Homelessness and addiction are hard enough to understand before we start debating solutions. One organization focuses on housing. Another emphasizes treatment. A rescue mission sees the importance of relationships, accountability and spiritual change. Each may be responding to a real need, but it is difficult to compare their approaches when we cannot see how people move through the larger system.
That is where systems thinking becomes useful. It gives us a way to map the problem before choosing an intervention. AI tools can help gather and organize the data, while worldview analysis helps us ask a different question: What are we ultimately trying to accomplish for the people we serve?
Key Takeaways
- Map both entries and exits before judging whether an intervention changes a social problem.
- Verify AI-generated research and diagrams, especially geographic figures, labels and arithmetic.
- Case management can bridge the gap between emergency shelter and residential recovery.
- Data describes system behavior, while stated values clarify the outcomes we seek.
Table of Contents
- Start With Stocks, Flows and Feedback
- Build a Systems Map With AI Deep Research
- What a Homelessness Map Reveals
- Map Addiction as a Pathway, Not a Single Event
- Design a Continuum of Care With Different Levels of Support
- Ask Whether an Intervention Changes the System
- Use Data to Describe What Is, Then Examine What Ought to Be
Start With Stocks, Flows and Feedback
A systems diagram begins with a simple distinction. A stock is something that accumulates, like water in a lake. A flow is what enters or leaves it, like a river. System dynamics describes how those stocks and flows interact over time, including feedback loops that bring people back to an earlier state.
For a social problem, the stock might be people experiencing homelessness or people living with untreated addiction. The flows might include people entering homelessness, moving into housing, entering treatment, returning to addiction or reaching sustained recovery.
This is more useful than a single headline number. If the number of people experiencing homelessness stays steady, that does not mean nobody is finding housing. Many people may be leaving while just as many enter. To change the stock, we need to understand both sides of the lake.
Build a Systems Map With AI Deep Research
I use a five-step process to turn a complicated issue into a map that can guide discussion.
- Choose an issue that has meaningful states and transitions. Homelessness, addiction, mental illness and crime are possible starting points because people can move into and out of identifiable situations.
- Research the stocks and flows. Ask a deep research tool for the population in each state, the rates of movement between states and the sources behind its figures. Specify the geography. A national model and a county model need different data.
- Generate a diagram from the research. Provide the research report to an image-generation tool and ask it to depict the stocks as lakes and the incoming and outgoing flows as rivers. Name the categories that must appear.
- Create supporting materials. Gemini Notebook can turn a report into resources such as an audio or video overview, slides or an infographic.
- Iterate and verify. Improve the report and image until the categories, labels and arithmetic make sense. If you have reliable data about your own organization, add its role to the map.
A useful research request is specific about what the eventual diagram must explain. For homelessness, I asked for data that would support incoming and outgoing rivers and distinguish transitional, episodic and chronic homelessness. I first built a United States report, then a separate one for San Diego County.
Do not expect the first image to be right. My homelessness diagram took roughly five rounds of work on both the research and the image. AI can misspell labels, mix geographic levels or produce numbers that do not add up. The San Diego draft, for example, appeared to carry a national chronic homelessness figure into a county-level diagram. That is not a small cosmetic problem. It changes what the picture means, so it needs to be fixed before the map informs a decision.
What a Homelessness Map Reveals
The river metaphor helps explain why homelessness is not one uniform experience. In the national diagram, transitional homelessness is represented as a fast-moving flow, with an illustrative 80% share. Episodic homelessness, shown at 10%, is more like churning rapids as people move in and out of places such as hospitals, treatment centers and jails. Chronic homelessness is shown as another 10% of the flow, but its slower movement creates something like a stagnant lake.
Those percentages are figures used in the working model, not numbers to transfer uncritically to another place or year. The important systems insight is the difference between how many people enter a category and how long they remain there. A smaller inflow can still produce a large, persistent stock when the outflow is slow.
The map also makes different interventions visible. Permanent supportive housing, rapid rehousing and transitional housing create potential routes out. At the same time, active addiction, untreated behavioral health conditions and the trauma of unsheltered life can become choke points. These are areas where rescue missions and Christian social service organizations often concentrate their work.
Cost belongs on the map, but we need to know what a cost comparison measures. The example diagram contrasts approximately $40,451 in annual system costs for a person outside housing with about $24,000 for housing. The larger figure considers costs across public institutions, including medical and emergency services. That comparison helps explain a financial argument for housing, but a savings claim alone does not tell us whether addiction, trauma or other needs have changed.
The systems question is practical: Are people moving toward stable exits faster than others are entering or returning? If not, an individual program may be doing valuable work while the overall lake keeps growing.
Map Addiction as a Pathway, Not a Single Event
The same method works for addiction. I began with national data and then developed a California version, since state-level information is often easier to find than consistent city-level figures. The national working diagram places an estimated 42.4 million people in a lake of untreated addiction and distinguishes chronic from mild or moderate conditions.
