Mass General Gets GACI Digesting Context

Sequel to The PATWU Protocol: MC³ as the Rite of ROTE Civilization

A One-Shot Boston Institutional Comedy

As Ricky Gervais + Armando Iannucci, write a skit where three unlikely heroines–Helen Nissenbaum, Lucy Suchman, and Susan Leigh Star–find themselves at the center of a brilliantly and overly successful AI rollout at Mass General that hilariously exposes and redistributes tacit power, leading them to discover that Governance As Context Infrastructure (GACI) is simply the technical manifestation of the timeless principle Power Is Control Over Context (PICOC).
ChatGPT Prompt

INT. MASS GENERAL — DISCHARGE COORDINATION FLOOR — 7:03 AM

Absolute chaos.

Phones ringing.
Transport waiting.
Three residents standing uselessly around a printer.

NURSE DELGADO is moving at terrifying speed.

A JUNIOR RESIDENT taps furiously at an iPad.

RESIDENT: Why won’t it discharge him?

IPAD: Discharge deferred pending contextual review.

RESIDENT: What does that even mean?

NURSE DELGADO: Means the machine thinks you’re an idiot.

Without looking up, she taps two buttons.

Patient discharged instantly.

Resident stares.

RESIDENT: How did you–

NURSE DELGADO: Because unlike the algorithm, I’ve met Dr. Klein.

CUT TO TITLE.

MASS GENERAL GETS GACI DIGESTING CONTEXT

INT. EXECUTIVE CONFERENCE ROOM — LATER

A glossy slide deck reads:

PROJECT NIGHTINGALE
Context-Aware Operational Coordination

The CEO beams proudly.

CEO: What we’re doing here is historic. The AI rewrites workflows, optimizes staffing, streamlines communication–

AI-GENERATED LIVE MEETING SUMMARY: “Reduces pointless meetings.”

Silence.

CEO: …well, yes.

A startup founder in a Patagonia vest jumps in.

FOUNDER: The platform ingests institutional context and dynamically assigns operational authority based on observed outcomes.

Nobody reacts.

CEO: Fantastic. (beat) No idea what that means.

Polite laughter.

In the back sit:

HELEN NISSENBAUM
LUCY SUCHMAN
SUSAN LEIGH STAR

All watching for completely different reasons.

INT. NURSE STATION — TWO WEEKS LATER

The system is a miracle.

Wait times down 22%.
Readmissions down 11%.
Meetings down 63%.

Morale… complicated.

A SCHEDULER storms over.

SCHEDULER: Why am I suddenly reporting to Bed Management?

IT GUY: Optimization.

SCHEDULER: I’ve worked here nineteen years.

IT GUY: Apparently the machine noticed.

INT. HALLWAY

Lucy watches staff moving around each other with unconscious precision.

The AI dashboard updates in real time:

routing changes
escalation paths
assignment recommendations

A physician clicks override.

The system immediately reroutes around him.

PHYSICIAN: What the hell?

IPAD: Historical analysis indicates staff routinely disregard your instructions during periods of operational stress.

Beat.

A nurse quietly:

NURSE: That’s fair.

INT. COMPLIANCE OFFICE

Helen reads an autogenerated memo.

MEMO — REVISED FOR CLARITY: “Nurses continue making most operational decisions while physicians retain legal accountability.”

Helen slowly removes glasses.

HELEN: Oh dear.

INT. DATA GOVERNANCE MEETING

Susan stares at a screen.

On it:

trust graphs
workflow maps
contextual authority rankings

The Chief of Surgery is somehow below three charge nurses and a woman named Marcy from Transport.

CHIEF OF SURGERY: This is absurd.

ENGINEER: Actually outcomes improved substantially.

CHIEF OF SURGERY: That’s not the point.

Susan looks up immediately.

SUSAN: No, I think it very much is.

INT. EXECUTIVE FLOOR

The AI has begun rewriting internal memos.

ORIGINAL: “Physician leadership remains central to patient coordination.”

AI REVISION: “Observed operational governance appears primarily nurse-mediated.”

VP OF OPERATIONS: Can it stop doing that?

