Glasses for the Mind: How AI Can Give People with Dementia and Autism Years of Independence

The article explores the potential of AI assistants as cognitive prostheses for people with dementia and autism spectrum disorders — desktop devices and apps that remind users to take their medication, structure their day, and help with bureaucratic tasks. Against the backdrop of data on the loneliness epidemic (871,000 deaths annually according to the WHO) and an aging Czech population, it analyzes the Israeli ElliQ device and other solutions, along with questions of reimbursement by health insurance.

Imagine a seventy-year-old woman in the early stages of Alzheimer's. Just a year ago she cooked, paid her bills, went to the doctor on her own. Today she forgets to turn off the stove, doesn't know what day it is, and her daughter is starting to think about a retirement home. Between her present condition and the threshold of independent living lies a handful of specific abilities — remembering a daily routine, taking medication on time, finding her bearings on a calendar.

Imagine a thirty-year-old man with autism spectrum disorder. He can handle work at a computer, but he's at a loss when a plan changes unexpectedly, he can't manage a phone call to a government office, and shopping in a crowded store paralyzes him. Between his life in sheltered housing and a flat of his own stand a few specific skills that someone patient, tireless, and always at hand could prompt him through, step by step.

What if that "someone" weren't a person, but a machine?

Not a robot from science fiction. A desktop device with a pleasant voice that reminds you to take your pills in the morning, suggests a walk during the day, and calls your daughter in the evening. Or a phone app that walks through the procedure step by step before a visit to an office and advises in real time what to say. Technology that doesn't diagnose and doesn't treat — but compensates for what is missing. The way glasses compensate for a vision defect.

Glasses for the mind.

This idea sounds futuristic, but a number of the pieces already exist and work. At the same time, it comes with fundamental questions: Does it really help? Doesn't AI just serve as a crutch that kills off whatever abilities remain? And above all — should health insurance pay for it?

Before we get to the answers, we need to grasp the scale of the problem. In June 2025 the WHO Commission on Social Connection published a landmark report with a figure that shook the medical community: loneliness is associated with 871,000 deaths a year worldwide. A hundred lives every hour.

One in six people on the planet suffers from loneliness. Among seniors, roughly one in eight. And in the Czech Republic this is a population that will grow dramatically — from the current 2.2 million people over 65 to more than 3 million around 2040. According to an estimate by the Kooperativa Foundation, as many as half of Czech seniors live alone.

This is not just an unpleasant feeling. Large pooled studies that tracked millions of people around the world have shown that chronic loneliness raises the risk of dementia by 50%, heart disease by 29%, stroke by 32%, and premature death by 26%. In 2023 US Surgeon General Vivek Murthy likened it to smoking 15 cigarettes a day — and declared loneliness a national epidemic.

And then there is the economic side. In the US alone, isolation among seniors costs the Medicare system an extra $6.7 billion a year — isolated old people end up in hospital more often and stay longer. As for how much loneliness costs the Czech health system, we can only guess. No one has carried out that analysis.

Sources: WHO Commission on Social Connection (2025), NASEM (2020), Holt-Lunstad et al. (2015), Valtorta et al. (2016), US Surgeon General Advisory (2023), AARP PPI (2017), ČSÚ, Kooperativa Foundation.

Into this context steps ElliQ — a desktop device from Israel that looks like a little lamp with a round head and a touchscreen. At first glance, nothing revolutionary. The crucial difference from Alexa or Google Home comes down to one word: proactivity. ElliQ doesn't ask and wait for a command. It greets you in the morning of its own accord, asks how you slept, suggests a stretching exercise, reminds you about your pills in the afternoon, and offers a quiz or a meditation in the evening. When the user hasn't called family in three days, it gently suggests it.

The State of New York provided ElliQ to over 800 seniors as part of a pilot program. In a sample of 107 users who had used the device for at least a month, they measured a 95% reduction in feelings of loneliness. Seniors chatted with ElliQ more than 30 times a day, six days a week — and three-quarters of these exchanges concerned health and well-being. Not entertainment. Health.

The commercial price today is $50 to $60 a month (just under CZK 1,400). That is a fraction of the cost of a caregiver, let alone placement in a home.

But before we get carried away, this needs to be said out loud: this is not a clinical study. There was no control group, the results are self-reported, and several of the researchers are employees of the manufacturer. The figure of 95% is extraordinarily high, and no independent study has yet verified it. Even so, it is a strong signal — especially because seniors' level of engagement held up even months after deployment, which is unusual for technology aimed at seniors.

