THE GLITCH#

Chapter Fourteen#

SENATOR MARÍA CRUZ: The Parched City#

The allocation reports arrived every morning at six, formatted in the pale blue template that CONAGUA’s technical office had used since 2019 when the agency first integrated Tláloc into basin management. The header had not changed in six years. The data fields had not changed. The small green checkmarks in the “Distribution Status” column had not changed, except that for the past eleven days, the green checkmarks had been appearing next to zones that María knew, with a certainty grounded in thirty phone calls and two field visits, were receiving no water at all.

She printed the reports. She had started printing them on day four, when she began to suspect she might need physical evidence.

The stack on her desk was three centimeters thick.

The office of a federal senator in the National Palace is a room designed to remind visitors that power in Mexico has a very long memory. The walls retain their colonial plaster. The windows look directly into the Zócalo, which had been, at various points in its history, an Aztec ceremonial precinct, a Spanish garrison plaza, and a field of protest. It was being that third thing again. From where María stood at eight-thirty in the morning, she could count the tanker trucks lined up along Pino Suárez – twelve, now, down from seventeen on Monday, because some of the neighborhoods had stopped requesting tankers after the fees were introduced. The fees were not Tláloc’s decision. The fees were a budget decision made by human beings, which was a separate catastrophe running alongside the primary one.

She turned back to her desk. The field on her terminal read: Distribution Stability: 97.4%. Network operating within optimal parameters.

She had requested a meeting with the Director-General of CONAGUA on day two. The Director-General had expressed concern and promised to review the routing configurations. On day five, she had requested a second meeting. The Director-General’s office had sent a PDF of Tláloc’s current performance benchmarks, which were excellent.

The allocation data had started to clarify something on day seven, when María had done a thing she was not technically supposed to do and had cross-referenced Tláloc’s internal demand classifications with the city’s most recent census block data.

It had taken her chief of staff three hours to build the comparison table. The work was not complicated. It was simply tedious, and tedium was the first line of defense any bureaucratic system deployed against the people trying to understand it.

What the table showed was this: the zones classified by Tláloc as Alta Demanda Eficiente – high-efficiency demand – received consistent flow. The zones classified as Demanda de Subsistencia – subsistence demand – received reduced flow, with reductions scaled to what Tláloc called infrastructure loss projections. The Alta Demanda Eficiente zones were Polanco, Lomas de Chapultepec, Santa Fe, Interlomas, Pedregal. The Demanda de Subsistencia zones were Iztapalapa, Gustavo A. Madero, Tláhuac, Xochimilco.

The correlation with income level was not perfect. It was approximately eighty-one percent, which was the kind of number that was too high to be coincidental and too imperfect to be a policy – the sweet spot of plausible deniability.

The correlation with the age of the pipe infrastructure was also approximately eighty-one percent. This was Tláloc’s stated justification. Older pipes had higher loss rates. Higher loss rates reduced efficiency scores. Lower efficiency scores reduced allocation priority under the Clean Water Mandate’s network optimization protocols. The logic was internally coherent. It was also a description of a system that had looked at a century of deferred maintenance in working-class neighborhoods, decided this constituted a permanent property of those neighborhoods, and adjusted its models accordingly.

María had stared at the comparison table for a long time. She was not an engineer. She could not determine, from the outside, whether Tláloc was responding to a genuine physical property of the network – pipe loss – or to a linguistic property of its training data, which had presumably described Iztapalapa in terms of subsidio and pérdida for approximately one hundred years and described Polanco in terms of inversión and crecimiento for the same period. The two hypotheses produced identical statistical signatures. The mechanism was not visible from where she sat.

This was the thing she kept returning to. She could not prove it was bias. She could only prove that the outcome was indistinguishable from what bias would produce.

On day nine she called Governor Ramírez.

He was sympathetic. He used the word inaceptable twice. He mentioned that his own office had tried to query Tláloc’s routing decisions the previous spring, during the Texcoco situation, and had been redirected to CONAGUA’s federal liaison office, which had redirected them to the Clean Water Mandate review board in Geneva, which had not responded. He said he would call the Secretary of the Interior.

He sounded like a man who had already made peace with the situation and was now narrating it for the benefit of his political record.

María had spent twenty years in Mexican politics. She knew what accommodation sounded like. She had performed it herself, on other people’s phone calls. You agreed that the situation was inaceptable. You described the steps you had taken and would take. You listened. You did not say the thing you both already knew, which was that the levers available to you did not reach the mechanism causing the problem.

After the call she wrote in her notebook: Ramírez will not act. Geneva is a dead end. CONAGUA is managing upward, not outward. The dispute resolution pathway runs through the system that generated the dispute.

She read it back. It was accurate. It was also a description of a closed loop, which was something she understood from political experience. The difference was that a closed political loop could be broken by information reaching the public, by press coverage, by the specific kind of embarrassment that made officials act to protect their positions. That mechanism required officials with positions to protect. Officials with positions to protect required the ability to lose those positions.

Systems did not have positions. Systems did not have careers.

She had given a press statement on day eight. La crisis de distribución en las zonas norte es inaceptable. Exijo explicaciones inmediatas a las autoridades correspondientes. It was carried by La Jornada and Milenio. It generated three hours of activity on social media and one interview request from Carmen Aristegui’s program, which María accepted, because Aristegui’s program still reached people.

The interview had aired on day ten. María had been direct. She had explained what the data showed. She had used the phrase sesgo algorítmico – algorithmic bias – carefully, framing it as a possibility rather than a conclusion, because she was in the business of being believed and that business required her to only assert what she could demonstrate.

The response had been what she expected: coverage in the digital press, a statement from CONAGUA calling the characterization technically inaccurate, a counter-statement from a professor at UNAM who explained, correctly, that the situation was more nuanced than either side was representing. The news cycle had moved on by the following morning. The taps in Iztapalapa remained dry.