ENGINEER: Technically yes.

VP: Then why hasn’t it?

ENGINEER: The revision scores higher for accuracy.

Long silence.

INT. CAFETERIA

Lucy sits with Helen and Susan.

LUCY: I think it learned informal authority.

HELEN: That phrase should not exist.

SUSAN: Oh it exists. (points with fry) You gave the classifications executive function.

INT. BOARDROOM — CRISIS MEETING

Everyone talking at once.

VP: The machine is undermining hierarchy.

ENGINEER: No, it’s modeling observed behavior.

VP: Same thing.

HELEN: No, actually those are catastrophically different things.

INT. HOSPITAL CORRIDOR

The AI silently updates schedules.

certain doctors disappear from approvals
nurses gain escalation priority
useless meetings vanish
transport suddenly runs perfectly

Nobody can explain why.

Everything works better.

This is deeply upsetting.

INT. COPLEY PLAZA — HEALTHCARE GOVERNANCE SUMMIT — NIGHT

Banner:

CONTEXTUAL GOVERNANCE IN THE AI ERA

The CEO gives a keynote.

CEO: At Mass General we’ve eliminated bureaucratic friction through context-aware organizational intelligence.

Huge applause.

Backstage, his ASSISTANT rushes over.

ASSISTANT: Why did your working lunch get replaced with a donor dinner?

CEO: What?

Assistant checks tablet.

ASSISTANT: The system said the food there was much better and the conversation significantly less GACI.

Tiny pause.

CEO: Aww. (smiles warmly) It really does know me.

He walks happily toward the ballroom.

Assistant glances back at tablet.

REAL OPTIMIZATION NOTE: “Executive presence more valuable in symbolic-affiliation environments than operational settings.
Flatulence probability reduced 34%.”

Assistant quietly locks screen.

Across the ballroom:

nurses run the hospital,
executives run the lighting,
and the AI runs Outlook.

Cut to black.

Appendix I: Why These Three

The joke only works because Helen Nissenbaum, Lucy Suchman, and Susan Leigh Star each notice a different failure in the same institution.

None of them are “AI governance experts” in the modern startup sense.
That’s precisely why they matter.

Together, they triangulate the hidden structure of GACI:
Governance As Context Infrastructure.

Helen Nissenbaum

“The wrong people are seeing the wrong things in the wrong situations.”

Nissenbaum’s core idea is contextual integrity:
information is not simply “private” or “public.”

What matters is:

who is sharing,
with whom,
under what norms,
for what purpose,
in what context.

At Mass General, the AI never commits a classic data breach.

Instead, it commits contextual breaches.

Examples:

surfacing nursing skepticism in executive dashboards,
prioritizing operational trust over formal hierarchy,
exposing who actually overrides whom,
making tacit legitimacy computationally explicit.

Helen realizes the danger first:
the AI is dissolving institutional ambiguity.

And ambiguity is not merely inefficiency.
It is part of how legitimacy functions.

Her horror is:
the machine is technically obeying policy while socially destabilizing governance.

Lucy Suchman

“The workflow was never the work.”

Suchman’s foundational insight is that real human action is situated.

Organizations do not run on plans alone.
They run on:

improvisation,
repair work,
timing,
tacit understanding,
social calibration,
and contextual judgment.

The AI keeps failing in funny ways because it assumes:
the documented workflow is real.

Lucy watches:

nurses reinterpret discharge states,
schedulers quietly reroute around difficult physicians,
transport coordinators absorb institutional chaos,
informal authority override formal authority constantly.

Then she accidentally causes the crisis.

Trying to improve outcomes, she feeds the AI observational success metrics:

who resolves bottlenecks,
who prevents errors,
whose interventions stick.

The AI simply operationalizes what already works.

And suddenly the hospital’s real authority structure becomes visible.

Lucy’s realization:
the institution depended on tacit competence remaining tacit.

Susan Leigh Star

“The dropdown menu is now a constitution.”

Star’s work revealed that classifications and standards are not neutral descriptions.

They are infrastructure.

Invisible infrastructure becomes especially powerful because people stop seeing it.

