April 17, 2024
Artificial Intelligence (AI) Primer: Election Administration
The 2024 election cycle is the first in the boom era of generative artificial intelligence (AI). Officials are already preparing for new and potential challenges, such as deep fakes and more sophisticated malware. However, AI as a science was not created to interfere with elections. Rather, AI was developed to improve productivity, enhance automation, and give us new insights into how we work.
We ask: What are some ways in which election officials can responsibly leverage AI as a productivity tool to help run successful elections? And how do they do so while recognizing its limitations and appropriate guardrails for use as it relates to critical systems?
We propose a few simple steps you can take if you are considering adopting AI’s productivity tools in your own office. We also offer a few real world examples to demonstrate its potential.
Improve your own AI literacy
As AI finds its way into our work discussions and news feeds, many of us do not have a basic understanding of the topic. While election officials do not need to become experts, they should take some time to learn the basic concepts surrounding AI including the science behind AI, recent advancements, and ways in which consumers and businesses are already using it. A primer:
- AI is the science of teaching computers to make decisions and/or generate content in a humanlike manner.
- AI “training models” take in data – images, text, code, sounds, files – and use statistics and algorithms to learn about those inputs to make decisions, provide insights, and even generate new content in real time.
- A recent surge in consumer interest and engineering advancements have us witnessing exponential growth in AI’s use throughout our economy.
- Consumers and businesses are using AI to run advanced analytics, deploy self-driving cars, create new media, and interact with human-like personal assistants.
Develop standards around the use of AI productivity tools in your office
Below are some standards you can apply in your own office for responsible use of AI productivity tools:
- Consult with your IT experts, office of legal counsel and state authority to address any concerns they may have.
- Never upload confidential, private or proprietary information to an AI platform.
- Never deploy an unauthorized or third-party AI application to a secure system.
- Always proof and validate final work products generated or supported by AI. AI may be appropriate for drafting or a “first pass” but should not be used as a final product.
- Audit, test, and quality control any custom tools deployed in your office or to the public.
- Be transparent about your use of AI.
To learn more about how technology companies are approaching the use of AI in elections, read “A Tech Accord to Combat Deceptive Use of AI in 2024 Elections.”
Explore how AI’s productivity tools can support your office
There is a wealth of AI productivity tools available to businesses and consumers. Some are bespoke and some are already embedded in tools you use. Elections administration involves many rote tasks that are data- and text-driven, and that require precision, replication and analysis. AI may be a good fit for many of these types of tasks.
Communications
You can use AI productivity tools to summarize long PDFs (even multiple PDFs), take a first pass at drafting a research assignment, and create social media content. For example:
- You can prompt natural language processing tools to draft an outline and generate key points for a presentation on a topic of interest. They can even take a first pass at moving your own documents into a presentation format.
Budgeting and Procurement
AI excels at analyzing and summarizing text- and data-heavy files, making it a great candidate to analyze your budget and procurement materials. For example:
- If you receive multiple bids to an RFP, you can upload those bids as PDFs and ask for a summary and comparison. You can use that as a framework or guide as you begin your own thorough assessment and scoring of those same bids.
Translation
AI excels at translating between languages. In fact, many of the translation apps you are familiar with are likely driven by AI technology. For example:
- Use AI to make a first pass translation for public facing materials. You can then use a certified, official translator to verify the accuracy of the work product. Note that you should always defer to official translations for critical materials such as ballot content.
Ballot Proofing
Ballot proofing can be supported by AI. While leadership is ultimately responsible for accurate ballot content, AI can serve as another proofing team. For example:
- Upload PDF versions of sample ballots along with the official source material, such as an official spreadsheet of the ballot content. Prompt the AI tool to compare and analyze the files for typos, omissions, and even candidates listed in the wrong order or under the wrong contest. This can be an effective layer to add to your existing proofing plan.
Analytics
Everyone benefits when officials order enough ballots, have appropriate staffing levels, and detect outliers in our datasets. AI productivity tools are perfect for supporting your analytics. You can use plain language prompts to generate charts and models that previously were time consuming efforts. For example:
- You can upload a voter turnout report to an AI-driven analytic tool. Be sure to remove any confidential or personally identifiable information. These tools can quickly produce charts and graphs, forecast turnout, and identify variations in your data to better forecast turnout and plan for the upcoming election.
Your responsibility
Productivity tools are meant to support your work and not to make critical decisions for you. Election officials are ultimately accountable for administering successful elections. Always explore new technology responsibly, and defer to your existing best practices and authorities when in question.
Stay informed on emerging security concerns
Unfortunately, these same types of productivity tools have the potential to disrupt your operations when used by malicious actors. Deep fake videos can spread misinformation quickly. Synthetic audio might be used in phishing attempts. New types of malware – and advanced deployment of malware – may emerge.
Consider reviewing the Department of Homeland Security’s Risk in Focus on generative AI and elections to build awareness around potential threats and how to mitigate. Another resource produced by The Elections Group, Avoiding F(AI)kes: Practices for Verifying Communications With People You Trust, is a helpful resource for election officials and their teams to help identify potential AI fakes, misinformation, and synthetic media.