The Trojan Horse

TAM-RWR.ZPF-03 · The Reshaped World, The Zero-Person Frontier · The Approximate Mind

The moment Sandra Purcell understood what her organization actually does, she was not in the office. She was on the phone with a substitute volunteer who had just completed a Tuesday delivery to Mrs. Chen’s apartment on Balboa Street.

The substitute reported a successful delivery. Correct meal, correct temperature, correct dietary specifications, accepted at the door, signed for. Sandra thanked her and was about to hang up when Delores called from home. Delores had been the regular volunteer on the Balboa route for four years and was out with a back injury.

“Did anyone check on her? She sets out two cups on Tuesdays.”

Sandra called Mrs. Chen. It took six rings. Mrs. Chen had been crying. Not about anything specific, she said. About the accumulation of days.

Sandra is the program director for a Meals on Wheels chapter serving the western neighborhoods of a mid-sized city. She has been in the role for nine years. On her desk is a framed photograph of her daughter’s third-grade soccer team, twelve children in oversized jerseys squinting into the sun. The photograph has nothing to do with meal delivery. Sandra keeps it because her daughter is now twenty-three and works in data analytics in another state, and sometimes Sandra needs to look at something that reminds her that time is doing what it does to everyone.

The delivery manifest on her screen lists addresses, meal types, dietary restrictions, delivery windows, and volunteer assignments. It does not list: “Sets out two cups on Tuesdays.” “Puts the orchid on the table when she’s having a good day.” “Answers the door slowly when her hands are bad.”

Delores knew all of this. The manifest knew none of it.

What the Manifest Tracks

#

The formal logic of Meals on Wheels is nutritional delivery. A population of homebound older adults cannot reliably feed themselves. The program provides balanced meals, delivered on schedule, at no cost to the recipient. The metrics are delivery reliability, dietary compliance, and coverage: how many eligible recipients are being served and how many are on the waiting list.

By these metrics, the program works. In Sandra’s chapter, delivery reliability is above 96 percent. Dietary compliance is near-perfect because the meals are centrally prepared. The waiting list is long, which is a measure of demand rather than failure, but it means that every operational inefficiency, every late delivery, every route that takes longer than it should, is a meal that someone on the list does not receive today.

The case for autonomous delivery is made in the language of the manifest. Robots do not call in sick. They do not get back injuries. They do not take longer on some stops because they are talking to the recipient. They do not need parking. They operate in weather that keeps volunteers home. They can run routes at times that volunteers cannot cover. A fleet of delivery robots serving Sandra’s territory could, by the manifest’s metrics, increase coverage by 30 percent and eliminate the waiting list within two years.

Sandra has seen the pilot proposals. She has read the outcome projections. She does not dispute the numbers.

She disputes what the numbers are measuring.

What the Manifest Does Not Track

#

Delores has been delivering to Mrs. Chen’s apartment every Tuesday and Thursday for four years. In that time she has learned things about Mrs. Chen that no intake form captured and no algorithm would think to ask.

She knows that Mrs. Chen sets out two cups on Tuesdays because Tuesday was the day her husband used to come home early. He has been dead for six years. The second cup is not for Delores. It is for the habit of expecting someone.

She knows that when the orchid is on the table, Mrs. Chen is having a day where she wants to be seen as a person who keeps beautiful things. When the orchid is in the bedroom, she is not.

She knows that Mrs. Chen’s arthritis is worse in her right hand than her left, that the pill bottles are the main difficulty, that Mrs. Chen will not ask for help opening them but will accept it if offered without fuss. She knows that Mrs. Chen’s son calls on Sundays and that the calls are short and that Mrs. Chen spends Monday recovering from whatever the call contained.

None of this is in the manifest. None of it is in Mrs. Chen’s medical record. None of it is the kind of information that a system designed around nutritional delivery would think to collect, because it is not nutritional information. It is the residue of one person paying attention to another person over time, accumulating a portrait that has no institutional home and no operational function.

Except that it does have a function. Delores has flagged three concerns to Sandra in four years. One was a medication change that Mrs. Chen had not understood. One was a fall that Mrs. Chen had not reported. One was a period of two weeks when Mrs. Chen stopped setting out the cups and stopped putting the orchid on the table and answered the door in the same housecoat every visit, which Delores recognized as something different from Mrs. Chen’s ordinary days and which turned out to be a reaction to a new blood pressure medication that was making her foggy and flat.

In each case, the flag came from knowledge that no intake form would generate and no remote monitoring system would detect. The medication confusion was invisible to the pharmacy because Mrs. Chen confirmed she understood the instructions. The fall was invisible to anyone not present because Mrs. Chen did not call anyone. The medication side effect was invisible to the prescribing physician because Mrs. Chen reported feeling “fine,” which is what she reports about everything.

The meal was a Trojan horse. What it carried was presence.

The Trojan Horse Taxonomy

#

Mrs. Chen’s situation is specific. The pattern is not.

