📨 AI for Social Impact Newsletter

👋 From the Editor

Hi! I’m Joanna. I am on a mission to help folks in the social impact sector understand, experiment with, and responsibly adopt AI. We don’t have time to waste, but we also can’t get left behind.

Let’s move the sector forward together. 💫

🧠 EDUCATION

• Hallucinations: Hallucinations are just a fancy way of saying that large language models (LLMs) make stuff up. LLMs don't actually "know" anything, rather they predict what word should come next based on patterns of an insane amount of data. They are “generative”, so their job is to generate output no matter what. You can always add language to your prompts like “always cite sources”, “if you don’t know, say I don’t have verified information” or my go-to “don’t make stuff up”. 😂

• Algorithmic Bias: AI training data comes from the internet, which means it's heavily skewed toward content created by people who have historically had access to technology (read: predominantly white, male, English-speaking folks from the Global North). Check out Dr. Joy Buolamwini, a computer scientist, poet, and founder of the Algorithmic Justice League. In her work, The Gender Shades Project, she brought to light algorithmic bias in commercial AI products. You can also try Latimer, an LLM that was trained with diverse histories and inclusive voices.

✨ INSPIRATION

• When Good Intentions Miss the Mark: Pennsylvania's Allegheny Family Screening Tool was designed to help social workers make more objective decisions about which families to investigate for child neglect, but instead it did the opposite—it created bias against families with disabilities. The encouraging news? The controversy sparked important conversations about algorithmic fairness, with Oregon dropping its similar tool due to equity concerns.

• Recipe for Disaster: Hallucination in action. Microsoft's AI-powered travel guide recommended the Ottawa Food Bank as a "tourist hotspot," encouraging visitors to arrive "on an empty stomach", confusing a Canadian nonprofit helping food-insecure families with a trendy restaurant. 😬 I suppose the silver lining of this spectacular AI failure is free marketing for the food bank?

🚀 ACTIVATION

• One Story, Four Formats: The beauty of using LLMs is the efficiency they create. Try this one out for size: “Here is a program outcome: [paste your story/data and mission of your organization]. Please help me adapt this into four versions: 1) a compelling slide headline + 2 bullet points for a fundraising presentation, 2) a 100-word impact story for a grant proposal, 3) a social media post with emotional hook and call-to-action, and 4) a personal acknowledgment letter opening that connects the donor to this specific impact. Please let me know if there is information missing that would strengthen any of the stories, and don’t make stuff up.” 😉 And that’s how one story becomes a communications toolkit!

• AI-Powered Report Planning: AI can be your strategic thinking partner to write any kind of report. Simply adapt this prompt to your needs: “I'm creating a [type of report] for [mission, programs]. My audience includes [donors, board members, trustees, community members, etc.]. This is my first time creating this report, so please ask me one question at a time to help me build a comprehensive outline that covers all the essential sections. Thanks." Then answer each question the AI asks, and watch as it builds your personalized outline section by section!

And if you were wondering, yes, I say “please” and “thank you” to a robot. 🙏🏼

🤖 What I Asked AI This Week

Where do squirrels go when it rains? 🐿️

If you want to know the full answer, hit reply! Spoiler alert: umbrella tails! 🆒

👀 ICYMI

If you’re new here, welcome! You can check out the archive of past newsletters here.

💬 What Do You Think?

This newsletter is meant for you, so I’d love to hear what you think. 💌 Reply anytime!