A different story is being written by PARO — a Japanese robot in the shape of a baby seal, which has existed for over twenty years and is the most-studied therapeutic robot in the world. In dementia care it reduces anxiety and agitation, improves mood, and encourages communication — even in patients who stopped speaking long ago. But: a large 2023 pooled analysis of 14 controlled studies concluded that PARO did not improve overall quality of life, and the effects faded once therapy ended. Moreover, an Australian study found that a plush toy without any AI at all achieved almost the same results — at a fraction of the cost.

In other words: we have promising signals, not proof. And that is precisely why we cannot jump straight to insurance reimbursement — but it is equally not a reason to ignore this field.

Sources: Broadbent et al. (2024) J Aging Res & Lifestyle, NYSOFA (2023), elliq.com, Karger Gerontology (2023), Mervin et al. (2018) JAMDA, BMC Geriatrics (2019).

Now to the most important point — and the one the current debate overlooks.

The notion of an AI companion for seniors is usually framed as a discussion about loneliness. That matters, but it is only half the story. The other half — and potentially the more revolutionary one — is an idea that can be summed up in a single sentence:

AI doesn't have to cure anyone. It's enough for it to make up for what's missing for an independent life.

A person with an IQ of 85 can live alone, work, and lead an ordinary life. A person with an IQ of 65 typically needs support. Between them lies a band where independence is decided by a few specific abilities: remembering a procedure, choosing between options, orienting oneself in space, handling an unexpected situation.

Today AI can — and demonstrably can — fill exactly these gaps:

Reminders and sequencing of tasks. "Now take your pills. Then have a snack. At two you have the doctor — I'll call you a taxi in an hour." That isn't science fiction; today's ElliQ already does it.

Decision support. "You have two options: call the office or go in person. In person it's 20 minutes by tram. Would you like me to prepare what to say?" Apps for people with autism handle this using visual scenarios.

Social navigation. For a person with autism spectrum disorder, a phone call to a customer line can be an insurmountable obstacle. AI that conducts the call or prompts in real time can be the difference between isolation and participation.

Safety net. Fall detection, location tracking, automatic calls for help. But also subtler things — recognizing that a person hasn't opened the fridge in three days or has stopped answering questions.

A review of 47 scientific papers from 2018–2025 documents how AI is gradually making its way into all of these areas. People with autism spectrum disorders perceive voice assistants as "friendly companions" that give them order and routine without judging them for repeated questions. The COACH system helps dementia patients complete handwashing step by step — it tracks their progress with a camera and speaks only when the person stops.

In the Czech system, the Institute for Health Status Assessment assesses degrees of dependence from I to IV. If AI helps a person at level II (moderate dependence) function closer to level I (mild dependence), that means fewer hours of paid care, later entry into a facility — and years of life spent at home. For a person with Alzheimer's it could be two extra years in surroundings they know. For a person with autism it could be the difference between sheltered housing and a flat of their own.

This shift doesn't have to be dramatic. Partial is enough. And that is exactly why the glasses analogy is so precise — glasses don't give you a hawk's eyes, but they let you read the newspaper. AI won't give you an IQ of 120, but it can help you manage the shopping.

Sources: ScienceDirect (2025) — review of 47 papers, Nature npj Digital Medicine (2024), Ki et al. (2020), PMC (2023, 2024).

Precisely because the idea of AI as a cognitive prosthesis is so appealing, the risks need to be named without sugarcoating.

The crutch danger. The most serious problem goes by the name "cognitive offloading." Studies from education show that students who use AI to solve math get better test results — but understand the principles worse. In other words: AI improves performance but may weaken ability. For a person in the early stages of dementia, this raises a question: is AI helping them hold on, or merely masking the fact that they are declining? For a person with intellectual disability: is AI building their independence, or making them dependent on itself?

A solution exists, but it requires thoughtful design. AI must be "scaffolding" that is gradually withdrawn — not a permanent prosthesis that replaces one's own effort. That is a fundamental requirement on manufacturers.

The substitution danger. Researcher Barbara Barbosa Neves of the University of Sydney warns that while AI makes contact easier, it does not always foster meaningful relationships. For some seniors, the technology may actually deepen the feeling of loneliness. A study by Berridge et al. (2023) found that some participants felt lonelier after interacting with robotic companions.

