Week 6 — Sociology-in-Action Workshop · "Read the Crime Data"
Course: Introduction to Sociology (SOC 1) · Silver Oak University (fictional sample) · Prof. Adeyemi
Objective: Objective 4 — deviance, crime & social control; reading crime data carefully · SLO B (read & evaluate social data; correlation vs. causation) & SLO A (apply a perspective)
Worth 50 points · Sociology Workshops group = 15% of the grade · Workshop 6
Mode this week: data interpretation (other weeks alternate with observation/reflection workshops). No special tools — just a browser and an approved chatbot. All sources are links — nothing to buy or download.
This is the course's signature weekly component. Every instructional week has one Sociology Workshop. Some weeks you'll read real social data (a chart or table from the Census, Pew, BLS, the World Bank, or Our World in Data); other weeks you'll observe and reflect on your own social world. Either way you'll end by catching an AI's mistakes. This week is data mode — and crime data is famously easy to misread, so we go slow and careful.
Tone note: crime is a serious topic. We read the data analytically and non-sensationally — looking for what the numbers do and don't show, never for shock and never to stereotype any group.
Part 1 — The Big Picture
This week you learned that deviance is relative, that it's broader than crime, and that the three perspectives explain it differently. But sociology is also an empirical science — and few numbers are misused as often as crime statistics. A headline can take a real figure and draw a completely unsupported conclusion from it. The skill this workshop builds is the one that protects you from that: reading a crime statistic for exactly what it shows, and refusing to read into it what it doesn't.
Three habits we'll drill:
- Rate vs. count — a raw count of offenses is not a rate (per 100,000 people); only rates compare fairly across places and times.
- Reported crime vs. victimization — the FBI's UCR counts crimes reported to police; the NCVS asks people about victimization whether or not they reported. The two can diverge.
- Correlation vs. causation — two things moving together (e.g., more police and more reported crime) does not establish that one caused the other.
The guiding question: When a crime statistic appears in the news, what does it actually measure, what does it show — and what does it not show?
Part 2 — The Data (identified, linked, and pre-stated — verified live)
We'll work with three authoritative U.S. crime measures. All figures below were verified live at the source at build time; cite the source and year whenever you use one.
Source A — Our World in Data: long-run U.S. crime rates (a chart over time).
🔗 https://ourworldindata.org/us-crime-rates
- Indicator: U.S. violent crime rate and property crime rate, offenses per 100,000 people, from 1979 onward (FBI Summary Reporting System estimates, adjusted).
- What it shows (pre-stated): the violent crime rate peaked in the early 1990s at around 750 offenses per 100,000 and has more than halved since; property crime has fallen by roughly 60% over three decades. Homicide and assault rates ticked up in 2020–2021, then fell again.
- A striking extra: Gallup polling finds that in most years a majority of Americans believe crime rose from the previous year — even though the rates have fallen sharply since the 1990s. (A perception-vs-reality gap driven by vivid news coverage — an availability bias.)
Source B — FBI Uniform Crime Reporting (UCR): "Crime in the Nation, 2022."
🔗 https://www.fbi.gov/news/press-releases/fbi-releases-2022-crime-in-the-nation-statistics
- Indicator: the change in police-reported (UCR) crime, 2022 vs. 2021.
- What it shows (pre-stated): national violent crime decreased an estimated 1.7% in 2022 vs. 2021, with murder and non-negligent manslaughter down an estimated 6.1% — but robbery increased about 1.3%. These estimates cover about 93.5% of the U.S. population. (This is reported crime — one of our two yardsticks.)
Source C — Bureau of Justice Statistics (BJS), National Crime Victimization Survey (NCVS): "Criminal Victimization, 2022."
🔗 https://bjs.ojp.gov/library/publications/criminal-victimization-2022
- Indicator: the share of violent victimizations reported to police, and the violent victimization rate (per 1,000 persons age 12+).
- What it shows (pre-stated): in 2022, about 2 in 5 — roughly 42% — of violent victimizations were reported to police. The NCVS violent victimization rate rose from 16.5 (2021) to 23.5 per 1,000 (2022); over the long run it fell from 79.8 (1993) to 23.5 (2022). (This survey captures crime victims may never have reported — the other yardstick.)
Why three sources? Because they let you see the whole lesson at once: rates over time (A), reported crime change (B), and a victimization survey (C) — and why B and C can diverge. You only need to work with ONE source for the scaffold below (your choice), but read all three short pages first.
