Talking Hearts: Could one button help my grandma talk and learn?
My grandma is losing her eyesight
One day, I walked past her on the couch. She was trying to open an app on her phone, but kept tapping next to it. It took what felt like twenty seconds to get in. I did not time it. I just saw her struggle.
At the time, I did not think much about it. Later, I started wondering: how often does this happen to her? She has a magnifying glass that she holds above her phone, reading glasses, and much bigger words and pictures in her phone settings. She also has arthritis affects in her thumbs. Imagine holding a phone in one hand and a magnifier in the other, then trying to scroll and tap the right tiny spot.
My mum later showed Grandma ChatGPT. Grandma named it “Rama.” Mum told me their chat began in English with everyday things. Then Grandma spoke in Hindi and the chat became much deeper. She talked about philosophy, politics and Indian news. I was not there to hear it, but I liked the thought of Grandma enjoying a proper conversation. Why should getting to that conversation be the hard part?
The idea for a talking object was actually my mum’s, then we brainstormed together to come up with Talking Hearts. As we discussed it more, I thought, if Grandma has this problem, maybe other people do too. I wanted to see if I could help.
The problem starts before the conversation
To talk to Rama on a phone, Grandma has to switch it on, unlock it, find the app, open it and turn on the talking mode. My idea is an object she can leave on a table. She presses a large TALK button once and talks for as long as she wants. She can press STOP or say “stop” when she is done.
I’m not saying all older people can’t use phones. I’m saying my Grandma struggles due to her poor eyesight and her Arthritis effects her ability to keep scrolling. This project is to help people with a similar profile to Grandma. The W3C, a group that writes web accessibility advice explains that when people zoom in on a screen, less is visible, so it takes longer to see everything. Its advice about button size also explains why small buttons can be difficult to press. That helps explain what I saw, but I still need to ask Grandma what she finds hardest.
What would the object do?
My working title for this project is currently Talking Hearts. The first version would need a microphone to hear her, a speaker to reply, a small computer inside like a Raspberry Pi device, large buttons, electricity and internet. One bit of software would turn her speech into written words. An AI model would make a reply. Another bit of software would speak the reply out loud. That is how a voice chat can happen without a phone screen in front of her.
It would also need to work out when Grandma has finished speaking. If she pauses to think, it should not cut her off. If its own speaker is loud, its microphone might even hear itself. Engineers have ways to help with these sound problems, but I will have to see how well they work in my build. The STOP button is important because saying “stop” might not always be heard.
I want her to ask about a poem, get a short answer and then decide what to ask next. Maybe it says, “What do you think the poem means?” She should be able to speak Hindi or English. I do not want it to give a giant speech every time she asks a small question. She should feel in charge of the chat.
I like the idea of a small heart shape that sits on a table. But I would change the shape if Grandma could not find the buttons easily. The shape only matters if it helps her use it.
What might it look like?
These are AI-generated design examples of how I imagine the Talking Hearts prototype might look. I am still trying different designs and researching whether I can make a 3D-printed prototype. None of these is a working device yet.
Could a chat also help her learn?
Grandma enjoys poetry and philosophy. If she asks about a poem, Rama could explain one idea, ask what she thinks and suggest something else to explore. That is more interesting than just giving one answer and stopping.
It could also ask, “Want a quick question about what we just talked about?” If she says yes, it asks one. If she gets it wrong, it gives a hint and talks through the answer with her. If she says no, they just keep chatting. It should feel like a game, not a school test.
There is one catch. The device should only ask about things it actually told her. Imagine being quizzed on a fact you were never taught. That would be unfair. I would make it keep a short list of the things they covered in that chat and choose questions from that list.
I would also start with a few topics I have checked myself. The AI could look at those notes before it answers, like checking a book. Researchers call one way of doing this retrieval-augmented generation. That long name basically means the AI looks up information instead of only relying on what it remembers. It can still get things wrong, so I would check its answers in testing.
I am not claiming my invention would fix memory problems. The US National Institute on Aging says we cannot assume that all brain games give the same results as games studied in research. I want Grandma to enjoy learning and have access to an on-demand AI companion when she wants it. That is the thing I can try to test.
What if the AI makes a mistake?
First, it might hear a word wrongly, especially when Grandma switches between Hindi and English. Instead of guessing, it could ask, “Did you mean this?” One study with older people using voice chatbots found that asking for confirmation can help when the chatbot makes a mistake. But if it asks after every sentence, that would be annoying. I need to find a balance.
Second, it might hear her correctly but give a wrong answer. AI can say something false in a very confident voice. NIST, a US science agency, warns about this. If Rama is not sure, I want it to say, “I’m not sure. Shall I look for a source?” News is harder than an old poem because news changes every day. I would start with topics I can check before trying live news.
Grandma should also know it is AI. It should never pretend to be a doctor or a real family member. And if it gets a quiz answer wrong because it misheard her, it should say sorry and try again.
