Editorial · product notes
We hold 142,592 records.
We default to hiding the photos.
By Steven Haller · March 5, 2026
The People Not Numbers database, which our newsroom maintains alongside our colleagues at peoplenotnumbers.org, currently holds 142,592 records on people who are or have recently been incarcerated under the Michigan Department of Corrections. Across those records there are 520,719 sentences. Eighteen percent of those records — about 25,600 of them — arrive from the source with no photograph, because the source itself withheld the image. The rest arrive with a photograph attached.
When you load any individual record on our site, the photograph is hidden.
You can choose to reveal it. The control is right there. It is one click. We do not gate it behind a form, a sign-up, an email address, or a click-through warning. We do not make a moral show of it. We simply default the image to off, and we tell you what is behind the toggle, and we let you decide.
This is a product decision. It is a small one. Almost nothing in the contract that grants the public access to MDOC records under Michigan's FOIA carve-outs requires it. The records are public; the photos are public; many other sites display the photos at full size, on the index page, next to the person's name, the way a corporate intranet displays an employee headshot. We have chosen not to.
I want to write down why, because I am tired of being asked.
data being public does not mean the design has to be cruel
The most common version of the question, from technically inclined readers, is: if the data is already public on the state's own site, what difference does it make how you present it?
The answer is: the state presents it once, on a state-branded mugshot lookup tool that nobody outside of corrections-tech and a handful of journalists has bookmarked. We present it on a fast, well-indexed, mobile-friendly, search-optimised site that anyone with the person's name can find in a quarter of a second. The state's site is a fence around the records. Our site is, by design, a window. We have not changed what is public. We have changed how easily the public can humiliate someone with it.
So we put the photo behind a toggle. The cost to a researcher, a journalist, a family member, or counsel is one click. The cost to a stranger Googling someone's name out of idle curiosity is a small but real moment of having to decide if they actually need to see this person's face. That moment is what the design is for.
the four defaults, and why
There are four design defaults across every record page that we get asked about. I will lay them out plainly.
Photo: off.
Already covered above. The control is one click, labeled show photo. Below it we display the year and source of the photograph (always MDOC; usually a booking image from intake; in some cases years out of date relative to the person's current appearance). We do not display marks, scars, tattoos, or distinguishing-feature text without the photo reveal. If you do not toggle, you see none of it.
Date of birth: year only.
The state publishes full DOB. We display year. The reasoning is identity-theft-adjacent: a full DOB plus a name plus a place of incarceration is a sufficient seed for a great deal of low-grade identity attack. We accept the cost of imprecision for journalists doing demographic analysis; the full DOB is available on request, with a one-line reason, to credentialed researchers. We have approved every single such request that has come in. There have been thirty-one of them in the last year. None has been refused.
Marks and scars: gated.
This was the easiest call. The "marks, scars, and tattoos" field on a corrections record is a legacy of how the system was designed to identify people if they escaped. It is a field designed to make a body a fingerprint. Displaying it next to a name, by default, on a public website, in 2026, is the digital equivalent of shouting a stranger's body description at them across a crowded room. We gate it. You see it if you toggle the photo. The two reveal together, because they are the same kind of information, and they ought to be the same kind of choice.
"Request review": on every record.
Every page has a small link at the bottom that says request review of this record. It is a real link to a real form that goes to a real human being on our small editorial team. The form asks: what is inaccurate, what is out of date, who is the person submitting, and how do they know. We do not require ID. We do not require legal counsel. We do not charge. We review every submission. We have made changes to 347 records on the basis of these reviews in the last twelve months — most often to remove records of people who have been released and now show up on the site for a state record that has not been refreshed in eight months. We are slower than the state. We are also less embarrassed about admitting it.
what the design is not doing
Several things this design does not pretend to do.
It does not hide the records. It does not pretend the data does not exist. It does not require a login to read what is already on the public record. It does not allow the subject of a record to remove themselves from the database; doing that would defeat the journalistic purpose. It does not score, predict, or recommend. The site does not tell you whether a person is "high risk." It does not generate parole-likelihood scores. It does not run a face-recognition pipeline.
And it does not — this matters — claim the design solves the underlying problem. The underlying problem is that the state of Michigan maintains a public-by-default photo-and-bio dataset of 142,592 currently or recently incarcerated people, and there is no statutory mechanism for any of those 142,592 people to opt out, regardless of charge, regardless of outcome, regardless of how long they have been home, regardless of whether their conviction has been vacated. The state could change that with one paragraph of legislation. We have written about that, and will keep writing about it. Our defaults are a patch, not a fix.
the cost, in product hours
I am going to put a number on the cost of being humane, because if you cannot put a number on it, you do not actually know what you mean.
The "photo off by default with a one-click toggle" feature took two of our engineers about a working day, including the storage of the user's per-session preference, the accessibility audit on the toggle, and the change to the social-share card so that screenshots taken from the page do not silently include the image. The "request review" form, end to end, took about three working days, most of which was the back-end review queue, not the front end. The DOB-year-only and marks-and-scars-gated changes took about an afternoon each.
The cumulative engineering cost of every dignity-leaning default on the site is, generously, two engineer-weeks. The cumulative reader benefit is harder to measure, but our session logs show that 78% of record pageviews do not toggle the photo on. Three out of four readers, given a choice, choose not to look. The state's design does not give them the choice.
what other operators of public-dataset interfaces should take from this
The argument is small. The argument is: when you build an interface over a public dataset, you are not just exposing the data — you are choosing the defaults for how millions of strangers encounter it. That is a product responsibility. It is not a legal question. There is no FOIA-compliance officer to escalate it to. It is yours.
If you operate one of these interfaces — a court-record search, a mortgage-default scraper, a civic permit lookup, any system whose "users" include people who never asked to be there — I would ask one question. What does your design assume about the person being searched, and what does it assume about the person searching? Most of these systems assume the searcher is the customer and the subject is the product. Ours tries not to. We are not done. We are working on it. The next change, currently in review, is to display sentence length in "years remaining" instead of "years served," because that is what families ask for. It will ship in May.
Letters welcome. The door is open by letter. We try to answer letters from inside facilities first.