Artificial Intelligence, Misinformation, and Elections

Artificial Intelligence chatbots like ChatGPT are popular sources of information; according to PEW Research “about four-in-ten U.S. adults say they use chatbots for information searching.” Google is getting ahead of the trend, and is making more of their AI chatbot features available right in the search bar of Google Search via AI Mode and the AI Overview.

While AI chatbots can make difficult research easier, there are potential drawbacks to using them. It is especially important to know about these drawbacks to using chatbots with the midterm elections coming up. For example: it can be very difficult to research about local candidates, and according to NPR, people are turning to chatbots to find information about them.

Using AI Chatbots for Research

There are two important things to remember if you choose to use AI to look up issues and candidates.

1. If AI cannot find an answer, it might make something up.
2. AI can remember your previous questions, and might only show you the results it thinks you will like. For example, if you consistently search for either republican, democratic, or independent candidates, future answers it gives you will skew towards the same viewpoints.

The best tip to deal with these shortcomings is to add a rule to your prompt. Try adding “Always show me your sources” to the end of your question. You can also try adding, “Tell me when you can’t find an answer.” Neither will remove the bias or hallucinations 100% of the time, but you can then click the links of sources the AI response gives you to double-check the results.

AI Generated Content about Candidates

AI also makes it easier than ever for people who want to push a certain agenda to create fake videos and images that misrepresent candidates. Unfortunately, there’s no longer a sure way to tell if an image or video is AI generated. Here are some signs that might be present if an image is AI generated.

1. Garbled text: if any written words are nonsense or unintelligible, it could be AI.
2. Overly flawless people: if it looks like an Instagram filter has smoothed out all flaws, it could be AI.
3. Cluttered images: if there’s a lot going on—images full of vague background lights and swooshes, or several graphics squashed in one image—it could be AI.

However, the AI image generators keep getting better. You may never know if an image is AI generated or not, but what matters is how you respond to the image. If you see a post villainizing or hero-worshipping a candidate, check to see who posted the image. The poster might be biased for or against the candidate. If an image makes you feel anger or disgust about a candidate, try a regular internet search like [candidate’s name + topic] or [candidate’s name + scandal] to check if they really did what the post claims.

Local Election Resources

GPL has created a blog of voter resources for the election. Check it out for links to reliable information about local candidates. Find the blog here.

The League of Women Voters will also be in the Lobby on Saturday, October 17, 10 AM–2 PM, Monday, October 26, 5:30 PM–7:30 PM to offer help registering to vote and finding answers to your questions about the November election. Find more information here: The League in the Lobby: Voter Information Events.