Sensors and AI in life with illness: assistance and the limits of trust

Epigraph

“A wise man, therefore, proportions his belief to the evidence.”

David Hume, An Enquiry Concerning Human Understanding, section X, part I, §87. Selby-Bigge's 1902 edition, Gutenberg 9662. The context concerns experience and testimony before a discussion of miracles; it is not a medical decision rule.

About this article

Imagine Nina living with a long-term condition. An app reminds her about an agreed appointment, a sensor transmits a particular reading and AI turns several notes into a confident account: “Your condition is stable.” Each function looks like part of one caring system. Yet a reminder does not establish that the appointment happened, a number does not explain all her experience and a coherent account is not necessarily a professional assessment. The differences can be especially hard to notice after the system has genuinely made some everyday tasks easier.

This article helps identify one technological task and the boundaries of trust in its output. It considers measurement, notification, interpretation, data access and someone able to respond. It selects no device, sets no medical thresholds and authorises no autonomous diagnosis. Trust need mean neither complete devotion nor total refusal. Hume and Aristotle help distinguish grounds for confidence, the ability to produce a result and the point of an action. Philosophy verifies no product's accuracy. It helps ask what became easier and what responsibility still needs actual arrangements.

I. One screen can conceal several different actions

A number, notification and suggestion may appear together while having different origins. A sensor might detect a particular quantity, software display a transmitted value and another component describe it in words. Transformations occur between stages. Consistent attractive presentation does not establish equally reliable components.

When speaking to a service or professional, name the function plainly: what does the device obtain, calculate or merely forward, and which action is it intended to support? Someone need not become an engineer to ask whether an on-screen phrase is a measurement or a program's explanation. If the answer is unknown, “smart” should not stand in for it.

Imagine Nina seeing a missing-entry notification. It can help her remember a task without establishing whether the corresponding real-world action was missed: perhaps it was not recorded. Equally, a “completed” mark does not establish how an action was performed. The article suggests no treatment change based on these marks. It separates an interface event from something that needs clarification with a person.

This remains relevant with extensive support needs. Someone else may keep a journal through an agreed arrangement, but their entry remains an entry and their opinion an opinion. Data must not entirely replace a nonspeaking adult. Their accessible response and preferences have a significance a graph need not express.

II. Verify the particular promised function

Success at one indicator does not establish the ability to measure another. A medical label, popular brand or similar-looking case does not specify every capability. Examine the particular product, intended use, version and conditions rather than technology's reputation in general.

An instructive example is the FDA communication issued on 21 February 2024, checked again for this article. It warns about watches and rings claiming to measure glucose independently without piercing skin and distinguishes them from apps displaying readings from appropriate measuring devices. This is a particular US warning, not rejection of all wearable medical equipment or a worldwide authorisation list.

The distinction matters: identical-looking numbers can result from different processes. Where care depends on accurate measurement, discuss a suitable device with the health care team. This article proposes neither testing products through risks to one's own health nor changing medication doses. A seller's successful demonstration is not a clinical evaluation.

Not every uncertainty calls for abandoning everything useful. A verified everyday function may remain useful while an unverified promise is not relied on. An appointment reminder and an assessment do not receive joint approval because they share an app. This is more precise trust: it concerns a particular task and its grounds, rather than an entire object at once.

III. AI can help frame a question and still get an answer wrong

A clear summary is convenient when someone struggles to assemble notes for a professional conversation. But brevity, warmth and confidence do not verify facts. If source information is incomplete or mixed, a smooth narrative can conceal precisely what should have remained unknown.

A WHO release about generative models in health care identifies risks of false, inaccurate, biased and incomplete statements, alongside excessive reliance on automated output. The 18 January 2024 release was read; no full reading of its underlying guidance or present-day error rate for any service is claimed.

Imagine Nina asking for questions based on her notes. Distinguish literal extraction, added supposition and medical conclusion. “You wrote about these two episodes” differs from “This means the disease is stable.” A plausible-looking citation also needs checking for existence and support for the particular claim.

If verification is unavailable, a consequential medical answer should not be called reliable merely for reassurance. It can become a question to an appropriate professional, or its unsupported conclusion can be removed. This does not make every patient responsible for independently reviewing every study. A health-related system does not shed responsibility for its claims by appending “Check everything yourself”.

