Some I know might be involved even. And, you know, I’m not wholly sure:
After Mr Trump swept to a convincing victory in the US presidential election despite many pollsters having the race on knife edge, Mr Warner said: “The last month has shown us we can catch 20-storey rockets out of the sky, but we still can’t predict elections.
“The reason is simple: while physical sciences have had centuries to mature, behavioural science is just beginning its journey.”
In a LinkedIn post, he added: “Imagine the potential when we finally crack the code of human behaviour. This is exactly what we’re building at Electric Twin.”
I think the idea runs into an at least potential problem.
If physics says this is going to happen then this is what happens.
Humans not so much. There’s that difference between expressed and revealed preferences for example. Now, true, with voting that’s not entirely a killer – because voting is express, as is an opinion poll.
But even then we’ve still a problem. With physics we’re working out what happens in what conditions. With voting polling we’re working out which choice is to be made. Which is different….phyics doesn’t have choices. At least not at the macro level it doesn’t. Polling is, to stretch the analogy waaaay too far, more like quantum. Where we trying to determine the state – which state among different possibilities – at any one time. And that’s not something we can do with accuracy at the individual level now, is it?
Chaotic systems. And, people are not bound to make decisions by sure and certain phenomena nor laws, they can do anything for any reason, or no reason at all. And they change their mind when they see what others do or think or appear to think…
There is a bit physics in all of this. The results change because of observation. People might give dishonest answers or just the first thing that comes into their heads.
Pollsters can mess this up in their methodology: asking the wrong questions, asking the wrong people ( not getting a properly representative subset ) or not asking enough people.
Another, more important problem is the pollers remove Don’t Knows from their data, who are the people whose opinions we really need to know, because they might decide an election.
I’m not sure if it’s true of course, but Harris’s campaign had all their internal polls showing her losing, while public polls kept showing it as close or with her ahead. Perhaps most of the polling companies were trying to drag her over the line?
Someone at Google said something similar. The idea was that Google can work out how people think so what was the point of running an election. It would be cheaper just to let Google work out how people would ahve voted and let Google declare the winners of the election.
Anything that purports to be able to predict human behaviour on the basis of logic or rationality is snake oil. As but one example, we bombed Serbia for 78 days in 1999 and still have no idea why Milosevic held his hands up on day 79; more to the point, neither did he!
Widespread knowledge of physics changes human behavior but not the physics. However public knowledge of predicted human behavior would modify that human behavior to the point that the knowledge was wrong.
Also, informed perversity. If the subjects of the monitoring know what is being predicted by the monitoring, they change their behavior. As Hari Seldon taught us, the only way to predict human behavior is with huge numbers of humans in aggreate, and in secret.
Ah, andyf thinks the same way I do. Worryingly – at the same time! 😀
About as predictable as long range climate.
Negative rainfall and so on!
“It would be cheaper just to let Google work out how people would ahve voted and let Google declare the winners of the election.”
Excellent idea, what could possibly go wrong?
A lot of confidence in Physics there, when what we call physics is in fact physical observations taken and interpreted by the same fallible perverse people as the polls.
Clmate change is an example. When you look into the facts, the observations and the conclusions, you find that predictions are not consistent, they depend on the opinions and biases of people.
RK
Climate science is a perfect example.
Chaotic system
Ask the wrong question
Use the wrong dataset
Not have a large enough sample
Ignore the Dont Knows
Predetermined answer
It might be well to listen to the ‘Freakonomics’ podcast (recently revised and updated) on academic fraud and misrepresentation in the ‘behavioural sciences’. It’s not hard to form the opinion that much of what is researched and published in that field is (at best) fuzzy and lacking in any sort of rigour, never mind repeatability. Abigail Shrier on the extent and weaknesses of ‘therapy culture’ is very good in this area too.
Electoral polling especially is subject to inherent and confirmation biases, and especially so when the electorate is very-sharply divided, as in the late US election. From what I read, the serious pollsters on both sides had very little doubt of the outcome, while the partisan and ideologically-driven polls all managed to give themselves a wonderful warm and fuzzy feeling for about 6 months (Harris leads nationwide by 6 points and even more in all 7 swing states!) at the unfortunate expense of being completely and utterly wrong. And yet, these same shamans of the statistics will be duly trotted out next time, to give the same hopeful predictions. It’s almost as if they serve some different political purpose than actual prediction.
llater,
llamas
The ‘polls’ were intentionally fraudulent, designed to dishearten Trump voters and persuade waverers to back the winning side.
The bookies told a rather different story.
“If physics says this is going to happen then this is what happens.
Humans not so much. ”
Thats a bit of a problem for economics then isn’t it, given its the study of human economic activity?
