Correlation, causation, or rando incoherence? 🤔

My experience with AI deployed on a "science" based research service...and how to do better.

My inquiry was straight forward - a time-series dataset of consumer perception trends for the three largest U.S. wireless carriers - AT&T, T-Mobile, and Verizon - on a commercially available online consumer research service.

I quickly and easily found the data I was looking for through the service’s intuitive and easy-to-use AI-driven chatbot. After asking my question, it described how it would break down the problem and structure its internal data queries. The chatbot returned cited results with guided discovery buttons to facilitate a deeper exploration.

I was impressed...until alas...it capped off the exchange with this (Verizon example):
👇👇👇

The first two bullets are nonsensical. The ones that follow are worse.

It never occurred to me, or perhaps anybody, to consider whether a wireless carrier's popularity might influence consumer preferences for Cobb or Caesar salad...and certainly not over time. And then...creepy mannequins! Oh my!

I struggle to imagine what this product's designer…