Community advisory board
A group of people from the community a project serves who meet with the research team regularly. They review plans, ask questions, and tell us when something won't work in real life. Their input shapes the project rather than just reacting to it.
Knowledge graph
A way of storing information as a web of connections instead of rows in a table. Rather than a list of shelters, it holds the relationships between them: which one serves which need, where each is, and when it's open.
Community co-design
Building something together with the people who will use it, starting at the first sketch instead of asking for feedback at the end.
Social determinants of health
The conditions where people live, work, and grow up, like housing, food access, income, and transportation. They shape health as much as medical care does.
Institutional review board (IRB)
A committee that reviews research involving people before it can begin. Its job is to make sure participants are treated fairly, understand what they're agreeing to, and are protected from harm.
Focus group
A guided group conversation, usually 60 to 90 minutes, where participants share their experiences and opinions on a topic. It's meant to surface how people actually think, not to reach agreement.
Qualitative codebook
A shared guide for labeling what people say. It defines each theme we're tracking so that everyone analyzing the conversations applies the same meaning to the same words.
Mixed methods
Using numbers and stories together. Survey data can tell us how many people something applies to, while interviews and focus groups tell us why.
Predictive modeling
Using patterns in past data to estimate how likely something is for someone now, such as who may end up in the emergency department in the next few months. It gives a probability, not a verdict, and it can only learn from the data it was given, so who was missing from that data matters as much as the math.
Screener
A short set of questions people answer before joining a study, used to check who fits what the study is looking at.
Natural language processing (NLP)
Teaching a computer to work with human language: the words in clinic notes, survey answers, or transcripts. It can sort thousands of documents by topic or pull out every mention of a symptom, which is useful when there's far more text than a team could read by hand.
Human in the loop
Keeping a person in the decision at the points that matter. The system can sort, draft, or suggest, but someone qualified reviews it before it affects care or a person's access to services. It's how we catch the confident mistakes.
Agentic AI
An AI setup that carries out a task in steps instead of answering one question at a time. Given a goal, it can look something up, decide what to do next, and use a tool to do it. More independence means more can go wrong unnoticed, so these systems need clear limits and human review.
Large language model (LLM)
The kind of AI behind chatbots. It predicts likely wording, which makes it fluent but also means it can state something confidently that isn't true. Pairing it with verified data is how we keep answers trustworthy.