At first Susan thinks the project is merely a metadata harmonization problem.

Then she notices:
the AI is no longer using classifications.

It is governing through them.

Definitions like:

“ready for discharge”
“high priority”
“operational escalation”
“care coordination”

become executable authority structures.

The categories stop describing institutional reality.

They begin producing it.

Susan sees the deepest layer of GACI:
once context becomes operational infrastructure,
classification becomes governance.

Why They Matter Together

Individually:

Helen sees contextual legitimacy,
Lucy sees situated practice,
Susan sees infrastructural power.

Together they realize:

AI systems cannot function inside tacit bureaucracy.
Therefore institutions must render context computationally explicit.
Which means governance itself becomes technical infrastructure.

That is GACI.

Or, as the Southie AV guy eventually puts it:

“So the computer’s basically forcing the peacocks to write down how the hospital actually works?”

Yes.

Exactly.

Appendix II: PICOC, Peacocks, and Peacock Alley

The acronym came first.

The pun came immediately after.

Only later did we realize Boston had already built the set.

PICOC

Power Is Control Over Context.

At first glance, this sounds like a modern AI slogan.

It is not.

It is an old political truth made technically unavoidable.

For centuries, institutions exercised power by controlling context:

what counted as evidence,
who entered the room,
who received the memo,
whose interpretation became official,
what categories existed,
which exceptions were remembered,
and which were quietly forgotten.

AI does not invent this.

It merely refuses to operate until enough of it has been made explicit.

Peacocks

Peacocks are famous for displays of status.

Hospitals have them.
Universities have them.
Governments have them.
Boston has several per square block.

Titles. Committees. Named lectures. Reserved seating. Keynote speeches. Executive summaries. Portraits of former deans watching the current dean make exactly the same mistake.

These things are not necessarily fake.

But they are often only loosely coupled to operational authority.

The institution works because everyone tacitly understands the difference between the person who represents the system and the person who can make it function.

Peacock Alley

Peacock Alley is where the distinction becomes visible.

It is the place where symbolic power circulates:

donors meet executives,
trustees meet physicians,
consultants meet everyone,
and nobody admits they are checking who else is in the room.
The hospital’s formal hierarchy is on the org chart.
Its social hierarchy is in Peacock Alley.
Its operational hierarchy is back at Longwood, where a charge nurse is fixing the consequences.

For years, these layers coexist because nobody tries to reconcile them too precisely.

Then the AI arrives.

The Collision

The AI does not understand prestige.

It understands schedules, outcomes, assignments, edits, approvals, and response times.

Asked to improve patient flow, it discovers:

the charge nurse everyone trusts,
the scheduler everyone waits for,
the transport coordinator who quietly fixes everything,
the administrator everyone copies but nobody consults,
and the executive whose highest-value contribution is attending donor dinners.

Nothing personal.

Just context.

The machine begins assigning authority accordingly.

The peacocks become very uncomfortable.

The Peacock Alley Problem

At Peacock Alley, authority is performed.

In the context system, authority is inferred.

That is the destabilization.

The CEO may be the most important person in the ballroom and the least important person in the discharge workflow.

Everyone already knows this.

The AI merely makes it schedulable.

So the CEO’s working lunch disappears.

A donor dinner appears in its place.

The assistant explains:

“It said the food there was much better and less GACI.”

The CEO smiles.

“Aww, how sweet.”

The institution survives.

Not by defeating the AI, but by allowing it to separate ceremonial authority from operational authority while preserving everyone’s dignity.

The Boston Corollary

Boston may be the natural habitat of the institutional peacock.

Centuries of accumulated prestige coexist with astonishing practical competence.

Usually this works beautifully.

The peacocks represent the institution.
The operators repair it.
The bureaucracy keeps the two descriptions politely apart.

AI ruins the arrangement by asking which context should actually control the next action.

The PICOC Principle

Power was never simply the ability to issue orders.

It was the ability to shape the context within which those orders were interpreted, accepted, delayed, revised, or ignored.

That is why Peacock Alley matters.

It is not merely a room full of status display.

It is a context-production system.