Indeed: which is exactly where the expressed and revealed preferences idea comes from. At least econopmics – well, neoliberal such – is insistent upon what people do, not what they say.
As others have already said, the aim wasn’t to try to predict the outcome, rather to keep a hopeless candidate in the race for as long as possible in the hope that something might turn up to win it for her.
Still, I like that $1.5 billion was spunked by wokists trying to polish a turd (and failing) 😎
There is a tendency for people to give the answer that they think the questioner wants to save the stress of an unwanted argument so mainstream political parties habitually apply a discount to their canvass returns when privately (i.e. not to the MSM) forecasting results. The more the MSM demonise Trump, Reform, AfD, RN, etc the more individuals will assume that the pollsters want them to say “Harris” or “Labour” or “SPD” or “Macron” so the polls will be not only wrong but wrong to a greater extent.
All the fancy electronic technology built into Ben Warner’s new venture isn’t worth a can (let alone hill) of beans compared to some guy exercising *good* judgement on the right discount to apply to the crude numbers from the pollsters return. [Nadhim Zahawi’s “yougov” achieved a significant improvement in accuracy by seeming to be an impersonal machine with no preference for any particular answer so a much smaller discount needed to be applied but it still threw up a small bias in favour of the views of the MSM]
In a LinkedIn post, he added: “Imagine the potential when we finally crack the code of human behaviour. This is exactly what we’re building at Electric Twin.”
And on LinkedIn as well, so you know it’s a sure-fire money maker.
LinkedIn is the modern Council of the Wise, where you can learn such startling things as “What The Rapid Takeover of Syria by ISIS taught me about B2B Sales”
As pointed out above: So long as pollsters leave “none of your fvcking business” out of the results as “irrelevant to the question”, polls will always be wrong.
You randomly poll 1000 people.
200 actually reply.
100 of those replies are “do not contact me with your polls again”
50 will be on board with any idea pushed forward in the poll.
50 will be “don’t know” /against.
Poll result: A majority of people want [X].
Message never published: 90% of people don’t even care about [X], and hate unsolicited rude questions about their private opinions.
Physics is repeatable. By repeating the experiment we can measure gravity, temperature, whatever to ever greater accuracy.
Behaviour or psychology is not repeatable (often / usually). Less than half of psychology tests show the same result. And some aren’t repeatable at all. Try doing the Milgram experiment today, even if you were allowed to. (Actually there’s some evidence that some of the test subjects worked out that it was a fraud from the get go.)
Silicon valley types have come up with what they think is a new theory of human behaviour. The “preference cascade”. This is just stability until there’s a “tipping point” where opinions change in an instant. Examples:Trump or Starmer.
When might the preference cascade happen to force policy change on immigration or glabal worming? Search me, gov.
On the $1.5 billion, or whatever the final number turns out to be – maybe not so much to prop up a failing candidate – you need lots of money, sure, but not that much – as to take the unique circumstance of Presidential election finance rules to a) pay off past debts b) pay for future activities c) do some discreet money-laundering. After all, after January 20th, the Biden crime syndicate will be out of business – having nothing to sell – and they need a suitable slush fund to maintain their extended family. And lots and lots and lots of money can be easily and freely paid to ‘consultants’ and ‘production companies’ and ‘analysts’ and ‘security providers’ and ‘web services agencies’ and a hundred other ill-defined rat-holes. Dig a little deeper – as the FEC reports will eventually do – and the network of ‘friends and family’ making bank off ‘servicing’ the campaign will be – intetesting.
However, I agree as others have noted, that the big public spends highlight the amazing mis-direction of the campaign. They spent money as though they were expecting to win if they could just attract more upper-middle-class women voters. Oprah Winfrey and Taylor Swift? Puh-leeze. Working-class families don’t give a single solitary toss what these amazingly-privileged women think, and in fact are actively put off at being lectured to by them. The whole campaign operated in a parallel universe, which may be why they were so amazed at the real-universe outcomes.
llater,
llamas
There’s real money to be made on betting markets in being right. And yet with that incentive, polling companies still get tight elections and referendums wrong. All that effort with online polling, face-to-face, focus groups, new phrasing, scraping data – the obvious conclusion is that what Hayek said about economic planning could also apply to human voting behaviours. There’s always a few % that is unknowable.
“Silicon valley types have come up with what they think is a new theory of human behaviour. The “preference cascade”. This is just stability until there’s a “tipping point” where opinions change in an instant.”
Well duh. Its not a new idea. Lots of things are obvious, but everybody ignores the facts until there’s a fracture in the consensus for some random reason, then because human beings are social animals rather than rational beings they all pile in to the new paradigm, whatever that might be, regardless of facts. See the current issue over Muhammed Fayed – it was obvious to anyone with eyes that he was a wrong ‘un, but no-one dared say so, until someone did, and now everyone piles in calling him the greatest sex pest since Genghis Khan.
