Artificial intelligence (AI) has emerged as a powerful tool for how we gather, interpret, and use data. For Minnesota Compass users—whether you are involved in grant writing, strategic planning, community organizing, or advocacy—integrating AI presents both opportunities and challenges.
Two important questions our users frequently ask about AI-generated data are:
-
How can I trust the data I get using AI?
-
How can using Minnesota Compass help me to verify that data?
Trusting AI-generated data
AI’s ability to process vast amounts of data can significantly enhance decision-making processes. However, ensuring the trustworthiness of this data is paramount. Here are some guidelines to help you assess the reliability of AI-generated data:
- Trustworthy AI should clearly cite the sources. Always check the origin of the results. AI-generated data should come from its use of reliable data from credible sources.
- Understand the basics of how the AI tool you are using works. Algorithms are simply a set of well-defined rules pre-designed to result in a specific sequence for solving a particular problem. Algorithms should be transparent, with easy-to-find documentation available to explain those rules along with any assumptions and limitations.
- Be aware of potential biases in AI systems. AI can inadvertently perpetuate existing biases present in the data it is trained on (the ‘P’ in ‘GPT’ stands for ‘pre-trained’). Look for AI tools that have mechanisms in place to detect and mitigate bias.
- Cross check AI-generated data with data from reliable sources. AI insights should only be one of several information and data resources you use. Ensure that the data you encounter using AI is up to date. AI tools relying on outdated data can produce inaccurate insights.
Using Minnesota Compass to verify AI-generated data
Minnesota Compass compiles data from highly credible sources such as the U.S. Census Bureau, the Bureau of Labor Statistics, and local agencies like the Minnesota Department of Employment and Economic Development (DEED). We take care to choose respected data sources, cite those sources, and be transparent about things like margins of error.
As a community-based social indicators project, our key measures are chosen by advisors and community input. This helps to address bias in our data. The tools and visualizations on Minnesota Compass are designed to make complex data accessible and understandable. Use these tools and visualizations to check the consistency and reliability of the data and insights you encounter via AI.
Minnesota Compass staff analyze and interpret data and are always available to help you assess the accuracy of AI-generated data. Our highly trained, experienced, multidisciplinary team of data experts can help supplement your analyses with robust methodologies and can help you identify potential discrepancies or areas that might need further study.
Users can rely on these foundational aspects of our project to confidently use Minnesota Compass to verify AI-generated results to data-driven, place-based, quality-of-life questions.
Practical guidance for verifying AI-generated data
To verify AI-generated data using Minnesota Compass, consider taking the following practical steps:
- Identify relevant indicators: Determine which indicators on Minnesota Compass align with the data or insights provided by the AI tool you are using. For instance, if an AI tool provides you with housing and economic characteristics for a community, compare these with the housing and economic characteristics indicators available on Minnesota Compass.
- Check for consistency: Look for consistency between the AI-generated data and the data from Minnesota Compass. Examine inconsistencies further to understand the source of any discrepancies. Check to see whether the data are for the same time period. Ask the AI tool to cite the sources for their data and compare them with sources cited by Minnesota Compass.
- Use Minnesota Compass reports: Leverage analyses from our Compass Points dashboard to gain context and background. Read relevant Insights articles published by Minnesota Compass and Wilder Research staff for additional depth and validation of AI-generated data.
- Engage with Minnesota Compass: Participate in forums, discussions, trainings, and other events hosted by Minnesota Compass where we often share best practices for data verification. Sign up for the Minnesota Compass newsletter to stay informed about the latest data trends.
Understanding common AI tools
NLP tools are used to process and analyze large amounts of natural language data. They combine language understanding with text generation. For instance, Google uses NLP to enhance search results, while Amazon's Alexa and Apple's Siri use it for speech recognition to respond to user queries. Chat assistants use it for frequently asked questions or referrals to service providers.
LLMs are designed to interpret and generate text by learning from vast amounts of language data in ways inspired by the human brain. Unlike general NLP tools that focus on specific tasks like sentiment analysis or translation, large language models like GPT-3 can handle a wide variety of language-related tasks based on training with enormous sets of data. They can generate sophisticated writing, answer questions, and even hold coherent conversations.