Multiple streams feed that lake, including alcohol, nicotine, marijuana, pain relievers and other substances. From there, people may enter outpatient or residential treatment. Some move toward recovery; others relapse, transfer to another form of care, become incarcerated or die. Mental illness can overlap with addiction, complicating the pathways.
The diagram includes illustrative relapse figures of 40% to 60% and a higher one-year figure without aftercare. Those figures should be checked against the population and treatment being modeled rather than treated as universal rates. Their role in the map is to make a feedback loop impossible to ignore. Getting someone into treatment matters, but so does what happens after treatment.
That creates several points of engagement for ministries. Friends and church communities may help someone take the step into care. Residential recovery programs can provide sustained support. Relationships and aftercare may help people remain on a recovery path. In the model, reaching five years of stability represents a substantially stronger position than simply completing an initial treatment episode.
Design a Continuum of Care With Different Levels of Support
Once we can see the flows, program strategy becomes clearer. A city’s continuum of care includes organizations with different capacities, boundaries and purposes. Government and secular agencies can offer scale and broad access. Christian organizations often work through closer relationships and more intensive expectations, but they cannot serve everyone and do not have government’s funding capacity.
I find it helpful to think about progressive engagement. Jesus served crowds, but he also invested more deeply in smaller groups. Social services likewise need ways to offer wide initial access while allowing people who want more intensive support to move into it.
Here I prefer the word boundary to barrier. A shelter can have relatively few entry requirements and still maintain expectations that help people feel safe. A dry shelter may set different rules from a low-barrier shelter that accepts people under the influence. As people move toward treatment and residential recovery, the level of structure generally rises.
The middle tier is especially important. Traditionally, a rescue mission might have offered a choice between a shelter bed and an intensive residential recovery program. That leaves a gap for someone who needs more than a bed but is not ready for, or does not need, the full residential model.
Intensive case management can fill that gap. It brings greater expectations and individualized support than a shelter stay without requiring the commitments of a residential program. Springs Rescue Mission is one example of a mission investing in case management. As transitional homelessness has become a larger part of the challenge, this middle tier deserves much more attention.
Residential recovery remains a distinct option. Christian programs commonly run for nine to twelve months and ask participants to accept substantial structure and accountability. Their Christian identity may be central to how they approach change. That can be a strength for people open to faith-based care, while others may prefer a secular or different faith-based provider. A functional continuum should make room for those choices rather than pretending one organization is the right match for everybody.
Ask Whether an Intervention Changes the System
Donella Meadows’ framework of twelve leverage points helps us ask how deeply an intervention acts on a system. Adding capacity or a buffer can be valuable, especially when it protects someone in immediate danger. But it is different from changing the feedback loops, goals or assumptions that keep the problem in place.
This is where I question programs that measure success only while someone remains inside a service. Shelter, housing and harm reduction can provide real immediate benefits. Still, if our intended outcome includes recovery from addiction or sustained stability, we must ask what happens afterward. Does the intervention change a person’s longer-term path, or does it mainly delay a return to the same conditions?
Systems thinkers describe patterns such as fixes that fail and shifting the burden: a short-term response relieves pressure but leaves the underlying dynamic untouched. These are questions to investigate, not labels to apply automatically to every housing or harm reduction program. We should examine outcomes over time and be clear about whether we are measuring safety, housing stability, treatment engagement or lasting recovery.
Our organizations need the same scrutiny. Single-loop learning asks whether we execute a program well. Double-loop learning asks whether that program fits what the community actually needs. Triple-loop learning asks whether our underlying theological vision and understanding of human flourishing are guiding us well. Better operations cannot compensate for pursuing the wrong goal.
Use Data to Describe What Is, Then Examine What Ought to Be
A systems diagram is a powerful descriptive tool. It can show how many people are in a stock, where they come from and which pathways lead out. It cannot, by itself, settle every ethical question about what we should prioritize.
This is the distinction associated with David Hume’s discussion of moral reasoning: an account of what is does not automatically establish what ought to be. We still need to state our values. Are we aiming to reduce immediate deaths, expand access, lower public costs, support recovery, foster spiritual transformation or pursue several of these together? Those goals affect the interventions we choose and the outcomes we measure.
From a Christian worldview, I want the goal to include more than survival. I want to see people move toward genuine, lasting transformation. That conviction does not make emergency protection unimportant, and it does not remove the need for careful evidence. It means we should be honest when an intervention keeps someone safer today but has not yet demonstrated the deeper change we hope to see.
The opportunity is to bring these approaches together without flattening their differences. Use AI-assisted research to build a map, verify it against credible data, locate the points where your organization can contribute, and make your goals explicit. Then keep revising the model as you learn. Complex social problems will not fit neatly into one diagram, but a good diagram can help us ask far better questions and work together more intelligently.