There is a category of service delivery in which the nominal function, the thing the system was designed to provide, was never the real function. The real function was human contact with people who might not otherwise have it, and the nominal function was the vehicle that delivered it.

School transportation is a Trojan horse. Ray has been driving the same bus route in a small city in the upper Midwest for twenty-three years. The route serves forty-one children across three schools. The nominal function is moving children from their homes to their classrooms and back. Ray does this. He also knows which children are quiet in a way that means something and which are quiet because that is who they are. He has called in three welfare checks in his career, and all three were warranted. He knew to call not because of any training protocol but because he had driven the same children for long enough to know what their ordinary looked like, and what he was seeing was not ordinary.

The autonomous school bus that will eventually replace Ray will be safer by several measurable standards. It will not get drowsy. It will not be distracted. It will not make the errors that human drivers make in traffic. It will deliver children to school reliably and return them home on schedule.

It will not notice the child who has stopped talking.

Library home delivery is a Trojan horse. In rural counties and small cities, library systems deliver books to homebound patrons. The nominal function is access to reading material. The delivery volunteers report that the reading material is often incidental. What the patron wanted was the conversation at the door, the ten minutes of being treated as a person with intellectual interests rather than a patient or a dependent. The book was the reason for the visit. The visit was the reason for the program.

Postal carrier routes in rural communities are a Trojan horse. The letter carrier is, in some rural geographies, the only person who comes to the door on a regular schedule. The nominal function is mail delivery. The carrier’s actual function, in communities where the nearest neighbor is two miles away and the nearest town is twenty, includes an informal welfare check that no one has formalized because no one needed to: the carrier noticed when the mail accumulated, when the dog was out but the person was not, when something had changed.

Community health worker visits are a Trojan horse. The CHW’s nominal function is health education, medication management, screening. The CHW’s actual function, in the communities where CHWs are most effective, is being the person who shows up. The clinical literature on CHW effectiveness consistently finds that the most important variable is not the health information delivered but the relationship between the CHW and the patient. The information is the vehicle. The relationship is the cargo.

The Measurement Problem

#

Every one of these services has outcome metrics. Meals on Wheels tracks delivery reliability and nutritional adequacy. School transportation tracks on-time arrivals and safety incidents. Library delivery tracks circulation numbers. Postal service tracks delivery times. Community health worker programs track screening rates and medication adherence.

In every case, the metrics track the horse, not what was inside it.

This is not an oversight in the way that a missing line item is an oversight. It is structural. The relational function that the human carried was invisible to the system before automation arrived, because the system was designed around the nominal function. Nobody designed Meals on Wheels to provide social contact. The social contact happened because a human being was delivering the meal, and human beings, when they see the same person twice a week for four years, start to know things about that person. The knowing was a byproduct of the logistics, not a feature of the program.

Byproducts do not generate metrics. They do not appear in budget justifications. They do not survive cost-benefit analyses because they were never on the benefit side of the ledger. When the cost-benefit analysis compares autonomous delivery to volunteer delivery, the costs are visible and the benefits are visible and the byproduct, the relational function, the knowing, the Trojan horse’s cargo, is not on either side of the equation.

The autonomous system improves the metrics. The metrics improve because they were designed to measure the nominal function. The nominal function was never what mattered most to the person at the door.

The Design Question

#

I keep returning to a question that the architecture of these programs makes difficult to ask. If the relational function was the real function, if the meal was a Trojan horse for presence, if the bus ride was a Trojan horse for noticing, if the book delivery was a Trojan horse for conversation, then should the system have been designed around the relational function rather than the nominal one?

The question sounds naive. Governments fund nutritional programs because nutrition is a measurable outcome. They do not fund “someone comes to your door and pays attention to you” because attention is not a line item in a budget. The political economy of social services runs on measurability: Congress funds what can be counted, evaluated, and reported. Presence cannot be counted. Attention cannot be evaluated. The Trojan horse architecture is not an accident. It is the only design that could have survived the funding environment.

Which means the relational function was always dependent on the nominal function for its delivery vehicle. And when the delivery vehicle is automated, the relational function has no carrier.

This is the specific problem the Trojan horse poses. Not that the system failed to value human contact. The system never saw human contact, because human contact was smuggled in through a logistics operation that happened to require human hands. When the logistics operation no longer requires human hands, the smuggling route closes. What was smuggled does not find an alternative route. It stops arriving.

I wonder whether a society that understood what the Trojan horse was carrying would have designed the system around the cargo rather than the vehicle, and whether we still could, and whether the funding environment that made the Trojan horse necessary in the first place has changed enough to make a direct approach possible, or whether presence still cannot be a line item.

What Arrives Instead

#

The robot arrives at Mrs. Chen’s door at 11:40 a.m. on Tuesday. The meal is correct. The temperature is correct. The dietary specifications are met. Mrs. Chen opens the door, takes the container, and closes the door. The interaction takes fourteen seconds. The system logs a successful delivery.

