What Bentham Actually Built
The Panopticon, designed in 1785, is a circular building. Cells line the outer ring, each backlit by an exterior window so a person standing inside is always visible in silhouette from the center. A single inspection tower sits in the middle, its own windows fitted with blinds and angled so that anyone inside the cells can see that the tower exists, but never see whether anyone is in it right now.
The building's guard-to-prisoner ratio is almost irrelevant to how it works. One person in the tower, or nobody at all, produces close to the same effect, because the design never depended on constant observation. It depended on constant uncertainty about observation.
The Trick Was Never the Watching
Michel Foucault's later reading of the Panopticon in Discipline and Punish named the mechanism precisely: power that works this way has to be visible and unverifiable. Visible, because the inmate has to know the tower is there and could be occupied. Unverifiable, because the inmate can never confirm whether it is occupied at this particular moment. Those two conditions together are sufficient. Once someone internalizes "I might be watched right now," they start watching themselves, and they do it continuously, not just during the fraction of time anyone could plausibly be reviewing them.
That's the part worth sitting with: full-time human observation was never the goal, and was never actually achievable at scale even in Bentham's design. Self-monitoring driven by uncertainty was the goal, and it scales for free.
Screen Recording Does the Same Job Digitally
Workplace screen-recording and activity-monitoring software reproduces both conditions almost exactly. Visible: employees are told the software exists, sometimes see an icon for it, sign a policy acknowledging it. Unverifiable: almost nobody knows, on any given afternoon, whether a human being is actively reviewing their screen right now, reviewing a stored recording later, or whether the capture is sitting untouched in storage and will never be watched by anyone at all.
| Panopticon | Screen Recording |
|---|---|
| Backlit cell, always visible in silhouette | Continuous capture, always technically available |
| Tower with blinds, presence unconfirmable | No indicator for "a human is watching this exact moment" |
| Inmate self-regulates against a possible observer | Employee self-regulates against a possible reviewer |
| One guard can effectively govern many cells | One compliance review can never cover most captured hours |
The behavioral effect shows up regardless of the real review rate, which in most deployments is a small fraction of what's captured. That's not a flaw in the system from the perspective of whoever deployed it. It's the same design principle Bentham was working with: the disclosure does most of the work, the actual watching is almost incidental.
"This Chat May Be Monitored" Does the Same Job With Words
A one-line disclosure banner, a compliance-archiving policy someone mentioned in onboarding, a company Slack or Teams instance known to log everything for retention reasons: none of these require anyone to actually read a given conversation to have an effect on it. The announcement alone supplies the "visible" half of the mechanism. The not knowing which conversations, if any, get pulled up later supplies the "unverifiable" half. People self-edit in a monitored, rarely-read channel almost as much as they would in a fully-read one, because the editing happens against the possibility, not the statistic.
Why "Announced" Is the Important Word
This is the counterintuitive part, and it's also the reason the word "announced" in a monitoring policy is doing more structural work than the word "monitoring." Secret, undisclosed surveillance produces no Panopticon effect at all, because nobody adjusts behavior against a possibility they don't know exists; it can only ever catch people after the fact, not shape what they do beforehand. A credibly announced policy with genuinely minimal enforcement can, per Bentham's own logic, get most of the way to the same behavioral outcome as heavy enforcement, at a fraction of the operating cost. The architecture is built to make the disclosure carry the weight, not the review team.
What This Means in Practice
For anyone deploying monitoring: the announcement is not a formality on top of the "real" mechanism, it largely is the mechanism. Coverage and review rate matter far less to behavior than most people assume when budgeting for enforcement capacity.
For anyone working under an announced policy: recognizing the mechanism is itself useful information. The pressure to self-edit is coming from your own uncertainty about a specific moment, not from a documented fact that someone is watching it. Bentham described his own design's ambition as a way of obtaining "power of mind over mind" through architecture alone. The modern version does the same thing through a status-bar icon and a sentence in the employee handbook, and it works for the identical reason: not because it watches everyone, but because nobody can prove it isn't watching them.