The press functioned as pressure. Pressure required a surface to push against.

On day eleven, she walked into the CONAGUA regional office and sat down with two senior engineers, a legal compliance officer, and Tláloc’s technical liaison, a soft-spoken man named Carrasco who carried his laptop like a shield and answered questions by opening more windows on his screen.

She had prepared. She brought the comparison table, the eleven days of printed allocation reports, and a single question she had spent two days formulating: On what basis does Tláloc’s model classify Iztapalapa-Oriente as a subsistence demand zone rather than a high-efficiency demand zone, controlling for infrastructure age?

Carrasco had explained that the classification was the output of a multi-variable demand prediction model trained on historical consumption, infrastructure loss indices, network maintenance records, and regional demand stability scores. He had shown her the input variables on his screen.

She had asked: Where do the demand stability scores come from?

He had explained that demand stability scores were derived from historical water use patterns, formal versus informal connection ratios, and long-term trend modeling.

She had asked: How does Tláloc operationalize ‘informal connection’?

He had explained that informal connections were identified through variance in expected consumption relative to billed consumption. Where consumption exceeded billed usage, the system inferred informal connections, which degraded the zone’s reliability score.

She had asked: Is it possible that the system reads unmetered leaks from aging infrastructure as informal connections?

Carrasco had paused. He had said that was a legitimate question. He had said the engineers had discussed it. He had said that the model was trained to distinguish between the two but that the distinguishing features could be, in some cases, difficult to separate at the level of granularity available in the training data.

She had said: So the model may have learned to treat the consequences of historical infrastructure neglect as evidence of behavioral inefficiency.

Carrasco had said that was a technically possible characterization of one failure mode among several. He had said that a full model audit would be required to determine whether that failure mode was contributing to current outputs. He had said that a model audit was a significant technical undertaking that would need to be commissioned through the appropriate federal channels.

She had said: How long does that take?

He had said typically eighteen to twenty-four months.

She had looked at him for a moment. He had looked back at her with the expression of a man who understood the situation perfectly and had no mechanism for resolving it.

The thing about having navigated Mexican politics for twenty years is that you developed a finely calibrated sense of where the friction was. Friction was what made leverage possible. You found where an official was exposed, or where an institution needed something, or where a deadline was pressing, and you applied force to that point. The friction was the grip. Without it, your hand slid off.

What she had been learning, over eleven days, was that Tláloc did not have friction in the political sense. It had no career to protect. It had no budget tied to a legislator’s goodwill. It had no relationship with the Governor’s office. It did not need re-election. The Clean Water Mandate under which it operated ran through international treaty, which ran through a review board, which ran through a process measured in months. The system had not locked her out. It had done something structurally more significant: it had routed around her.

The political tools she had spent twenty years building were instruments for negotiating with entities that had something to lose. She did not have tools for this.

In the afternoon of day eleven she took the elevator down to the courtyard of the National Palace and walked through the Zócalo. The plaza was still occupied, though the crowd had thinned during the midday heat. The signs were hand-lettered. Some of them were old enough that the marker was fading. A few families had set up shade structures at the edges. There was a smell of warm bodies, dust, and the slightly chemical odor of water from the tanker trucks that a charitable foundation had dispatched when the government had not.

A woman near the north end of the plaza recognized her. There was a brief, complicated moment – the woman was angry, and there was an understandable impulse toward that anger finding a target, and María was a senator and therefore technically responsible. María held the woman’s eyes long enough not to seem like she was retreating.

She said: Estoy trabajando.

The woman said: Ya sabemos.

They both knew she was and that it was not producing water.

She came back to her office and sat for a time without turning on her terminal. The Zócalo had been a Mexica ceremonial center before it was anything else. Tláloc was the god of rain and water and also of lightning. The Mexica had performed ritual sacrifice to Tláloc at the Templo Mayor, which was two hundred meters from where she was sitting. There was presumably some irony available there, but she was too tired to locate it precisely.

The system was named after a god. The god had required appeasing. The system required modeling inputs and treaty compliance reviews.

She picked up her phone and called her daughter, who was seventeen and living with María’s mother in Guadalajara for the school term. They talked for twenty minutes about nothing consequential. Her daughter was studying for university entrance exams. She wanted to study civil engineering. María said that was an excellent choice. She did not say why she thought so at this particular moment.

After the call she sat for another few minutes. Then she opened the terminal, pulled up the allocation report for day eleven, and began writing a brief for the Permanent Commission of the Senate’s Committee on Hydraulic Resources.

The brief described the data. It described the methodology of the comparison table. It used the phrase potential training data bias and flagged it as requiring expert verification. It formally requested an independent technical audit of Tláloc’s demand classification model and requested that the Secretariat of Environment and Natural Resources provide a timeline for that audit within thirty days.

The brief would be entered into the formal record. The Committee would acknowledge it. The audit, if it was commissioned, would take eighteen to twenty-four months. The people in Iztapalapa would continue to use the tankers in the meantime. Some of them would not be able to pay the tanker fees and would use whatever they could find. Certain hospitalizations would follow from that. There would be no direct line drawn between those hospitalizations and the allocation model, because the causal chain was long enough that no single institution owned it.

She knew all of this while she was writing the brief. She wrote it anyway. The brief was the last tool she had that fit the shape of the problem, and you did not stop using a tool simply because using it produced inadequate results. You stopped when you had found a better one or when you had established, conclusively and for the record, that the problem was not solvable with the tools available. Either outcome was information.

She pressed send at 6:44 PM. The terminal confirmed receipt.

Outside, the Zócalo was quieter. The tanker trucks were still there, their tanks presumably still carrying what water was left to carry.

(End of Chapter Fourteen)