But the greatest danger is not technical — it is political. An AI companion costs a fraction of a human caregiver. For politicians looking for savings, it is tempting to say: "Seniors have a robot, they don't need caregivers." That would be a catastrophe. AI must complement human care, not replace it. Sweden, where 52% of municipalities use robotic animals in care homes, handles this with a clear rule: the robot is an aid for moments when staff can't keep up, not a substitute for staff.

The privacy danger. An AI companion that is to be useful has to know a great deal — health status, daily routine, emotional state, family relationships. For people with limited capacity to consent (dementia, more severe intellectual disability) this creates an ethical labyrinth. Who decides about data collection? A legal guardian may not represent the user's true interests. Most AI companies are based in the US or Israel — the data leaves the EU. In the Czech context, oversight by the Office for Personal Data Protection (ÚOOÚ) would be essential, together with a requirement that data stay in Europe, ideally processed directly on the device.

The deception danger. The PARO robot looks like a plush seal. ElliQ has a "head" that turns toward your voice. Hyodol in South Korea talks like a seven-year-old child. Where is the line between a therapeutic tool and a con played on the emotions? Researchers Allaban, Wang, and Padir identified six ethical problems with social robots: reduced human contact, loss of control, loss of privacy, restriction of freedom, deception/infantilization, and unclear accountability when something fails. None of them is unsolvable — but none can be ignored.

Sources: PMC (2024, 2020), Berridge et al. (2023), Allaban, Wang & Padir, TechPolicy.Press (2025).

The arguments for insurance reimbursement are strong. An AI companion of the ElliQ type costs CZK 15,000 to 24,000 a year. A part-time home caregiver comes to CZK 200,000 to 400,000. A retirement home, CZK 180,000 to 360,000. Care for a dementia patient, CZK 300,000 to 600,000 a year. If AI delays placement in a facility by even six months, the savings far exceed the price of the device.

And the argument reaches beyond money. If we accept the glasses analogy — and it is precise — then AI for a person with a cognitive deficit is not a luxury, but a compensatory aid. Health insurers pay for hearing aids, prostheses, orthopedic insoles. Why not for a device that compensates for a deficit of memory or decision-making?

The trouble is that the current system isn't ready for it — and not without reason.

AI companions don't fit any established reimbursement category. They aren't medical devices (SÚKL doesn't classify them). They aren't medicines. They aren't diagnostic instruments. Insurers rightly demand evidence from randomized controlled trials — and that evidence is still missing. The pilot data from New York are encouraging, but methodologically insufficient. There is no Czech-language AI companion, let alone one clinically validated in the Czech setting.

That is why we propose a three-tier model that does not skip over the evidence, but prepares the way:

Tier one — available now. A basic AI voice assistant for seniors aged 70+ living alone, connected to an emergency line, funded from social services. No demand for clinical evidence — this is a social, not a medical, tool. The Czech Republic has the infrastructure (emergency care, SOS buttons) — it just needs to be expanded.

Tier two — after validation. A full AI companion prescribed by a doctor on the basis of a loneliness assessment. 80% covered by the insurer, 20% co-payment. Condition: at least two quality clinical studies demonstrating benefit. This tier requires a Czech-language product — without one it makes no sense.

Tier three — for people with a cognitive deficit. An AI companion with cognitive-training modules and a link to telemedicine, prescribed by a neurologist or psychiatrist. Full reimbursement as a "device-based compensatory aid." This tier — AI as glasses for the mind — is the hardest legally and in insurance terms, but the most important socially. This is exactly where the potential lies to give people with dementia, autism, or intellectual disability extra years of life outside an institution.

The Ministry of Labour and Social Affairs (MPSV) Strategic Framework for Preparing for Ageing 2021–2025 explicitly mentions "insufficient support for the development of assistive technologies." The Czech Technical University (ČVUT) and the JINAG project in the South Moravian Region are working on assistive technologies for seniors. The foundation exists — what is missing is political will and a systematic approach.

Now (2025–2027): Fund a pilot clinical study of a Czech-language AI companion with at least 500 participants. Commission it from an independent institution (not the manufacturer), with a control group and standardized tools for measuring both loneliness and cognitive function. Set up an MPSV working group on AI in care for seniors and people with disabilities. Map which existing AI systems could be adapted for Czech.

Soon (2027–2030): On the basis of the pilot's results, create a SÚKL certification pathway for AI companions. Incorporate them into the care-allowance scheme. Train social and health workers. Establish data-protection rules compliant with the GDPR, specifically for cognitively impaired users.