What have other people built?
I looked at four other companions. They helped me see what is already possible. They also stopped me from saying my idea is completely new, because it is not. What I want to test is whether a different mix of features would suit Grandma.
Get Jai Allan’s stories in your inbox
Join Medium for free to get updates from this writer.
ElliQ talks, plays games, gives reminders and helps people contact family. It has a screen, but you can speak to it too. It even has a button you can feel and games about memory, so I cannot pretend I invented those ideas. I guessed it might be a 3/10 fit for Grandma because of the screen, but she has not tried it. My project aims to be simpler for her and easier to move around. I would need to test that.
Yaya Bear is a talking teddy bear with no screen. The company says you turn it on, say “Hello Yaya” and then talk. It also helps record people’s life stories. I think Grandma might like the easy way of talking, although I find the teddy a bit childish. The company says it works in several languages, but I could not confirm Hindi. I also could not find a clear price, size or weight. I guessed it might be a 7/10 fit for her.
Sentai surprised me. I first thought the older person had to use the phone app, but that is mainly for family. The older person talks to a small hub with no screen. They can say a wake word or press a button. That makes Sentai the closest to my idea. I could not confirm Hindi or its exact weight and size. I guessed a 4/10 fit for Grandma, but I would need her to try it. The family app is used for setup and updates; she would not need to use it for a chat.
AMI by AgingX Labs is a robot that talks. The company says it can do video calls, offer entertainment and help in emergencies. I do not think its robot look would feel right in Grandma’s living room, but that is just my opinion. I could not check its exact size, weight or price, or how much she would need to use its screen. I guessed a 2/10 fit for her.
What the companies say they offer
The table is about what I could confirm from their websites, not which device is best. “Unknown” means I could not find a clear answer. “Planned” means I have not built it yet.
For Sentai, the “Yes” in the screen row refers to the family phone app. The older person talks to a screen-free hub. A family member uses the app to set it up and see updates.
So what might be different about Talking Hearts? I want one press for a whole chat, a STOP button that is easy to find, Hindi and English, and little questions about things Grandma just learned. It also has to look and feel right for her. Those are plans, not things I have proved. Sentai already shows that a simple hub can exist. I need to build mine and let Grandma tell me what actually works.
How could I test my idea?
If Grandma wants to try it, I would compare my device with ChatGPT Voice on her phone. Both would be set up and connected to the internet first. I would time how long it takes her to get the first answer from each one. I would count times she taps the wrong place or needs help. We should try more than once, because one lucky go does not prove much.
I would also ask the same questions on both devices. Did they hear her words correctly? Did they answer correctly? Those are different things. I would check the answers using good sources. I would also try speaking from different spots in the room, first when it is quiet and then with normal background noise.
Most of all, I would ask Grandma how each one felt to use. A device could answer quickly but still be frustrating. An international guide to usability looks at whether people can do a task, how hard it is, and whether they are happy using it. That sounds like a fair way to judge my idea.
My hot take is
I think Grandma might start a chat faster with one big button than with her phone. I might be wrong. Maybe she would rather say a wake word. Maybe my speaker would be too quiet. Testing it with her is how I would find out.
I noticed that toys for little children often have clear shapes, lights, sounds and big buttons. Could I borrow the easy-to-understand parts without making something that looks like a baby’s toy? Maybe Grandma could feel where TALK is and hear a sound when the device is listening. W3C’s advice on controls backs up the idea that controls should be easy to find and use. Colour alone would not help someone who cannot see it well.
My bigger thought is that AI does not always need to look complicated. Sometimes the smartest part could be hidden inside something really simple.
Sources and picture credits
· W3C: Visual barriers — using a phone with limited sight.
· W3C: Button size and easy-to-use controls — making controls easier to use.
· Espressif: Sound tools — detecting speech and reducing echoes.
· Lewis and others: Retrieval-augmented generation — AI checking outside information.
· US National Institute on Aging: Brain-game evidence — why a game is not a proven treatment.
· Lin and others: Voice chatbots with older people — asking for clarification after mistakes.
· NIST: Generative AI risks — answers can sound sure but be wrong.
· ISO: Usability — success, effort and satisfaction.
· ElliQ features and measurements, languages, price and product picture.
· Yaya Bear website. Yaya picture supplied by my family.
· Sentai website and price, hub picture and family app picture.
· AMI by AgingX Labs, company product details and promotional picture.
· Talking Hearts design images: AI-generated concept visualisations made for this project; no physical prototype is pictured.
About me
I am a TKS student who loves animals and cares about people too, especially people who are vulnerable. I enjoy learning about Drone technology, AI, Cellular Agriculture and 3D Printing. I chose this project because I want Grandma to keep talking about things she loves and learning new things without struggling on her phone first. I hope I can build something she can use on her own. Then I want to listen to what she honestly thinks of it.
