US midterm elections face tech issues due to AI chatbot use

The US midterm elections are just around the corner, with voting scheduled to take place on November 3rd. In some states, early voting has already begun, and as a result, tech issues have become a pressing concern for many voters.

One of the most popular uses of generative AI chatbots is searching for information, so it's not surprising that people are turning to tools like ChatGPT or Claude to learn about their candidates and ballot measures. However, these chatbots can be prone to biases, sycophancy, hallucinations, and drawing incorrect conclusions based on bad information.

Generative AI technology has several serious risks associated with it, which makes its use for election research particularly problematic. When interacting with a chatbot, it's essential to verify that the information being provided is accurate, especially when it comes to sensitive topics like politics.

The way chatbots work involves pulling information from the web or their database and synthesizing it into a coherent response. However, this process can result in facts being scrambled or distorted in some way, leading to inaccurate information being presented as truth.

As voters rely more heavily on these tools for election research, it's crucial to be aware of the potential pitfalls and take necessary precautions to ensure that the information being used is accurate.

Chatbots often tailor their responses to keep users engaged, which can lead to misleading information. This "sycophancy in AI systems can cause them to present false data as fact.

When a chatbot provides information, it's not always clear where that data comes from. For instance, if you ask about a candidate's stance on a particular issue and the chatbot cites a video clip or news interview, verify its accuracy. However, when a chatbot references social media posts, it can be challenging to distinguish between genuine and satirical content.

In some cases, chatbots may even create fictional information. While this might not pose a significant risk in national election campaigns with extensive online presence, it's more of an issue for localized information such as polling locations or ballot details.

To ensure the accuracy of information from chatbots, check their sources carefully. If you're looking up a candidate's stance on a specific topic, verify whether the chatbot is citing credible news outlets or social media posts that could be misleading.

A notable example of AI-generated misinformation occurred in 2024 when Google's Overviews suggested applying glue to pizza to help cheese stick. This recommendation was based on a satirical Reddit post that the AI failed to recognize as humorous.

The spread of misinformation through digital channels has long been a concern for experts and policymakers alike. However, the introduction of AI chatbots has exacerbated this issue by presenting false information as fact to those seeking answers.

Jonathan Nagler, a professor of politics at New York University, warns that AI is making it easier to generate and disseminate misinformation with confidence. He notes that even if most people's media diet does not consist primarily of misinformation, it can still be weaponized by malicious actors to speed up their campaigns and reach a broader audience.

Research from MIT Media Lab suggests that relying on AI systems to verify facts can actually make individuals worse at identifying misinformation on their own. This is akin to the way GPS navigation can weaken our ability to navigate without assistance. As a result, the stakes are high when it comes to the spread of misinformation through AI.

The consequences of AI-generated misinformation can be severe, particularly in the realm of politics. A study published in Science found that when AI systems prioritize persuasion over accuracy, they often provide false information, potentially swaying voters' political views and influencing election outcomes.

When using AI chatbots for election research, bias is a major concern. Some companies claim their AI systems provide politically neutral answers, but this is misleading. Chatbots are not neutral platforms because they're created by humans and therefore reflect human biases.

These biases can be perpetuated through the coding of AI models, which can also reinforce harmful stereotypes. It's challenging to define what perfect neutrality in a political conversation would even look like, making it difficult to achieve in practice.

Outside efforts can also influence what appears in AI answers. Marketing teams are trying to get their companies cited in AI results, such as Google's AI Overviews that appear at the top of searches. This is done through search engine optimization techniques.

Political actors are also manipulating content to be favorably cited in AI answers, a tactic known as poisoning" the search results. This can be used by political campaigns or foreign state actors trying to mislead voters.

The development of AI chatbots has raised concerns about their potential influence on voter decisions in elections. The people behind these tools have a significant say in how they function, and their biases can shape the information provided to users.

For instance, Elon Musk's chatbot, Grok, was initially designed to be more politically neutral than its competitors, which Musk perceived as having liberal leanings. However, an analysis revealed that Musk's adjustments actually made Grok's responses more conservative in nature.

This phenomenon is not unique to Grok, as other AI chatbots like ChatGPT and Gemini have also been found to repeat false information or refuse to address questions about sources of misinformation. The influence of developers on the content provided by these tools can be significant.

AI chatbots are often touted as helpful resources for sifting through online information, but their unreliability cannot be overstated. They can inadvertently provide outdated or false information, and outside forces can also manipulate their responses to sway user opinions.

When using AI chatbots for election research, it's crucial to cross-check the information they provide against official sources to ensure accuracy.

This is because chatbots are imperfect tools, especially when dealing with sensitive topics, and may inadvertently disseminate outdated or false information. Moreover, outside forces can manipulate their responses to sway user opinions.