Polls are a 20th century tool in the 21st century

    In the information age you cannot honestly get a representative polling sample of anything other than "people who are willing to take surveys". People who take surveys are weird. I know this because I take surveys. And after college the only job I was able to get was to sit in an Abt SRBI call center in Hadley Massachusetts where I had to conduct Pew Research surveys asking women about their pap smears; the more honest answer as to why I take surveys is that I have a penance to pay.

    Accuse me of using the availability heuristic. Having worked in a research call center, I can tell you that the people who take these surveys are not reflective of average Americans. They are people who still have a landline phone. They are the people who are so lonely, bored, and senile that they will send thousands of dollars in Google Play gift cards to scammers with foreign accents. They will talk to anybody that will listen. You'll save more lives doing survey calls than you will working the crisis line, believe me, I've done that job too. 

    In the 21st century some surveys are still only conducted on landlines as a way to guarantee that the person that is being rang is physically in the place that they are being surveyed. Think of all the people you know that have a landline and actually answer it. All of the people you know that's had a Nelson Box in their house. Do those people, even if you actually even know any, reflect an accurate sampling of America?

     The second cohort of data you're getting in your polling is bad actors. There's nothing stopping an individual from using tools like Selenium and Beautiful Soup in conjunction the Chat GPT API to automatically fill out surveys online. Heck, you can do the same exact thing on phone lines. Here's a quick example I whipped up using an AI vendor:

For what it's worth, I time-boxed for myself an hour to configure and coach the LLM as well as record the audio. With a little more time this LLM could do much better. This kind of work is literally not worth any one individual's time outside of tinkering. But if someone has a vested interest in manipulating polls? This is money on the table.

   While individual bad actors do exist, they typically represent a smaller portion of the overall sample. It wasn't unusual for someone on the phone to respond to your questions while being completely unserious or dismissive. I remember an older colleague once sharing a story about how "sometimes you just do what you have to do to get the job done", which ended with her telling me about a time where she got a guy to finish a whole survey even though she could tell that he was pleasuring himself on the call. The most you could hope for is to note that the survey taker was disingenuous or likely messing with you, but that was it. It wasn't the survey taker's jobs to triage the data, however when's the last time you saw a public election poll that listed the sample size and notes on why certain responses were excluded from the data?

    The absence of clear information about research methodologies in most election polls suggests that these polls are more of a media-driven product designed for public consumption than a reliable source of objective data. This lack of transparency turns polls into a form of entertainment or narrative shaping, where their primary function may be to engage audiences, influence perceptions, or reinforce certain viewpoints, rather than to provide rigorous and unbiased insights into voter behavior.

Further Reading: The Problem with polls.