Human knowledge models: Learning applied knowledge from the data.

Area of Science:

• Artificial Intelligence
• Cognitive Science
• Machine Learning

Background:

• Current AI models create a disconnect between computational processing and human knowledge acquisition.
• There is a need for AI systems that align with human cognitive processes.

Purpose of the Study:

• Introduce Human Knowledge Models (HKMs) to replicate human computational abilities.
• Develop a new machine learning paradigm based on human cognitive research.
• Enable AI to generate human-understandable, computable, and applicable knowledge.

Main Methods:

• Formalized the definition of HKMs based on cognitive research.
• Developed a new machine learning approach by training models with human processing capabilities.
• Validated HKMs using diverse datasets from applied fields.

Main Results:

• HKMs demonstrated high predictive power and resistance to noise and overfitting.
• HKMs efficiently mine knowledge directly from data, rivaling complex AI in explaining patterns.
• Learned knowledge is understandable, computable, modifiable, and applicable by humans.

Conclusions:

• HKMs show significant potential for decision-making applications, especially where "black box" models are unacceptable.
• This research enhances understanding of human decision-making and its potential to approach ideal solutions.
• HKMs bridge the gap between AI processing and human knowledge discovery.