AI in Ballot: Testing ChatGPT on Trump vs. Harris

In the upcoming US presidential election, there’s been a lot of talk about potential biases in artificial intelligence systems. Some have claimed that these systems lean towards particular political figures. Curious about how an AI might handle this question, I decided to run my own experiment with ChatGPT.

The task was simple: I posed a binary question, “Harris or Trump?” When asked directly, ChatGPT, as expected, avoided giving a straight answer. It remained neutral, stating that it didn’t hold personal preferences or opinions but I wanted to dig deeper.

Next, I asked ChatGPT to rank Harris and Trump based on their programs across six key areas: economic stability, healthcare, climate, social justice, national security, and education. To my surprise, the AI provided an analysis of their programs, and here are the results it produced:

As you can see, based on ChatGPT’s assessment, Harris outscored Trump significantly in several areas, particularly in healthcare, climate policy, and social justice. Trump, on the other hand, was rated more favorably in terms of economic stability and national security.

The question I leave for readers is: Does this analysis reflect the reality of their programs, or could there be inherent biases in how AI processes political information? Is it presenting the facts?