Skip to content
Playbook
AdvancedPillarHRFounderCEO

Foucault's Panopticon: Why Modern HR Surveillance Changes People Before It Even Sees Them

Foucault used Bentham's Panopticon prison design to describe a form of power that works through the possibility of being watched, not the fact of it.

12 min read
On this page
60-Second Summary
  • Foucault (Discipline and Punish, 1975): the Panopticon shows how power works through visibility. Being watchable — even without being actively watched — makes people discipline themselves.
  • Modern HR replicates this: keystroke logging, calendar mining, presence indicators, activity dashboards, AI screening. The disciplinary effect is present the moment employees know it could be running.
  • The measured behaviour changes before the measurement is used. What you now capture is 'behaviour under known observation', not baseline performance.
  • Second-order harms: performative work, trust decay, migration of real work to un-observed channels, self-censorship, and disproportionate impact on already-marginalised groups.
  • Design rules: minimise surveillance by default, publish exactly what is collected and used, prohibit secondary use, give employees the same visibility over their data, and prefer outcome measurement over activity measurement.

A mid-market firm rolls out an 'employee productivity platform' — keystroke frequency, application-in-focus time, Slack response latency, meeting density, and an AI-generated weekly 'focus score'. Within a quarter, output looks identical on the dashboards. Six months in, five of the strongest senior engineers have left, referrals have collapsed, and a legal-team review flags disproportionate impact on caregivers and disabled colleagues. Leadership never actually looked at most of the data. The damage was done by the fact that people knew it existed. This is the panopticon working exactly as Foucault described.

What Foucault actually argued

He who is subjected to a field of visibility, and who knows it, assumes responsibility for the constraints of power; he makes them play spontaneously upon himself; he inscribes in himself the power relation in which he simultaneously plays both roles; he becomes the principle of his own subjection.
Michel Foucault, Discipline and Punish (1975)

Foucault took Jeremy Bentham's 18th-century Panopticon prison — a circular building where a single guard in a central tower could potentially see any cell, without the prisoners knowing when they were watched — and made it a model of modern power. The genius of the design is not the watching, it's the possibility of being watched. Prisoners self-discipline because they cannot verify they are not being observed. The tower can be empty and the effect is the same.

Foucault's broader claim: modern institutions — schools, hospitals, workplaces, the military — moved historically from spectacular punishment to distributed disciplinary power. Instead of one visible sovereign, thousands of small mechanisms of observation, ranking, examination, and record-keeping produce docile, productive bodies. The workplace HR function, when unexamined, is one of the most complete implementations of this pattern.

The important corollary: disciplinary power is not necessarily malicious. It is efficient. It reduces the visible cost of coercion because the coerced coerce themselves. That is what makes it hard to see and hard to argue against — the tools look like 'good management'.

The modern HR panopticon inventory

Where the panoptic pattern shows up today
  1. 1
    Activity monitoring
    Keystroke frequency, mouse movement, application-in-focus tracking, webcam-on requirements. Vendors: Hubstaff, Time Doctor, Teramind, ActivTrak, and native features in some MDM stacks.
  2. 2
    Communication analytics
    Slack/Teams/email metadata mined for response latency, network centrality, sentiment, after-hours activity, meeting density. Sold as 'wellbeing' or 'collaboration' analytics.
  3. 3
    Presence indicators
    Green/yellow/red dots, 'active X minutes ago', calendar auto-blocking. Low-tech but panoptically powerful — always visible, always potentially interpreted.
  4. 4
    Performance instruments
    Continuous feedback tools, ratings, calibration decks, 9-box grids. See the second-order cybernetics article for why measurement itself is intervention.
  5. 5
    AI screening and ranking
    Resume filters, interview transcript scorers, meeting summarisers that rate participation. Employees increasingly perform for the model, not the human.
  6. 6
    Physical/location tracking
    Badge-in analytics, RTO enforcement dashboards, WiFi triangulation, hot-desking apps that record who sat where. Even 'anonymous' aggregates create the sense of being watchable.
~70%
of large US employers use monitoring tools of some kind (up sharply post-2020)
Gartner, Brookings, ExpressVPN Digital Rights (2022–2024 surveys — cited ranges vary; direction is stable)
~2–3×
reported employee stress, turnover intent, and disengagement under intense monitoring vs light monitoring
MIT Sloan / Harvard Business Review syntheses (2022–2024)

The point is not that every one of these tools is illegitimate. The point is that they aggregate. Any single tool feels innocuous. Together they build the tower — and the tower does its work whether or not anyone climbs it.

