How do robots know that a person has been well-fed? The setting is set in the near future where helper robots have become commonplace and are often assigned to help around households, hospitals, retirement homes and animal shelters. Despite the advancements in AI and robotics these machines aren't sentient and operate purely on a set of rules and guidelines. For a lack of a better word, they lack 'common sense'. A robot tasked with making meals and feeding a person won't distinguish between a well-fed and a starving person unless it was designed to do so. Considering the 5 basic human needs (food, water, air, and shelter), the ability to tell a person's nutritional status should be there by default.
The frame challenge for this particular question is that the average robot lacks the specialized medical equipment to make these quick assessments. Short of pricking everyone's hand every few hours, the robots have no way to tell what's going on inside the person. Not only would equipping every single household robot with blood monitors be prohibitively expensive but it would also add an additional fear factor to the already stigmatised robots (fear of needles + fear of robots). Robots need to evoke a sense of well-being, so a method of scanning a human without them knowing would be ideal.
Common examples of when this would be useful: What if that person has just eaten a meal but gets a sudden drop or spike in blood sugar? What if a human has a case of malabsorption due to an inflamed gut? What if their owner is anorectic and is only pretending to eat? What if a kid hasn't eaten their greens in a while?
Basically this would prevent humans from lying to the robots. "Oh! I'm sorry, I haven't eaten anything today" the human would say and the robot would retort with "my sensors indicate that you've been snacking again, Samantha".
Ideally the robots would make use of their already existing senses to evaluate a human's health. Those senses would include sight, hearing, touch as well as infrared and electromagnetism.