Dignity in automated care

The strongest case for a cooking machine is not convenience. It is the person who wants to cook for themselves and cannot use a stove today: someone recovering from a stroke who wants their own food, a woman of ninety who has fed others all her life and would rather feed herself, a young adult whose kitchen was not built for them. For them a machine that cooks safely is not a luxury. It is the difference between being fed and eating.

That is also where the risk of indignity is highest. A machine can feed a person without asking them anything. It can decide what is good for them from a profile. It can serve a tray and leave. It can be the reason nobody else comes.

Cookwala has made a few choices here and left most of the hard questions open.

The person chooses; the machine checks. Nothing in the standard computes what a person should eat. Rule packs flag sodium or sugar on a menu and say where the rule came from; they do not decide the menu. Clinician-set targets are entered locally by a clinician; the machine never diagnoses. Weight goals are opt-in; the default is "eat well". A person who wants the salty dish they have eaten for seventy years is told the number and served the dish.

Texture and serving are part of the recipe, not an afterthought. A recipe can carry a texture level, a serving form and a serving temperature, so a meal for someone who swallows with difficulty is the same dish prepared properly, not a different, lesser meal.

Presence is a safety rule, not a social policy. Steps that cannot be verified by a sensor need a person present. We have resisted the temptation to make presence a requirement for other reasons, because a rule that says "a person must be there" for someone who lives alone is a rule that denies them the machine. The social question of who visits is not solved by a schema and should not be hidden inside one.

The hard case, stated plainly. The sensor ladder's last rung for sautéing, searing, frying and reducing is a person. The people this essay is written for are often the people who cannot be that person: they cannot stand at the pan, see the colour or smell the smoke. For them the current standard refuses exactly the steps that make a meal taste like home, and "will you watch?" is a question they cannot answer with yes. We do not have a rule that closes this gap. The honest options are better sensors on cheaper devices, a remote attendant who is reachable and accountable, or a recipe written to avoid those steps; each has a cost in taste, in money or in dignity. It is listed as an open problem, not solved by this essay.

No score of a person, ever. The household context profile forbids behavioural scoring. How fast someone eats, what they leave on the plate, whether they argue at dinner: a machine may notice these to serve them better and may never turn them into a number about them, and may never send them anywhere.

The cook's name stays on the plate. A machine that cooks a dish from a care home's own cook, or the resident's own recipe, serves something that belongs to someone. We think this matters more than it sounds.

What we have not answered: whether being fed by a machine, however well, is a loss that should be named rather than designed around; whether the people who do this work today should be protected from the machines or given them as tools; who speaks for the person who cannot answer the machine's question. The standard's governance reserves a seat for a labour or consumer voice; a seat for the people served, including disabled users, is proposed and not yet in the charter. That is a start, not an answer.