When an artificial neural network (ANN) takes a psych evaluation

Dr. Been Kim, who is a senior researcher at Google Brains has conducted an experiment where an artificial neural network (ANN) took one of the tests of Gestalt therapy. The results did not amuse the scientists but helped them find one more feature that differs from the human brain from ANN’s core.

What is an artificial neural network?

ANN is a computer system that is designed after a human biological neural network. Simply put it’s a learning machine that is trained like a human brain to solve the tasks that are set for it. As ANN is designed after the human brain it has levels of neurons: from upper to lower (they have different functions — from supporting voluntary movements is a spinal cord to conducting conscious movements).

The thing is that theoretically if it’s made according to the image of the human brain, it should be working like that? Thank god no, and artificial intellect works better. In the situation where the mathematics is needed, of course.

Gestalt therapy and ANN

Let’s memorize what Gestalt therapy is? It is a branch of psychotherapy that is based on experimental and phenomenological methods. Consciousness as the foundation of this therapy and conscious mind which organizes the pieces of chaos in a meaningful whole. Here is how we describe the Gestalt effect: something more than just some parts. It means that our brain analyzes the glimpses of things and makes a guess that it means something.

Dr. Been Kim has set an experiment where the ANN should have to go through the Gestalt test and guess what’s on the picture. The team of the scientists generated three datasets which included: triangles, triangles with incomplete corners and triangles with removed edges. Then they trained the network what a triangle is and test it on the dataset.

Difference between human and AI

Trained to complete the triangles, ANN seems to exhibit the law of closure. The networks better classified the illusory set as triangles than those with randomized connection weights or networks trained on white noise.

It means that the ability of the network to complete the chaos creating objects is connected with the ability to generalize. In the case of humans who undergo the same test, they classified the incomplete triangles as a sign. It can better be shown with the prism of semantics science with its expression of an object and the meaning of it.


When taking psychological tests of Gestalt therapy, an artificial neural network (ANN) recognizes the expression of the object (where a human sees a forest, AI recognizes trees), and the question “why” is still waiting to be answered.




Stories on science, critique, and social behavior. Facebook: https://www.facebook.com/savashonair/

Love podcasts or audiobooks? Learn on the go with our new app.

Recommended from Medium

Tsinghua AI Association of International Students (TAIS) news roundup 6 November

Understanding neural networks and backpropagation-Part I

Matching Attributes: How Alibaba’s AI Learned to Pick an Outfit

AI — The Shift From Cloud to Local Device

I have a very robust and well organized note taking system (heavy sarcasm) and found this gem in…

Could Automation and AI break the BI adoption barrier? (Part 1)

Be Prepared for 2018: the Year of the Bot

3 Steps To Maximize ITSM Efficiency Through AI And Reduce IT Support Costs

Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store
Savash on-air

Savash on-air

Stories on science, critique, and social behavior. Facebook: https://www.facebook.com/savashonair/

More from Medium

Marriage on our own terms

Insects and Perception — Reflection on 19th Century Insect Technics

I want to sink into your ocean.

Why We Won’t Be Using Telegram Today (Or Any Other Day)