Bedtime Stories for Public Servants: The Algorithm That Blinked

The views expressed are those of the author and do not necessarily reflect the views of ASPA as an organization.

By S. Mohsen Fatemi
August 18, 2025

The cubicle wall still held a thumb-tacked copy of the old Eligibility Rulebook, its pages yellowed at the corners but no one reached for it anymore. Three days ago, the state benefits agency had archived the binder and uploaded Model 8B, an artificial intelligence scoring system that scanned every Supplemental Nutrition Assistance Program application and spat out one of three verdicts: Green (process), Yellow (double-check), Red (fraud-risk—halt).

At 4:34 p.m. on rollout Day Three, Caseworker Jordan Kim stared at a pulsing red banner over Application #14577. The screen said the applicant’s address history did not match DMV records; confidence score 0.92. Mandatory hold.

Jordan exhaled, long and slow. She had talked to this client—Brianna Harrell, single parent, two jobs, recent eviction—just yesterday. The mismatch was a landlord typo on the move-out notice. The child support payments were real, the bills current, the fridge probably half-empty.

Across the floor, monitors glowed with the same new interface, a neat teal-and-gray grid that made years of procedural labyrinth look elegant. Speed was the slogan. Ten-day turnaround slashed to forty-eight hours—the governor would trumpet the metric at tomorrow’s press briefing. Neutral competence 2.0, the Innovation Office called it.

But every override of a Red score now triggered an audit trail and a “variance justification” form routed to headquarters. One paragraph wrong and the Performance Review month would be brutal. The new hierarchy was invisible yet absolute; instead of section chiefs wandering the aisles, a silent algorithm watched everything.

Jordan glanced at the clock—twenty-six minutes to closing. Her manager, Mr. Ramos, stood near the printers, face washed blue by dashboard analytics. He was scrolling the live adoption numbers, grinning like a man guarding good news. She imagined walking over to explain Brianna’s eviction paperwork. She also imagined Ramos’s eyebrows lifting at the words “manual override.”

Jordan opened the old binder anyway. Rule 12-c, Fraud Screening, read: Discrepancies in residency shall not by themselves constitute presumptive fraud; contextual verification may satisfy eligibility. The sentence was underlined in pencil from years ago. She snapped a photo with her phone, just in case. The room hummed with fans and quiet keystrokes, as if the whole agency were waiting to see who would blink first: the caseworker or the code.

She clicked Override. A dialog bloomed: Provide variance justification (1,000 characters max). Her fingers hovered over the keys.

Rule 12-c permits contextual verification where administrative data conflict. Landlord error acknowledged in supporting documents. No prior adverse findings. Manual override consistent with due-process equity.

She attached Brianna’s eviction notice, the binder photo, and her call log. Then she cc’d Ramos, the compliance unit, and the governor’s liaison. Total transparency—as bright as noon.

The red banner disappeared. Application #14577 turned Green.

A hush settled over Jordan’s headset. Somewhere, a performance metric ticked in the wrong direction. She closed her eyes, half-expecting a reprimand to ping before five.

Instead, an email popped from Ramos: “Thanks for logging the rationale. Let’s talk tomorrow—might be worth refining the model on address errors.” No exclamation points but no threat either.

Jordan powered down. The office smelled faintly of lemon from someone’s forgotten mug. She stepped into the elevator and whispered the rule to herself—just a flicker of reassurance—as the doors slid shut.

A Pause with the Civic Sage

The corridor just off the main floor carries a late-day stillness. Overhead lights cast a sterile glow on worn linoleum. A water cooler gurgles quietly, its tank fogged with condensation. A box of retired forms rests against the wall, the word “OBSOLETE” stamped in red along its side. Jordan stands near the cooler, phone in hand, Rule 12-c still open on the screen. The Civic Sage finishes slicing a lemon and drops a wedge into the reservoir.

Civic Sage (securing the lid with a soft click): Algorithms are mirrors, Jordan. They reflect what we’ve built into them—nothing more.
Jordan Kim: Mirrors don’t see hunger. Brianna’s kids need food, not pattern recognition.

The lemon slice floats in slow circles. Light catches the edge of the water tank, casting shifting shapes on the floor.

Jordan Kim: Rule 12-c lets me act. If the audit doesn’t say I shouldn’t have.
Civic Sage: Neutral competence isn’t about hiding behind systems. It’s about noticing when the system misfires and owning the repair.
Jordan Kim (rubbing the back of her neck): Every override paints a target. One more and I’m flagged.

A distant printer chirps then stops. Footsteps echo faintly down the hall.

Civic Sage: Then paint the target in daylight. Show your work. Make the record part of the repair.
Jordan Kim (quietly): Daylight over dread.
Civic Sage (offering her a paper cup of lemon water): Equity isn’t an exception. It’s the measure of whether the system deserves the public’s trust.

Loose Ends

Code is the newest civil servant and justice is its shy probation officer. Rules once lived on paper and wore signatures like wax seals; now they glide in silicon microseconds, cloaked in probability scores that few can read. Yet the canon of public administration endures. Neutral competence demands we interrogate the tool, not genuflect before it. Due process demands transparent footsteps—from data source to final stamp—so every rejected claim can appeal to daylight. And social equity, the third pillar, demands we trace each false positive back to the human it mislabels then mend the mirror that warped their reflection. Until an algorithm can testify, stewardship belongs to those with beating hearts and accountable pens—who override when fairness calls and log the reason so the public record sings.

---

Author: S. Mohsen Fatemi is a PhD candidate in the School of Public Affairs & Administration at the University of Kansas, where his research examines sustainable energy governance, policy, and justice. He is the creator of Bedtime Stories for Public Servants, a narrative series that bridges public administration theory and practice through storytelling. He can be reached at [email protected]. His website is www.mohsenfatemi.com, and his X/Twitter handle is @MohsenFatemiii.

(1 votes, average: 5.00 out of 5)
Loading...