How does artificial intelligence make its choices? Science breaks the black box

Artificial intelligence plays an increasingly important role in our lives. But if you don’t get a loan “because the computer says so,” it’s natural to feel unfairly treated. We would prefer to have artificial intelligence that can explain how it arrived at its predictions. Fortunately, there are all kinds of techniques for this.

If you undergo an MRI or CT scan, there is a high probability that the radiologist will have help from artificial intelligence. For example, there are artificial intelligence systems that automatically stain organs or determine the dimensions of tumors upon examination. This helps, for example, in developing a radiation plan for treating cancer.

Doctors are increasingly adopting this type of artificial intelligence. Not only because the systems work better and thus save time, but also because they are easier to verify. If the system makes an error, the doctor quickly sees that something is wrong and can correct the error.

Read also:

“Because the computer wants it”

But this controllability does not yet exist in many other healthcare applications, says Peter Bosman, head of the research group Evolutionary intelligence At the Center for Mathematics and Computer Science (CWI) in Amsterdam. “There are systems that give you a diagnosis using a patient’s symptoms and characteristics as input. This saves a lot of time, but doctors are reluctant if such a system cannot explain how the diagnosis was arrived at.”

This type of AI system becomes one Black box It’s called: You enter data and a prediction comes out, but you don’t know what’s going on inside. This is a problem in health care. If your doctor tells you after a CT scan that she recommends treatment, you – just like your doctor – want an explanation.

See also  Gyms are getting rid of tanning beds after a call from dermatologists

Developing artificial intelligence systems that can explain how they arrive at their predictions is the subject of the research field explainable artificial intelligence Explainable artificial intelligence. You can read all about this in KIJK 3/2024. Order this edition In our online storeOr simply via the button below.

Text: Queen Verfluzim

Image: Shutterstock

Megan Vasquez

"Creator. Coffee buff. Internet lover. Organizer. Pop culture geek. Tv fan. Proud foodaholic."

Leave a Reply

Your email address will not be published. Required fields are marked *