AI in healthcare

The idea

In healthcare, AI helps doctors by spotting patterns in pictures and records quickly, but people make the decisions and give the care.

What is AI's job in a hospital?

Many hospitals now use AI. What do you think its job is?

• To do the doctors' work, so doctors are not needed
• To help doctors with some jobs, while people stay in charge
• Only to do paperwork

Who does what?

Each card is a job in a hospital. Decide: can AI help the doctor with it, or does it need a person, for judgement and care?

Some cards are tricky. Tap a group. Read the reason, then press Next card.

A helper, not a replacement

The jobs on one side were about pictures, records, readings and speech: huge amounts of data, where fast pattern-spotting helps. The others were about care, choices and responsibility.

Even when AI helps, the doctor looks at what it shows and decides. AI can be wrong, so a person must check.

In healthcare, AI helps doctors by spotting patterns quickly, but people make the decisions and give the care.

Hospitals use AI as a tool. The doctor and the patient stay at the centre.

Notes

In healthcare, AI helps doctors by spotting patterns quickly, but people make the decisions and give the care.

When you feel unwell, tell a trusted adult and see a doctor. A chatbot cannot examine you.

Check yourself

An AI program compares a skin photo with thousands of others and highlights a patch that looks unusual. What happens next?

Kabir has a bad stomach ache. A friend says, “Ask a chatbot what is wrong.” What is the best next step?

A hospital wants to use patients' records to teach an AI to spot disease. What must it do about the records?

• The patient is told at once what the illness is, since the AI found it. A highlight is a hint, not a diagnosis. AI can be wrong, so a doctor must decide what it means.
• The highlight is ignored, since it came from a machine. A highlight can point the doctor to something easy to miss. It is a useful hint, as long as a person checks it.
• A doctor examines the patient and decides what it means — correct. Yes. The AI helps by pointing, and the doctor uses judgement and knowledge of the patient.
• Tell a parent or teacher, so that a doctor can check him — correct. Yes. A doctor can examine Kabir and ask questions. A chatbot cannot examine anybody, and it can be wrong.
• Take the medicine that the chatbot suggests for stomach aches. Doctors choose medicine by looking at the patient's health. A chatbot has no record of Kabir's health, and it can be wrong.
• Wait a few days and see if the pain goes away on its own. Pain can mean something that needs a doctor. Telling a trusted adult early means help is not delayed.
• Share them with a company that asks, because it will improve the AI. Health records are very private. They must not be handed out, even for a good purpose, without the patients' permission.
• Use them freely, because records belong to the hospital and not the patients. The records are about real people, so patients have a say. A hospital looks after the records but does not get to ignore their privacy.
• Use them with permission and strong protection, and remove names where possible — correct. Yes. Health information is personal, so patients' privacy must be protected when it is used to teach AI.