IV. Naming a recipient does not create assistance

Imagine an app sending Nina's sister a message. She checks her phone between work meetings but has not accepted round-the-clock duty and cannot immediately travel. A contact created a delivery address, not the whole response. Notification, reading, responding and the resulting help are different events.

Before relying on a proposed chain, clarify who accepted involvement, when, which information they receive and what they can do. A professional service has conditions too: not every subscription includes attendance and not every automatic message reaches a clinician. Check the actual service's current documents and answers; no particular service is assessed here.

An absent alert must not automatically become evidence of wellbeing. This article does not identify the cause of every silent system; that depends on the device and situation. Do not reassure someone with an unverified promise of continuous monitoring. Nor should a dangerous event be staged at home to test whether help works.

A backup route concerns the organisation of real assistance. Agree clinical responses to deterioration with the health care team. Immediate danger requires appropriate urgent help instead of waiting for an app icon. If an intended participant can no longer respond, that changes the arrangement substantially. A newer phone does not itself create another person, free time or an available service.

V. More data does not always mean more care

A reading, location trail, voice recording and view into a home provide different visibility. Agreement to report one event does not automatically explain permanent access to daily life. Before discussing monitoring convenience, identify the data, recipients, task and retention.

Imagine Nina agreeing to show a particular record to a professional but declining a shared family dashboard. That does not prove rejection of assistance. Another person may knowingly choose wider support. Neither choice follows automatically from diagnosis, age or kinship. The method of response must be accessible; communication difficulty does not turn silence into indefinite permission.

Ask about actual service conditions: possible actions, how access ends and what happens to information already shared. A helper's promise to look only when necessary differs from technical access restrictions. Ending future access does not establish the disappearance of previous copies. The article establishes no universal retention period or legal ground for processing.

Where someone depends on assistance and fears refusing, “Do you agree?” may not reveal their wishes. Separate necessary care from optional expansion of monitoring and seek appropriate support when pressure is involved. Declining additional visibility should not mean earning isolation. At the same time, a family agreement neither verifies a whole system's security nor replaces a competent answer to access questions.

VI. Hume and Aristotle ask trust different questions

Hume suggests proportioning belief to evidence. Here that helps distinguish a plausible account from verified grounds. His context, however, is a philosophical argument about experience and testimony before miracles, not a clinical scale or a demand that patients calculate device error probabilities.

In book VI of the Nicomachean Ethics, Aristotle distinguishes making from acting and later cleverness from practical wisdom. Reaching a chosen outcome successfully does not settle whether the purpose is good. This exposes why reliable information transmission does not, by itself, authorise a system to define someone's whole life.

Imagine a perfectly functioning dashboard showing a family every movement. Accuracy might be established while the desired scope of monitoring remains unsettled. Conversely, a respectful intention does not correct an inaccurate reading. The epistemic question—what deserves belief—and the ethical question—what should be done and for what purpose—connect without swallowing one another.

Aristotelian practical wisdom is not contemporary professional licensing, and ancient ethics does not inspect software. The comparison provides language rather than certification. We can value a useful capability without worshipping a system, and discuss care without making illness a test of virtue.

VII. Reconsider the part that actually changed

An arrangement can become outdated even when the object remains on a wrist. A function is added, another recipient appears, processing changes or a helper becomes unavailable. Someone's entire life need not be reassessed each time; understand the consequential change and the decisions depending on it.

Imagine an update offering automatic commentary on readings. Prior agreement to reminders does not select this interpretation. Clarify its purpose and limits without declaring all previous functions harmful. An unconfirmed capability remains unconfirmed. This article does not instruct anyone to disable a necessary medical system without an agreed professional route.

What helps preserve an arrangement is not an endless surveillance archive but understandable meaning: this task, this recipient, this verified part and this unknown. It can be discussed orally with support. People need not maintain complex spreadsheets for their boundaries to matter. A personal note nevertheless does not replace service documentation or verify actual implementation.

An alert that once helped is significant experience, not a guarantee for the next event. When a function fails, establish its scope and consequences rather than changing everything randomly. Consequential medical uncertainty requires appropriate assistance. Unverified changes do not become safe simply by being called personalisation.

