A machine learning model with human cognitive biases capable of learning from small and biased datasets

Area of Science:

• Artificial Intelligence
• Cognitive Science
• Machine Learning

Background:

• Conventional machine learning requires extensive data for concept generalization, unlike human learners.
• Human learning benefits from cognitive biases that enable rapid concept acquisition.
• Bridging the gap between human and machine learning efficiency is a significant challenge.

Purpose of the Study:

• To develop a machine learning method that leverages human cognitive biases for efficient learning.
• To reduce the data requirement for machine learning models in concept generalization tasks.
• To enhance machine learning performance in scenarios with limited and biased data.

Main Methods:

• Implemented a human cognitive model into machine learning algorithms.
• Compared the performance of the new models against established methods like Naïve Bayes, SVM, neural networks, logistic regression, and random forests.
• Focused on the spam classification task, a benchmark for data-intensive machine learning.

Main Results:

• The developed models demonstrated superior performance compared to conventional methods.
• Effectiveness was particularly pronounced when using small and biased datasets.
• Achieved higher accuracy in spam classification with limited data inputs.

Conclusions:

• Integrating human cognitive biases into machine learning significantly enhances learning efficiency with small datasets.
• This approach offers a promising direction for developing more human-like artificial intelligence.
• The method shows potential for improving performance in various machine learning applications beyond spam classification.