Mrs. Chen eats the meal. It is adequate. It is what she needs, nutritionally, and it is better than what she would prepare for herself, which is often nothing.

She has stopped setting out two cups.

This is not a dramatic change. Nobody will write a story about Mrs. Chen’s cups. The program’s metrics will not register their absence. Sandra will notice, if she visits, which she does less often now that the deliveries are automated and her operational workload has shifted from volunteer management to fleet maintenance contracts. Delores has moved to Sacramento to be near her grandchildren. She and Mrs. Chen exchanged addresses. Neither has written.

The pilot program’s six-month report will show improved delivery reliability, reduced cost per meal, elimination of weather-related service gaps, and expansion of coverage to 340 recipients previously on the waiting list. The 340 new recipients are receiving meals they were not receiving before. This is real. This matters. The waiting list represented people who were not eating adequately, and now they are.

The system is working better by every measure the system tracks.

The measure the system does not track is the one that would show what happened to Mrs. Chen on Tuesdays between the departure of Delores and the arrival of the robot. That interval is where the Trojan horse’s cargo used to be delivered. The interval is now empty. The meal still arrives. The presence does not.

The Soccer Team

#

Sandra’s photograph is still on her desk. Twelve children in oversized jerseys, squinting. Her daughter is the one on the far left, the smallest, the one whose jersey comes to her knees.

The photograph predates Sandra’s work at Meals on Wheels by six years. She was a different person when it was taken: younger, employed in logistics for a shipping company, thinking about efficiency and routes and on-time percentages, the same metrics she would later apply to a different kind of delivery.

She did not come to this work because of the metrics. She came because her mother was alone in the last years of her life, in a house in Daly City that was too big for one person, and the woman who delivered her mother’s meals twice a week was the only person who saw her mother on the days Sandra could not. The woman’s name was Graciela. Sandra’s mother called her “the one who comes.” Not by name. By function. The function was not delivery.

Sandra has been asked, in three separate meetings this year, whether the autonomous delivery program should be expanded to cover her remaining volunteer routes. The data supports expansion. The waiting list supports expansion. The budget supports expansion.

She has not said no. She has asked a question that the data cannot answer and that the meetings have not yet found a place for: what happens to the Mrs. Chens when the Gracielas stop coming?

The question is on her annotated version of the program review. It is not on the version that goes to the board.

The photograph of the soccer team is still on her desk. She does not look at it when she is thinking about delivery metrics. She looks at it when she is thinking about time, and what it does, and what it takes with it, and what gets left behind at the door.

References

#

Meals on Wheels and Social Isolation

Thomas, Kali S., and Vincent Mor. “Providing More Home-Delivered Meals Is One Way to Keep Older Adults with Low Care Needs Out of Nursing Homes.” Health Affairs, vol. 32, no. 10, 2013, pp. 1796–1802.

Sahyoun, Nadine R., and Rong Zhang. “Use of Meals-on-Wheels and Other Nutrition Programs by the Elderly.” Journal of Nutrition for the Elderly, vol. 25, no. 3-4, 2005, pp. 173–193.

National Foundation to End Senior Hunger. The State of Senior Hunger in America 2021: An Annual Report. Brandeis University, 2023.

Social Contact in Service Delivery

Holt-Lunstad, Julianne, et al. “Social Relationships and Mortality Risk: A Meta-analytic Review.” PLoS Medicine, vol. 7, no. 7, 2010, e1000316.

Courtin, Emilie, and Martin Knapp. “Social Isolation, Loneliness and Health in Old Age: A Scoping Review.” Health and Social Care in the Community, vol. 25, no. 3, 2017, pp. 799–812.

Perissinotto, Carla M., et al. “Loneliness in Older Persons: A Predictor of Functional Decline and Death.” Archives of Internal Medicine, vol. 172, no. 14, 2012, pp. 1078–1083.

Community Health Workers and Relational Effectiveness

Perry, Henry B., et al. “Community Health Workers in Low-, Middle-, and High-Income Countries: An Overview of Their History, Recent Evolution, and Current Effectiveness.” Annual Review of Public Health, vol. 35, 2014, pp. 399–421.

Kangovi, Shreya, et al. “Community Health Worker Support for Disadvantaged Patients with Multiple Chronic Diseases: A Randomized Clinical Trial.” American Journal of Public Health, vol. 107, no. 10, 2017, pp. 1660–1667.

Autonomous Delivery and Service Robotics

Locus Robotics and Industry Reporting. Trade press and operational analyses of autonomous last-mile delivery pilots, 2023–2025.

Hoffman, Donna L., and Thomas P. Novak. “Consumer and Object Experience in the Internet of Things: An Assemblage Theory Approach.” Journal of Consumer Research, vol. 44, no. 6, 2018, pp. 1178–1204.