VIII. Technology should leave room for an ordinary day

Life with illness need not become a continuous report of correct behaviour. Benefit may mean fewer repeated explanations, accessible conversation or room to do something enjoyable. Not every positive outcome is more measurement, and not every pause means neglecting oneself.

Imagine Nina finding an agreed-appointment reminder useful while declining an automatic daily judgement of her quality of life. Another chosen function may support medical care under specific conditions. These decisions should not be merged. Discuss a particular medical purpose professionally while retaining a human voice in the rest of life.

Without money, family or a usable app, the next step may be one accessible question to a service or professional about the actual task and available support. A purchase is not compulsory. An everyday reminder cannot be promised as a substitute for missing health care; small assistance need not be dismissed because it cannot solve everything.

The opening story may end modestly: Nina understands what a reminder does, who receives a message and which AI conclusion remains unverified. Trust has become narrower and clearer. It no longer requires believing the screen about everything. Technology can assist a particular part of life without becoming its sole narrator or deciding what its owner deserves.

What this means in practice

Choose one function in use or under consideration. Name the task and distinguish measurement, forwarding, program interpretation and a professional decision. No purchase or activation is needed for this reflection.

For the actual system, clarify data recipients, an accepted response role and how access can be reviewed. If writing is inaccessible, one oral question or a supported response can begin. An unconfirmed participant does not become on duty through a saved contact alone.

Do not change treatment, medical thresholds or device settings based on this article. Discuss a consequential medical conclusion and device suitability with a professional. Immediate danger does not wait for an alert or exercise. A useful outcome may simply be retaining a helpful function, naming a limit and preparing one question instead of promising complete protection.

Key quotation

“Practical wisdom is not the faculty, but it does not exist without this faculty.”

Aristotle, Nicomachean Ethics VI.12, W. D. Ross translation. Source. “The faculty” refers to the neighbouring discussion of cleverness in reaching an intended goal. The sentence distinguishes successful execution from good choice; it automatically attributes wisdom to neither machine nor professional.

Connections between traditions

Comparison	What it distinguishes	Limit
Hume §87 and Aristotle VI.12	Evidence sufficiency and a purpose's quality are different questions	Modern epistemology and ancient ethics constitute neither one doctrine nor a device standard
Aristotle VI.4–5 and an automatic dashboard	A successfully produced output does not settle good action	The analogy establishes neither AI consciousness nor clinical authority
Hume and a contemporary FDA warning	Trust concerns a particular claim and its grounds	The philosopher did not predict the device; the US warning has its own narrow scope

The map supports discussion of knowledge, execution and care without converting philosophy into medical evidence.

Primary sources and research

1. David Hume. An Enquiry Concerning Human Understanding, L. A. Selby-Bigge, second edition 1902, section X, part I, §§87–88. Selected paragraphs on confidence and testimony, not the whole work.
2. Aristotle. Nicomachean Ethics, Book VI, W. D. Ross translation, VI.4–5 and VI.12, including the cleverness context. Philosophical primary text, not clinical advice.
3. FDA. Do Not Use Smartwatches or Smart Rings to Measure Blood Glucose Levels, 21 February 2024; main warning and display/independent-measurement distinction. Not a worldwide catalogue of all devices.
4. WHO. AI ethics and governance guidance for large multi-modal models, 18 January 2024; news release, especially Potential benefits and risks and distributed responsibilities. The full guidance was not read or reproduced. Selected pages/passages checked on 24 October 2026. No individual device, version, account or service was tested.

Tags

sensor, artificial intelligence, illness, disability, adult, measurement, notification, interpretation, decision, trust, evidence, data, privacy, consent, support, recipient, assistance, accessibility, responsibility, uncertainty, update, limited resources, ordinary life, Hume, Aristotle, epistemology, practical wisdom, technology

About the authors

Andrey Zarinsh is the author and founder of United Wisdom. He explores inner transformation, the unity of traditions, intercultural dialogue and contemporary spiritual discipline. Instagram: @andrey.zarinsh.

Maria Evilkut is a co-author of United Wisdom. Alongside research, she seeks to test these ideas in practice and observe how principles of unity and mutual understanding work in real life between people and cultures. Instagram: @ilkutki.