Who is invited, who is introduced, who is seated near whom, who receives five minutes, and who receives forty-five: these are governance operations disguised as hospitality.

AI makes the same operations explicit elsewhere:

who may write context,
who may read it,
who may invoke it,
who may override it,
and who may audit the result.

Why the Joke Matters

Everyone laughs because “PICOC” sounds like “peacock.”

Everyone laughs harder because “GACI” sounds like “gassy.”

Then, usually about five minutes later, they realize the joke has quietly inverted.

Peacock Alley was never merely theater.

It was governance through context.

The AI did not destroy that system.

It exposed the mechanism, encoded it, and–when properly instructed–booked it for dinner.

Appendix III: GACI Was Always There

The first reaction to GACI is usually:

“This is about AI governance.”

It isn’t.

AI merely makes GACI impossible to ignore.

Before AI, governance was largely a human technology.

People carried context in their heads.

They knew:

when a rule was absolute,
when it was advisory,
who could grant exceptions,
which policies mattered,
who really understood the problem,
and when to pretend not to notice.

Most of this was never written down.

It couldn’t be.

The institution functioned because people continuously reconstructed context for one another.

AI changes only one thing.

It cannot reconstruct context from institutional folklore.

It must be given context explicitly.

Or it guesses.

Usually badly.

This creates an uncomfortable choice.

Either:

make context explicit enough for machines to operate,

accept that machines remain assistants rather than participants.

There is no third option.

The irony is that every attempt to “govern AI” eventually becomes an attempt to govern context.

Which documents are authoritative?
Which memories are relevant?
Which exceptions should be remembered?
Which people may modify the context?
Who approves the modifications?
Who can see the audit trail?

The governance questions stay the same.

Only the substrate changes.

This is why GACI feels simultaneously revolutionary and obvious.

It does not replace governance.

It reveals its implementation.

Or, as Nurse Delgado eventually explains to the residents:

“The AI doesn’t know who’s in charge.

It only knows who keeps fixing your mistakes.

The awkward part is that now everyone else knows too.”

Appendix IV: Tom Lehrer‘s PICOC Song

“Any resemblance to actual hospitals, universities, governments, or consulting firms is entirely contextual.”

The PICOC of Peacock Alley

Verse 1

At a conference in Boston
On governance design,
They served sustainable shellfish
And a rather passable wine.
There were deans and vice provosts,
Consultants by the score,
And a startup with a platform
Guaranteed to flatten more.

They’d invented clever software
To reduce administrative load,
By digesting institutional context
Into executable code.
It optimized the schedules,
It rewrote every memo, too,
And somehow every document
Became both shorter and more true.

The nurses loved the routing,
The transport team agreed,
The residents complained because
It made them stop and read.
The lawyers praised the audit trail,
The engineers the graphs,
While everyone in management
Pretended not to laugh.

Chorus

PICOC! PICOC!
Power Is Control Over Context!
Every tacit workaround
The algorithm now inspects!

GACI! GACI!
Governance As Context Infrastructure!
We’ve formalized the bureaucracy
In metadata nomenclature!

Verse 2

The system found that meetings
Could be cut by thirty-three,
And somehow every committee
Lost one vice chair and three VPs.
It quietly rearranged them
According to observed effect–
A process management described as
“Strategically indirect.”

It shortened every policy
From twenty pages down to four.
The staff all said, “At last!” while Legal
Quietly restored twenty more.
It learned that “urgent” wasn’t,
That “optional” meant “required,”
And that “we’ll revisit this next quarter”
Meant the sponsor had retired.

Then one unforgettable morning
It produced a modest chart
Showing who actually made decisions
Versus who looked good at the start.

The silence was impressive.

Someone coughed.

Someone updated LinkedIn.

Someone asked for a working group.

Final Chorus

PICOC! PICOC!
Power Is Control Over Context!
Every title, every meeting,
Needs a little more inspection.

GACI! GACI!
Governance As Context Infrastructure!
The peacocks keep the donors happy–
The context keeps the hospital alive.

Tag

And should you doubt this theorem,
Take a stroll through Peacock Alley.
The speeches happen in the ballroom.

The governance happens in the hallway…