A Hospice Bot Made Dying Sound Like an Order. She Stepped In.

Lina started the notebook during the first week of her grandfather’s hospice care. On the opening page, beneath April 2024, she drew two columns and labeled them “bot was enough” and “call me.” She expected to record a few practical examples for her mother, who spoke limited English, and the visiting nurses, who spoke limited Spanish.

Within nine weeks, the notebook held 37 entries.

The family had begun using a translation bot before hospice. Her grandfather could speak into the phone when he needed help with an appointment or wanted to understand a letter. Its sentences were imperfect, but everyone usually found the meaning. During hospice visits, the same speed felt necessary.

A nurse could request a medical interpreter, yet that introduced a wait, while the phone was already beside the medication chart.

At first, it handled small exchanges well enough. Her grandfather described where something hurt. A nurse explained that a delivery would arrive later. Lina’s aunt asked whether he wanted soup.

Those entries went in the left column.

Then a nurse tried to explain that his body might be getting tired and that the family could notice him sleeping more. The bot stripped away the uncertainty. Its translation sounded like an instruction to let the body stop, followed by a blunt statement that death was near.

Her grandfather pulled the blanket higher and looked at Lina’s mother. Lina’s aunt thought the nurse was asking the family to stop his medication. The nurse, hearing the alarm in their voices but not understanding the words, looked back at the phone.

Lina translated the sentence again. She kept the nurse’s uncertainty and explained that the changes might happen, rather than telling her grandfather they must. Nothing about his care changed during that visit, but the room did. Her mother stopped placing the phone between them and held it at her side.

In the notebook, Lina wrote a short description of the error under “call me.” She added that the bot had turned possibility into command.

Hospice workers often used indirect language to give the family space to absorb what was happening. They spoke of comfort, changes and the possibility that someone was getting closer to the end. The bot favored direct equivalents, and when there was no neat match, it sometimes dropped the softer verb or the phrase that showed uncertainty. What sounded measured in English could arrive in Spanish as a decision already made.

Lina understood both versions, which meant she also became responsible for choosing between them. Her mother wanted the nurse’s meaning softened enough that her grandfather would not feel abandoned. Her aunt worried that softening kept him from understanding his own condition. Her grandfather sometimes asked Lina to stop explaining around the subject and say the plain thing.

She had not expected translation to make her the judge of how much truth another person could bear at once.

The nurses began looking toward her before longer conversations. Lina lived 40 minutes away and worked from home, but her availability was less open than it appeared. Her phone stayed beside her laptop. During one week, she joined four visits by video while answering messages from relatives between work calls.

Her mother apologized each time. The apologies made Lina feel worse, not better.

The bot remained useful. Her grandfather preferred it when he wanted water, needed another pillow or was trying to tell a nurse that a sound from the medical equipment bothered him. He could address the phone without waiting for the family to decide who should speak for him. On those occasions, the tool gave him back a small measure of privacy.

Medicine created a different problem. During one visit, a direction about taking a comfort medication only when a particular symptom appeared came through as though the dose should be taken routinely. The nurse noticed the family’s confusion and stopped the exchange. A medical interpreter later restated the direction, while Lina wrote “medicine: human” across both columns of the notebook.

No one was harmed. Still, the near mistake changed how the nurses used the phone. They continued to accept it for immediate, low-stakes needs, but conversations about medication or a change in care moved to an interpreter whenever one was available. If the family could not wait, Lina translated and asked the nurse to repeat the main point so she could check what she had understood.

That arrangement protected one kind of safety by placing more pressure on a relationship. Her cousins began messaging Lina before visits to make sure she would be reachable. One relative accused her of making the process slower after she asked them to pause a care discussion until an interpreter joined. Lina knew the bot could produce a sentence in seconds.

She also knew that a fast sentence was not always the sentence anyone had meant.

Privacy sat underneath those choices. To translate, the phone had to capture a dying man’s voice and the details of his care. The family did not know how long the app kept recordings or whether its text was reviewed elsewhere. Her grandfather understood that the phone translated him, but Lina was not sure he understood that his words might leave the room in another sense.

The nurses avoided entering identifying information into the tool. That boundary helped, though it did not remove the discomfort. Pain, fear and family disagreement could still be sensitive without a name attached. Lina noted the privacy concern near the back of the notebook, then drew a line beneath it because she had no resolution to record.

Her grandfather’s preferences were not consistent. He disliked the bot when it made dying sound like an order. On another afternoon, he used it to tell the nurse that the family kept answering for him. The translation was awkward, but the meaning reached everyone.

Lina’s mother moved her chair back and let him finish.

As he became sleepier, the family used fewer words around him. The phone stayed on the table, available for ordinary requests. For discussions about what his final days might look like, a nurse arranged an interpreter or waited for Lina. The delay could feel heavy, especially when relatives wanted certainty that no language could provide, yet it gave them time to notice whether her grandfather was awake and wanted to take part.

After his death, Lina found the notebook among the papers her mother had gathered from the house. Some entries seemed almost trivial: water, blanket, equipment noise. Others had thick lines beneath them. On the last used page, she had written that the bot was fine for a family joke, then placed talk about dying in the other column.

Questions people ask

Can a translation bot handle conversations during hospice care?

In this family, the bot handled ordinary requests and short exchanges that could be checked through gestures or context. It became less reliable when nurses discussed medication, uncertainty or dying, because small changes in tone and grammar altered what relatives believed they were being asked to decide.

Why did gentle language about dying translate so badly?

The nurses used phrases that left room for possibility and gave the family time to respond. The bot sometimes replaced that uncertainty with direct statements or instructions. Lina could recognize the shift because she understood both languages, but relatives relying on only one version could not see what had disappeared.

How did the family decide when to involve a human interpreter?

Their boundary developed after specific misunderstandings rather than through a formal rule. The phone remained available for immediate comfort requests, while medication and changes in care usually moved to an interpreter or Lina. That reduced some risk, though it also made her the relative everyone expected to remain reachable.

Did using the translation app create privacy concerns?

The family knew the phone had to capture speech, but they did not have a clear picture of where the words went afterward or how long they remained available. Nurses kept identifying details out of the app when possible. Lina recorded the unresolved concern in the notebook, which now sits in her kitchen drawer.