Introduction

Social Media has been in the news lately with Elon Musk purchasing Twitter and taking it private. During the 2016 election, Twitter devolved into a crazy people echo chamber. When Twitter finally ejected Trump with a lifetime ban, things improved as most of the crazy people left in protest. Now with Elon in charge, although the policies are still the same, it appears all the crazy people are returning and testing every possible boundary.

I previously blogged on how repeated exposure to fraudulent information conditions us into believing it along with how Social Media divides us. This article continues in that theme, adding in elements for bad science, even to the degree of killing off supporters to reach an election goal.

The Creation of Echo Chambers

Both Facebook and Twitter provide news information, but they use algorithms to try and give you the news you want. Their algorithms look at the articles you click on, what your friends click on along with incorporating all the information they collect on you. As a result if you are a Democrat, you tend to get news articles that are friendly to Democrats, similarly if all your friends are Democrats, then everyone posts things in support of each other and all you ever hear is the single side of every issue. This places everyone in a nice comfy silo, but often results in shock when confronted with the real world, like via election results.

The Death of Sampling Theory

Back, long ago, when I attended the University of Victoria, I took a third year Statistics course on Sampling Theory. This is a rigorous science on how to poll a subset of a population and calculate the probability that the results of the sample can be generalized to the whole population. I greatly enjoyed this course and understand that if done correctly, proper polling of a subset of the population can give extremely accurate predictions. There is a lot of sophisticated Math and Statistics behind Sampling Theory along with strict methodologies on how to carry out the sampling. None of this methodology seems to be in use today by the current news organizations.

If you’ve read Isaac Asimov’s Foundation series, or seen the recent Apple TV series, then the fictional character Harry Seldon’s invented Psychohistory, comes across as Sampling Theory on steroids. I think Asimov foresaw Statistics and Sampling Theory becoming more powerful and influential, but he didn’t foresee Social Media and its rather negative effects.

Given the power of sampling theory, why have all the election polls become so inaccurate in the past few elections? Why was everyone predicting a Republican wave that was going to wipe out the Democrats? Unfortunately, the people conducting polls have become extremely lazy. Let’s consider a Twitter poll on who will run a particular senate race. A few points:

• When you post a Twitter poll, it is primarily going to be seen by your own followers and if you pay to promote it, the algorithms will promote it to people similar to yourself. This isn’t a representative subset of the population.
• There is no control over who responds to the poll. Anyone can respond to it whether they vote in that state or even if they reside in the US.
• There are groups of people actively trying to influence the election, whether foreign governments, activist groups or corporate lobbyists. These groups either employ or run “bot” farms that are directed to flood such polls with fraudulent responses.
• The people that use Twitter do not make up a representative subset of the population. Many age and ethnic groups are not represented at all. How will people that don’t use social media vote?

Professionals doing sampling theory correctly need to be careful that the subset being polled has the same makeup as the general population. This includes by sex, age, ethnic background, sexual orientation and dozens of other factors. They spend a lot of time making sure that everyone is represented, which includes gathering the results by multiple means such as online, phone, knocking on doors or even canvassing the homeless.

When conducting these polls you can calculate the margin of error, but you never see this reported. The smaller the sample size, the larger the probable error. The less representative your sample, the larger the error.

One example, of many, is the following poll suggesting Dr. Oz would overwhelmingly win the Pennsylvania senate race. Actually, he ended up losing to Fetterman.

What’s the sample size of this poll? How were responses validated? How representative of the population is the sample? What methodologies were used? Was a margin of error even calculated?

COVID and the Death of Republicans

For the past few years, we’ve spent part of the winter down in Yuma, Arizona. I would attend the local Amateur Radio club breakfasts early Wednesday and Saturday mornings. Things became so polarized that they tended to need separate Democrat and Republican tables. The sad thing I observed when we spent March down there was that all the Republican members were unvaccinated and believed if they got COVID, then they would use the Ivermectin or hydroxychloroquine tablets they had bought over the internet as recommended by FOX news and Dr. Oz.

This led to a situation where all the hospitals were full and you could only get an ambulance if you were having a serious heart attack. This has caused quite a shift in the makeup of the population down there. Before the pandemic, it was nearly all elderly Republicans. But with so many either dying unnecessarily of COVID, or complications of COVID with another condition, the demographics have shifted noticeably with the neighborhoods having more young families, often of an ethnic background.

With razor thin margins between the two parties, I think this made a difference where the house riding containing Yuma is now Democrat. Not sure if it was the smartest idea to promote an anti-vax agenda that resulted in hurting and even killing their base, namely the elderly.

Summary

Polling is a science. When done correctly, it is a great tool. When done poorly, the results are garbage. Modern news organizations have become lazy and most of the polls we see today are terrible and don’t represent the general population at all. There are still people doing correct polling, but these are generally ignored as they don’t provide the biased results being looked for, or take too long to conduct.

I’m not sure if there is an answer here, but social media certainly seems to be part of the problem. It would be nice to go back to unbiased news organizations that aren’t owned by billionaires pushing their own agendas. But with the death of print media, until someone figures out a way to monetize good reporting, it isn’t going to happen.