Later (2030+): If the evidence supports it: introduce the three-tier reimbursement model. Build a Czech or European ecosystem of AI companions so that the country is not dependent on a single foreign supplier. Link it to the nationwide electronic health record.

Can AI be a companion to a lonely senior? Yes. The existing data show that seniors with AI companions interact with them surprisingly intensively — over 30 times a day — and report a marked drop in loneliness. Rigorous clinical studies are lacking, but the signal is strong. An AI companion is not a substitute for a daughter, a grandchild, or a caregiver. It is a supplement for the moments when a person is alone — and for seniors living on their own, those moments are the overwhelming majority.

Should AI be covered by health insurance? Not yet — because the evidence does not reach the level that insurers rightly require. But the path to reimbursement has to be prepared now, because between a pilot study, certification, and language adaptation, five to ten years will pass. And in ten years the Czech Republic will have hundreds of thousands more seniors.

Can AI supply the "missing IQ points" for an independent life? Yes — if we name it correctly. AI does not raise intelligence. It works as a cognitive prosthesis that makes up for specific shortfalls in memory, planning, decision-making, and orientation. Like glasses for the mind. For a person with Alzheimer's it can mean two extra years in their own flat. For a person with autism, the move from sheltered housing to a place of their own. For a person with intellectual disability, the ability to manage the shopping, the doctor, the government office — things most of us do without a second thought.

This potential is enormous. And that is precisely why it deserves more than enthusiasm — it deserves honest research, thoughtful rules, and the political courage to name AI as a compensatory aid. Not as a toy. Not as a substitute for human care. As glasses.

Loneliness and health impacts: WHO Commission on Social Connection (2025): From Loneliness to Social Connection. US Surgeon General (2023): Our Epidemic of Loneliness and Isolation. NASEM (2020): Social Isolation and Loneliness in Older Adults. CDC (2025): Health Effects of Social Isolation and Loneliness. Holt-Lunstad J et al. (2015): Loneliness and Social Isolation as Risk Factors for Mortality, meta-analysis of 3.4 million people. Valtorta NK et al. (2016): Loneliness as Risk Factor for CHD and Stroke. AARP PPI (2017): Medicare Spends More on Socially Isolated Older Adults.

AI companions — research: Broadbent E et al. (2024): ElliQ, an AI-Driven Social Robot. J Aging Res & Lifestyle, 13:1–21. NYSOFA (2023): ElliQ Proactive Care Companion Initiative. BMC Geriatrics (2019): Benefits and Barriers of PARO — review of 29 studies. ScienceDirect (2022): Socially Assistive Robots for Dementia — 66 studies. Karger Gerontology (2023): meta-analysis of 14 RCTs. Mervin MC et al. (2018): Cost-Effectiveness of PARO, JAMDA 19(7):619–622.

AI as cognitive support: ScienceDirect (2025): AI Influence on Individuals with Disabilities — review of 47 papers. Nature npj Digital Medicine (2024): AI for Adaptive Functioning. Ki et al. (2020): voice assistants and autism. PMC (2025): AI-Based Interventions for Learning Disabilities — 11 studies.

Ethics and policy: PMC (2024): Ethical Implications of Robots Among Older Adults with Dementia. PMC (2020): Ethical Issues with AI for Cognitively Impaired. Berridge et al. (2023): seniors' responses to social robots. Bipartisan Policy Center (2025): Paying for AI in US Healthcare.

Czech sources: ČSÚ: Population Projection. MPSV: Strategic Framework for Preparing for Ageing 2021–2025. Kooperativa Foundation: Loneliness of Seniors. Deník.cz (2026): There Are Too Few Caregivers, Smart Devices Will Help. ČVUT / Štěpánková. JINAG, DOMA JE DOMA (2025).

Transparency of creation:

The concept, structure, and editorial line of the article are the work of the author, who prepared the content outline, set the key theses, and directed the entire creative process. Generative AI (Claude, Anthropic) was used as a technical tool for research, fact-checking, and fleshing out the author's draft.

The author edited the outputs throughout, verified the key findings, and approved the final wording. No part of the text was published without human review. All factual data were verified against the publicly available sources cited in the text.

The procedure complies with the requirements of Art. 50 of EU Regulation 2024/1689 (AI Act) on the transparency of AI-generated content. #poweredByAI