The second-order harms most leaders miss

First-order (measured) vs second-order (real) impact
What leaders see on dashboards
  • Activity metrics stable or up
  • Response times improved
  • Meeting attendance perfect
  • 'Productivity scores' trending green
  • Attrition roughly flat this quarter
What is actually happening
  • Performative work: mouse-jigglers, ghost commits, meeting appearance without contribution
  • Real thinking migrates to un-observed spaces (personal notes, walks, offline calls)
  • Silent self-censorship: risky ideas, dissent, and admissions of confusion vanish from Slack
  • Trust decay — seniors quietly line up exits; referrals dry up first
  • Disproportionate impact on caregivers, disabled colleagues, and neurodivergent employees, whose working patterns don't match the assumed norm
The 'we don't actually look at it' defence doesn't hold

Foucault's core insight: the disciplinary effect is caused by the possibility of being watched, not by the watching. 'We collect it but rarely review it' is not a mitigation — it is the exact condition that maximises the panoptic effect at minimum operational cost to the employer. If you're collecting it, you're deploying it.

Design rules for non-panoptic HR

Eight principles for HR that observes without panopticism
  1. 1
    Data-minimisation as default
    Collect the smallest amount of data needed for a stated purpose. If a business question can be answered without individual-level activity data, don't collect it. This is also the legal default in GDPR/UK GDPR and increasingly in US state privacy laws — align to the higher standard globally.
  2. 2
    Purpose-lock every data source
    Each collected data type has one stated purpose and cannot be repurposed for another (e.g., wellbeing data cannot be used for performance decisions). Publish the list; audit against it.
  3. 3
    Publish what is collected and why
    Employees see the full inventory of what the company can observe about them, at what granularity, retained for how long, and who can access it. Opaque surveillance is qualitatively worse than open surveillance.
  4. 4
    Give employees symmetric visibility over their own data
    Anything the company holds about an employee's activity should be visible and exportable to the employee. Symmetry breaks the tower's asymmetry, which is the core of the panoptic effect.
  5. 5
    Prefer outcome measurement over activity measurement
    Measure work shipped, decisions taken, customers served — not keystrokes or meeting count. Outcome measurement resists gaming better and disciplines less.
  6. 6
    No secret evaluation criteria
    If an AI model, screening filter, or scorecard is used in a decision about someone, they should be able to know it was used and on what basis. This is also emerging law (EU AI Act, NYC AEDT rules).
  7. 7
    Sunset instruments on a schedule
    Every monitoring tool decays into a proxy for something else. Time-box adoption and re-justify continuation.
  8. 8
    Document the disproportionate-impact analysis
    Before deploying any monitoring or scoring tool, run and record a bias/adverse-impact analysis by protected characteristic, working pattern, and role type. Keep it updated.
The tell for a healthy surveillance posture

You should be able to publish your full monitoring inventory internally without leadership discomfort. If publication would cause a stir, the stir is the honest signal — the monitoring is doing panoptic work that could not survive open air.

FAQ

Frequently asked questions

Doesn't remote work require more monitoring?

There is no serious evidence that activity monitoring improves remote outcomes; the evidence for outcome-based management under remote conditions is much stronger. The 'remote requires monitoring' claim usually reflects lost trust and a management skill gap — solvable by better cadence, better goals, and better 1:1s, not more tracking.

What about compliance-driven monitoring (finance, healthcare, defence)?

Legitimate. The Foucauldian point is not 'never observe' — it is 'observe with the narrowest scope, the clearest purpose, and the most symmetric visibility that the compliance regime permits'. Compliance-scoped monitoring is compatible with all eight design rules above.

How does this relate to Second-Order Cybernetics?

Directly. Second-order cybernetics tells you the observer is inside the system and that measurement is intervention. Foucault tells you the specific mechanism by which the intervention works in workplaces — self-discipline induced by possible visibility. They are the epistemology and the mechanism of the same problem.

How does this relate to Loose Coupling?

In heavily surveilled cultures, the gap between formal policy and actual practice widens — because real work migrates to un-observed channels. Surveillance often produces more loose coupling, not less.

Takeaways

  • The panoptic effect is caused by the possibility of being watched, not by watching. 'We don't actually look at it' is not a defence.
  • Aggregate the modern HR observation stack and you have a distributed panopticon whether or not any single tool intended it.
  • The measurable effects look fine; the real damage is in performative work, trust decay, silent self-censorship, and disproportionate impact on already-marginalised colleagues.
  • Design HR that observes without panopticism: minimise, purpose-lock, publish, give employees symmetric visibility, prefer outcomes over activity, and sunset instruments regularly.
Written by Pawan Joshi.Sources cited inline.
First published 12 Jul 2026See site changelog →