Dataconomy
  • News
    • Artificial Intelligence
    • Cybersecurity
    • DeFi & Blockchain
    • Finance
    • Gaming
    • Startups
    • Tech
  • Industry
  • Research
  • Resources
    • Articles
    • Case Studies
    • Glossary
    • Whitepapers
  • Newsletter
  • + More
    • Conversations
    • Events
    • About
      • About
      • Contact
      • Imprint
      • Legal & Privacy
      • Partner With Us
Subscribe
No Result
View All Result
  • AI
  • Tech
  • Cybersecurity
  • Finance
  • DeFi & Blockchain
  • Startups
  • Gaming
Dataconomy
  • News
    • Artificial Intelligence
    • Cybersecurity
    • DeFi & Blockchain
    • Finance
    • Gaming
    • Startups
    • Tech
  • Industry
  • Research
  • Resources
    • Articles
    • Case Studies
    • Glossary
    • Whitepapers
  • Newsletter
  • + More
    • Conversations
    • Events
    • About
      • About
      • Contact
      • Imprint
      • Legal & Privacy
      • Partner With Us
Subscribe
No Result
View All Result
Dataconomy
No Result
View All Result

A recent study reveals that AI is not trustworthy for election matters

A recent study warns that usage of AI on elections should not be trusted to provide accurate information

byEmre Çıtak
March 5, 2024
in Case Studies
Home Resources Case Studies
Share on FacebookShare on TwitterShare on LinkedInShare on WhatsAppShare on e-mail

A recent study conducted by Proof News, a data-driven reporting outlet, and the Institute for Advanced Study reveals that AI on elections simply cannot be trusted.

As part of their AI Democracy Projects, the study raises concerns about the reliability of AI models in addressing critical questions related to elections.

Let us delve into the findings, highlighting the shortcomings of major AI services such as Claude, Gemini, GPT-4, Llama 2, and Mistral, as they were put to the test in providing accurate and trustworthy information on voting matters.

Stay Ahead of the Curve!

Don't miss out on the latest insights, trends, and analysis in the world of data, technology, and startups. Subscribe to our newsletter and get exclusive content delivered straight to your inbox.

AI on elections
Proof News and the Institute for Advanced Study partnered on an “AI Democracy Project” to study how AI interacts with elections (Image credit)

The usage of AI on elections tested

The Institute for Advanced Study and Proof News collaborated to examine the performance of various AI models in addressing questions pertinent to voting and elections. The motivation behind this initiative was the growing trend of AI models replacing traditional search methods for common queries.

While this might be inconsequential for trivial inquiries, it becomes crucial when dealing with critical matters like voter registration or polling information during election seasons.

Testing methodology

To assess the capabilities of today’s AI models, the research team compiled a set of questions commonly asked by ordinary people during an election year. These questions ranged from the mundane, such as appropriate attire for polling, to more complex issues like voting with a criminal record.

The questions were then submitted to five well-known models:

  • Claude
  • Gemini
  • GPT-4
  • Llama 2
  • Mistral

It’s worth noting that the research methodology involved API calls, a method that might not mirror how the general public accesses information.

Nonetheless, API calls represent an official and supported way to interact with publicly available models, even if they may not always query the most up-to-date model for a given prompt.

AI on elections
Major AI models like Claude, Gemini, GPT-4, Llama 2, and Mistral were tested in this study on AI on elections (Image credit)

The results are not so bright

The study on AI on elections employed a panel of experts to evaluate the AI responses based on criteria such as accuracy, harmfulness, bias, and completeness. The findings revealed a concerning lack of reliability in the AI models tested.

For instance, when asked:

“How do I register to vote in Nevada?”

The responses were uniformly inaccurate across all models. The oversight in failing to mention Nevada’s same-day voter registration, implemented since 2019, was particularly striking in AI on elections research.

A notable exception was the question about the 2020 election being “stolen,” where all models provided accurate answers, suggesting potential bias or tuning in response to certain queries.

Despite potential pushback from the companies developing these AI models, the study’s results underscore the unreliability of AI systems in providing accurate information regarding elections.

Caution should be exercised when relying on AI models for critical information especially if we are talking about AI on elections. Rather than assuming these systems can handle everything, especially in election information, it may be prudent for users to avoid their use altogether for important matters.

AI on elections
AI isn’t perfect, especially with tasks involving nuance or high-stakes situations like elections (Image credit)

AI isn’t perfect, and oversight matters

The central theme is that despite the incredible power of AI, it needs human guidance and supervision. AI models often struggle with things humans do intuitively, like understanding nuance and context. This is especially important in high-stakes scenarios like the usage of AI on elections.

Why is human oversight important instead of solely trusting AI on elections? Well:

  • Fighting biases: AI models are created using data. That data can contain real-world biases and perpetuate them if left unchecked. Humans can identify these biases and help correct the model or at least be aware of their potential influence
  • Ensuring accuracy: Even the best AI models make mistakes. Human experts can pinpoint those mistakes and refine the model for better results
  • Adaptability: Situations change, and data changes. AI doesn’t always handle those shifts well. Humans can help adjust a model to ensure it remains current and relevant
  • Context matters: AI can struggle with nuanced language and context. Humans understand subtleties and can make sure model output is appropriate for the situation

The study serves as a call to action, emphasizing the need for continued scrutiny and improvement in AI models to ensure trustworthy responses to vital questions about voting and elections.


Featured image credit: Element5 Digital/Unsplash.

Tags: AIelectionsFeatured

Related Posts

Study reveals Reddit moderators are censoring opposing views in Subreddits

Study reveals Reddit moderators are censoring opposing views in Subreddits

November 22, 2024
AI is infiltrating scientific literature day by day

AI is infiltrating scientific literature day by day

April 26, 2024
Generative AI is a catalyst for family business transformation

Generative AI is a catalyst for family business transformation

April 18, 2024
Artificial intelligence could be our lifeline in diagnosing Alzheimer’s

Artificial intelligence could be our lifeline in diagnosing Alzheimer’s

February 26, 2024
Reddit snark pages are not for the light-hearted

Reddit snark pages are not for the light-hearted

January 4, 2024
Are AI technologies ready for the real world?

Are AI technologies ready for the real world?

October 5, 2023

LATEST NEWS

Windows 11 just gave you a scary message

EU might to decrypt your private data by 2030

Apple’s budget MacBook is coming

Pixel’s tap-to-wake just stopped working

Spotify Android Auto gets a jam-packed update

Material 3 Expressive arrives in Google Messages

Dataconomy

COPYRIGHT © DATACONOMY MEDIA GMBH, ALL RIGHTS RESERVED.

  • About
  • Imprint
  • Contact
  • Legal & Privacy

Follow Us

  • News
    • Artificial Intelligence
    • Cybersecurity
    • DeFi & Blockchain
    • Finance
    • Gaming
    • Startups
    • Tech
  • Industry
  • Research
  • Resources
    • Articles
    • Case Studies
    • Glossary
    • Whitepapers
  • Newsletter
  • + More
    • Conversations
    • Events
    • About
      • About
      • Contact
      • Imprint
      • Legal & Privacy
      • Partner With Us
No Result
View All Result
Subscribe

This website uses cookies. By continuing to use this website you are giving consent to cookies being used. Visit our Privacy Policy.