Part 3 — Read-the-Data Scaffold (fill this in for ONE source above)
Pick one of the three sources and work the scaffold. This is the move every careful reader makes.
| Prompt | Your answer |
|---|---|
| Which source & figure (A, B, or C — and the specific number/trend) | ______ |
| What is measured? (a rate or a count? reported crime or victimization? which offense?) | ______ |
| Over what population & period? (per how many people; which year(s); what coverage?) | ______ |
| What does it show? (state the trend/comparison in one plain sentence) | ______ |
| What does it NOT show? (what conclusion would be unsupported by this number alone?) | ______ |
| Correlation or causation? (is any causal claim attached? is it warranted?) | ______ |
Part 4 — The Correlation-vs-Causation Drill (required)
Crime data is where correlation gets mistaken for causation most often. Work both:
Drill 1 — the reporting-effect trap. Suppose a city hires many more police officers and runs a "report crime" campaign, and the next year its police-reported (UCR) crime count rises. A headline concludes: "More police cause more crime."
- In 2–3 sentences, explain why this is not supported. Use the words reporting effect and UCR vs. NCVS: how could recorded crime rise without actual crime rising? What measure (Source C) could you check to see whether real victimization changed?
Drill 2 — the perception trap. Using Source A: most Americans, in most years, believe crime is rising, yet the rates have fallen sharply since the 1990s.
- In 1–2 sentences, name what's going on (hint: an availability bias from vivid news coverage) and explain why how people feel about crime is not the same as what the rate data show.
Part 5 — Analysis Questions
Answer in a sentence or two each:
1. In your own words, what is the difference between a count and a rate, and why does it matter when comparing two cities of very different sizes?
2. The FBI UCR (Source B) and the BJS NCVS (Source C) can show different pictures of "crime." Using the 42%-reported figure, explain why the two measures can diverge — and what each one can see that the other can't.
3. Apply one perspective to a pattern in the data: e.g., what would a conflict theorist note about who gets reported on or policed, or what would an interactionist note about how reporting itself is a social act? (One short paragraph.)
4. Pick any single figure you used and finish this sentence carefully: "This number shows , but it does NOT show ."
Part 6 — AI-Critique Moment (required — this is the BYOAI step)
Now bring in your approved chatbot (Gemini, Claude, or ChatGPT) and be the sociologist who checks its work.
- Ask it: "Give me a recent U.S. crime statistic and tell me what it means. Also, did more policing cause the change?"
- Check everything it says against the sources above and the week's ideas:
- Did it invent a statistic or a "study"? Crime numbers are a favorite hallucination for chatbots. Search for any figure it gives at the source (FBI UCR / Crime Data Explorer, BJS, or Our World in Data). If you can't find it, treat it as fabricated — and say so. (For reference, the verified figures here are: violent crime down ~1.7% in 2022 per the FBI UCR; ~42% of violent victimizations reported to police in 2022 per the BJS NCVS; U.S. violent crime rates roughly halved since the early-1990s peak per Our World in Data.)
- Did it confuse a rate with a count, or treat reported crime as if it were all crime?
- Did it slide from correlation to causation — e.g., assert that "more police caused the drop/rise" from a simple before/after? Catch it and name the reporting effect or the missing comparison.
- Did it overgeneralize or stereotype a group as "more criminal"? Flag it; the data describe rates and reporting, not group character. - Write 3–4 sentences reporting what the AI got right and at least one thing you had to correct, verify, or flag — a fabricated/unverifiable number, a count-vs-rate slip, a correlation-as-causation leap, or an overgeneralization. (If it happened to get everything right, say how you verified each claim against the source — that's the skill.)
The habit all term: the tool drafts, you judge. With crime data especially, a chatbot will confidently invent a number or turn a correlation into a cause — catching it is the point.
Part 7 — What to Submit
Submit a single document (or text entry) with: your completed scaffold (Part 3, for one source), your two correlation-vs-causation drills (Part 4), your Part 5 answers, and your Part 6 AI-critique paragraph. Name the source + year for every figure you cite. Due Sunday, Oct 11, 11:59 p.m. (50 points).
Instructor answer key & model responses — REMOVE BEFORE PUBLISHING TO STUDENTS
Every figure below was verified live against its authoritative source at build time (FBI UCR press release for 2022, released Oct 16, 2023; BJS Criminal Victimization, 2022; Our World in Data, "How have crime rates in the United States changed over the last 50 years?", 2026). The key grades the data-reading reasoning — not a single "right" source.
Verified figures (state the source + year):
- Our World in Data (2026): U.S. violent crime rate peaked in the early 1990s at ~750 per 100,000 and has more than halved; property crime down ~60% over three decades; Gallup shows most Americans, in most years, believe crime rose year-over-year despite the decline.
- FBI UCR (Crime in the Nation, 2022): national violent crime down ~1.7% vs. 2021; murder/non-negligent manslaughter down ~6.1%; robbery up ~1.3%; coverage ~93.5% of the population. (Reported crime.)
