New AI model can predict the difficulty of logical thinking

Tjeerd Fokkens wanted to investigate and understand why people have difficulty thinking logically. He created a theoretical model based on a cognitive bottleneck theory, where each piece of information must be processed in a certain order and also overcome a limitation in the brain's capacity. It is this process that makes logical reasoning challenging for us humans.

"In the experiments I conducted, the model was able to predict the human difficulty with logical thinking with impressive accuracy. These results surprised me that a model based on a relatively simple idea could make such powerful predictions", says Tjeerd Fokkens. 

By integrating multiple machine learning techniques, he was able to speed up the model's calculations, making it useful in technical environments.

One such future application is debugging knowledge bases – advanced databases used in AI, semantic web development and industrial automation. These systems often contain complex programming errors. 

”Using the AI ​​model, the user interface can automatically select the most understandable explanations for these errors, which makes the troubleshooting process significantly easier and less cognitively demanding for those working with programming or domain development”, says Tjeerd Fokkens.

Another, more philosophical insight he gained while working on his thesis was how difficult it is to narrow down the concept because each individual thinks differently. 

”I managed to define what cognitive difficulty means, and take into account individual differences. The definition turned out to be surprisingly technical.”

Text: Monica Havström
 

Thesis title: Ordering Thought

Link to thesis:  https://hdl.handle.net/2077/89482

Tjeerd defended the thesis on November 7, 2025.