See also the Post office scandal, the blood transfusion scandal etc etc (I note the involvement of the State being a common factor). Ditto the Great Financial Crash – I knew something was coming down the pipe as early as spring of 2007, thats when I sold my share portfolio, the Crash wasn’t exactly difficult to predict. But all the BSDs knew better, until they didn’t and all rushed for the exits at the same time.
Eventually there will be a time when everybody will have known Climate Change and Net Zero was nonsense, and never supported it one jot. And also a time when the covid vaccines will be seen as the massive human rights violation they undoubtedly are. When the switch will flip on those is unknown, but flip it will one day. All the facts are there right now for anyone desirous of the truth, yet 99.99% will ignore that until they see the other sheep disappearing through the hole in the fence, and will follow them, regardless of where they are going. Better to be in the crowd under any circumstances than be out of the crowd and be right.
Preference cascades are, imo, about social proof.
From mid-2015 onward the entire US and global MSM did its damnedest to portray Trump voters as a minority of racist simpletons who were fooled by Literally Hitler.
This was less like a political campaign and more like what mean girls do at high school – Trump was Othered in the biggest tsunami of criticism the world has ever seen. Considerably more words – in print, on broadcast media and on the internet – have been spent monstering the 45th president of the United States than were spent on criticising Adolf Hitler when he was alive. Every single talking head, blue checkmark and Important Person was marching in lockstep.
Anyway, the social proof points that tipped the balance towards Trump came in two phases imo:
* When Trump was shot, and came up swinging like a badass – this was priceless and unforgeable proof of the man’s courage. It couldn’t be ignored. All of a sudden it was safe to say you like the guy.
* When Elon Musk adopted the Trump campaign – say what you like about Musk, but that was brave and it also cranked open the Overton window a little bit more. If the richest man in the world is happy to share a stage with DRUMPH, he’s not the unpopular, crazy, criminal fringe character they portrayed. Musk made it safe for rich people and nerds to come out of the closet and back the Orange Man.
Someone’s been reading too much Asimov. Spoiler alert: Hari Seldon isn’t real.
Also, the true plot point is The Mule.
My honest feeling about polling is that it’s a bit like cost/benefit studies in companies. A senior manager has already decided we’re going to do XYZ, so you cook up some numbers to justify it. I remember getting a cost/benefit study for an insurance product and they’d included the premiums as income rather than the commission. So, about 4 times out. I called someone and they just said “look, shut up WB, it’s happening, OK”. And I did it. Not my money being wasted.
Like you can bet that Lucasfilm had all sorts of data about Girlboss Star Wars to give to Disney executives. But did Kathleen Kennedy scrupulously analyse the market, or did she just want to do Girlboss Star Wars, so get the ammunition together?
And I think a lot of this comes from the principle/agent problem. Small entrepreneurs don’t do a whole lot of research. Roger Corman made his movies with monsters, car chases and women with big knockers in bikinis and only once lost money on hundreds of movies and that was because he just wanted to make a serious film about racism (The Intruder). But Roger Corman was spending his own money.
Grikath,
“Message never published: 90% of people don’t even care about [X], and hate unsolicited rude questions about their private opinions.”
There’s also a big thing about degrees with polling which is missed. Like 49% of people in the UK support re-entering the EU compared to 35% who want to stay out. But clearly, they don’t care that much because those millions of people aren’t giving money to Mirror Universe Nigel Farage or walking the streets of Swindon to take us back in. The parties that have rejoining the EU as an expressly stated policies are all the little wanky ones like the LDs, greens and SNP (less than 25% of the vote).
Thing is, though, that objective commentators absolutely did predict the election. The YouTube channels I watched in the months leading up to the election were fair and thorough in their examination of the polls, and spot on with the state-by-state results. As I commented here on the live thread as the polls closed, the only question was whether Trump would also win the popular vote. So who is the “we” in “we can’t predict elections”?
“ The parties that have rejoining the EU as an expressly stated policies are all the little wanky ones like the LDs, greens and SNP (less than 25% of the vote).”
Wait until Starmer’s request the regulators comes back. As a minimum they will say we need to be in the EU Single Market but for maximum economic growth we need to rejoin. At that point Single Market becomes official Labour policy and they will act as if it’s rejoin.
Gödel’s law is roughly
Any system of coded representation that can make of complete statement of itself is , a incomplete system.
@ Chris Miller
Not yet
“Hari Seldon” may not have been real, though his creator, Dr Asimov, in a short story “Franchise” came up with the same idea as “Electric Twin” in 1955 – one person was selected to be “The Voter” and he was quizzed by “Multivac” (Asimov’s mega-computer) from which the computer calculated the winner of the election.