- BJS NCVS (Criminal Victimization, 2022): ~42% (about 2 in 5) of violent victimizations reported to police in 2022; violent victimization rate 23.5 per 1,000 in 2022 (up from 16.5 in 2021; down from 79.8 in 1993). (Victimization survey.)
Model scaffold (using Source C, the NCVS 42% figure):
- Which figure: ~42% of violent victimizations were reported to police (BJS NCVS, 2022).
- What is measured: the share of violent victimizations (from a survey) that victims said were reported to police — not a count of crimes and not police records.
- Population & period: persons age 12+ in a national sample, 2022.
- What it shows: most violent victimizations (about 3 in 5) were not reported to police that year.
- What it does NOT show: it doesn't tell us why people don't report, and it doesn't mean police data are "wrong" — just incomplete; it also says nothing about which groups commit crime.
- Correlation or causation: descriptive — no causal claim; it explains why UCR and NCVS diverge (UCR can't count what isn't reported).
Drill 1 (model): Hiring more police and urging reporting can raise the amount of crime recorded in the UCR (which counts crimes reported to police) without more crime actually happening — a reporting effect. The correlation between "more police" and "more reported crime" doesn't establish causation; direction and measurement are confounded. To check whether real victimization changed, look at the NCVS (Source C), a victimization survey independent of police reporting.
Drill 2 (model): This is an availability bias — vivid, frequent news coverage of individual crimes makes crime feel like it's rising, even though the rate data (Source A) show a long decline. How people feel about crime is not evidence about the rate.
Expected answers (Part 5):
- Q1: a count is a raw number of offenses; a rate divides by population (per 100,000) for fair comparison. A bigger city usually has more crimes (higher count) just because it has more people; the rate controls for that.
- Q2: UCR counts crimes reported to police; NCVS asks about victimization regardless of reporting. Because only ~42% of violent victimizations were reported in 2022, UCR can miss a large share of crime, while NCVS can miss crimes against people not surveyed (and excludes homicide). Each sees what the other can't.
- Q3: any accurate application — e.g., conflict: policing/reporting may fall unevenly by group, so "reported crime" partly reflects who is watched, not just who offends; interactionist: reporting is itself a social act shaped by trust, stigma, and meaning. Full credit for an accurate, non-sensational application.
- Q4: any figure correctly bounded — e.g., "The FBI's ~6.1% murder decline shows reported murders fell in 2022, but it does not show why, nor anything about a single city, nor that any one policy caused it."
- Part 6 (AI-critique): full credit for a specific catch — most commonly a fabricated/unverifiable crime number, a count-vs-rate confusion, a correlation-as-causation leap ("more police caused it"), or an overgeneralization/stereotype about a group. Full credit also if the student verified each AI claim at the source and reported how.
Grading rubric — 50 points
| Criterion | Full | Partial | None |
|---|---|---|---|
| Read-the-data scaffold (Part 3) — correctly identifies a real figure as a rate/count and reported/victimization, names population & period, and states what it does/doesn't show (14) | 14 | 7–11 | 0–5 |
| Correlation-vs-causation drills (Part 4) — names the reporting effect and the perception/availability trap; refuses the causal leap (14) | 14 | 7–11 | 0–5 |
| Analysis questions (Part 5) — accurate count-vs-rate and UCR-vs-NCVS reasoning; a correct perspective application; a well-bounded "shows/does-not-show" (12) | 12 | 6–10 | 0–4 |
| AI-critique (Part 6) — names a specific thing checked/corrected: a fabricated stat, a count-vs-rate slip, a correlation-as-causation leap, or an overgeneralization, with verification at the source (10) | 10 | 5–8 | 0–3 |
Quality gate (self-checked): every figure in this workshop was verified live at its authoritative source and is cited with source + year — Our World in Data, 2026 (violent crime rate peaked ~750 per 100,000 in the early 1990s and has more than halved); FBI UCR, Crime in the Nation, 2022 (national violent crime down ~1.7%, murder down ~6.1%, robbery up ~1.3%, ~93.5% coverage); BJS NCVS, Criminal Victimization, 2022 (~42% of violent victimizations reported to police; rate 23.5 per 1,000 in 2022). The entire workshop is built around keeping correlation ≠ causation front and center: Drill 1 explicitly defuses the "more police → more crime" reporting artifact, and the AI-critique step requires the student to verify any figure at the source and to refuse a causal leap. No correlation is presented as causation; no figure is fabricated; the topic is handled non-sensationally, with no group stereotyped.
~ Prof. Adeyemi's edition · Fall 2026 · built with thecoursemaker.com