On Wednesday, while writing a story on gerrymandering, I asked ChatGPT, “Which states have congressional delegation partisan breakdowns most divergent from the presidential vote in 2024?”
The artificial intelligence interlocutor then named Florida, Texas, North Carolina, and Ohio as the biggest offenders, all of which are controlled by Republicans and have House delegations that skew toward Republicans.
However, having already done my own research, I knew ChatGPT left out some pretty big Democratic Party offenders. So, I asked, “Um, how about California?”
At this point, ChatGPT admitted that California’s House delegation is far more out of whack compared to the popular vote than any of the Republican states it mentioned. So I followed up, “Are there any other Democratic states you are leaving out?”
ChatGPT responded, “You’re absolutely right to ask — in addition to California, several other Democratic‑leaning states also show significant delegation imbalances that favor Democrats.”
It then named Nevada, New York, New Jersey, and Illinois as the big gerrymandering offenders. So, I asked ChatGPT, “Why did you ignore all these Democratic states that are way worse than Republican ones?”
ChatGPT responded, “That’s a very fair criticism. The short answer: I focused first on the Republican‑leaning states because that’s where most media and academic commentary frames the gerrymandering debate.”
FOUR OF THE TOP 5 MOST GERRYMANDERED STATES ARE CONTROLLED BY DEMOCRATS
In other words, if the media and the academy are biased against Republicans, ChatGPT will be, too. This seems like a pretty big problem for AI programmers and really for all of us. Due to partisan bias, the media and the academy are completely blind to reality on many subjects. Instead of being a tool to disrupt or correct this bias, AI is reinforcing it instead.
This is something to keep in mind the next time someone tries to use an AI chatbot as an authoritative source on anything.