Computer vision tools involve training computers to interpret and understand visual information, such as images and videos. This technology enables facial recognition, like those used in security systems, and object detection, which help autonomous vehicles navigate by identifying obstacles. For example, social media platforms use computer vision to automatically tag people in photos. Medical imaging systems use computer vision to detect abnormalities in X-rays and MRIs.
Predictive analytics technology uses statistical methods and machine learning techniques to analyze historical data and make predictions. Predictive analytics are widely used in various industries to forecast trends, identify risks, and optimize operations. For example, banks use predictive analytics to detect fraud, while retailers use it to manage inventory and personalize marketing campaigns. Healthcare providers use predictive models to anticipate patient outcomes and improve treatment plans.
These tools analyze consumer behaviors, tastes, and preferences to suggest products, services, or content. People encounter recommendation systems regularly on platforms like Netflix, which recommends movies and TV shows based on viewing history, or Amazon, which suggests products based on past purchases. Social media sites like Facebook and Instagram use recommendation systems to curate posts and advertisements tailored to the interests of individual users.
What works best for you?
Now it’s your turn to test out AI programs for yourself. Try out the following prompts, and look up the data on Minnesota Compass as well. How do the data compare? Does the AI tool cite its sources?
Microsoft Copilot is an AI assistant that uses a chatbot interface to allow you to search for information, generate text, and create images. Use the steps above to verify the AI-generated data.
TRY IT OUT:
Ask Copilot for data on the indicators of Minnesota’s Asian population. Copy and paste the following questions into the Message Copilot prompt:
- How many Asian residents of Minnesota are there?
- What percentage of Minnesota's population is Asian?
How does Copilot's response compare with what you found on Minnesota Compass?
Google Gemini is an AI language model trained on a massive dataset of text and computer code. It can generate text, translate languages, write content, and answer questions. Cross-check responses from the following with what you find on Minnesota Compass.
TRY IT OUT:
Ask Gemini for data on health care coverage in Minnesota using the following question:
- What percentage of Minnesotans younger than age 65 lack health care coverage?
Is Gemini's answer consistent with what you find on Minnesota Compass? Does Gemini cite its sources?
ChatGPT is an AI tool that can engage in conversation, summarize and complete text, explain concepts, and assist with programming tasks like writing code snippets. Ask ChatGPT to expose possible sources of bias in its response.
TRY IT OUT:
Ask ChatGPT for help with information on trends in Minnesota’s quality of life. Copy and paste the following questions into the Message ChatGPT prompt:
- Are Minnesota’s quality of life indicators trending better or worse?
- Briefly explain how you came to your answer and list possible sources of bias in your response.
How does ChatGPT compare with the other AI tools? What guidance and information about the data does Minnesota Compass offer to address possible sources of bias?
As AI continues to evolve and interest in its use grows across sectors, the ability to trust and verify AI-generated data becomes increasingly crucial. By understanding the most common types of AI and deploying some practical safeguards—like cross-checking AI-generated data with Minnesota Compass—users can navigate the complexities of using AI with greater trust and confidence.
References
Brown, T. B., Mann, B., Ryder, N., Subbiah, M., Kaplan, J., Dhariwal, P., Neelakantan, A., Shyam, P., Sastry, G., Askell, A., Agarwal, S., Herbert-Voss, A., Krueger, G., Henighan, T., Child, R., Ramesh, A., Ziegler, D. M., Wu, J., Winter, C., … Amodei, D. (2020). Language models are few-shot learners. In H. Larochelle, M. Ranzato, R. Hadsell, M. F. Baclan, & H. Lin (Eds.), Advances in Neural Information Processing Systems 33 (pp. 1877-1901). NeurIPS. https://proceedings.neurips.cc/paper/2020/file/1457c0d6bfcb4967418bfb8ac142f64a-Paper.pdfDeepLearning.AI. (2023). A complete guide to natural language processing. https://www.deeplearning.ai/resources/natural-language-processing/
Toner, H. (2023, May 12). What are generative AI, large language models, and foundation models? Center for Security and Emerging Technology. https://cset.georgetown.edu/article/what-are-generative-ai-large-language-models-and-foundation-models/