people want, as well as anyone who’s directly impacted by a natural disaster.
Steve Byrnes argument seems convincing.
If there’s 10% chance that the election depends on an event which is 1% quantum-random (e.g. the weather) then the overall event is 0.1% random.
How far back do you think an omniscient-modulo-quantum agent could‘ve predicted the 2024 result?
2020? 2017? 1980?
Daniel Kahneman notes that if sperm are randomized, the chance of Hitler, Stalin, and Mao all being born boys is 1⁄8. Re-run the 20th century with any of them being female and you get vastly different results. That thought experiment makes me suspect our intuitions for the inevitability of history are faulty.
Perfect predictions of certain physical systems are downright impossible due to tipping points, ie chaos theory.