Phoster

Research and Development

Machines Teaching Machines to Solve IQ Test Items

Introduction

Artificial intelligence systems can be evaluated with respect to their performances on IQ tests. The training of such systems can be optimized via machine teaching techniques.

Psychometric Artificial Intelligence

Psychometric artificial intelligence posits that artificial intelligence systems can be designed to exhibit intelligence by solving the same intelligence test items as humans. The same methodologies and techniques are of use for measuring both varieties of intelligence: natural and artificial. Advancements with respect to the measurement of one variety of intelligence entail advancements with respect to the measurement of the other.

Machine Teaching

Machine teaching is the control of machine learning. Machine learning algorithms define dynamical systems where states, or models, are driven by training data. Machine teaching designs optimal training data with which to drive learning algorithms to target models.

Machine teaching is also applicable to educational scenarios with human students. Machine teaching can optimize lessons to help students to achieve educational goals. If one can assume a cognitive learning model of a student, one can use machine teaching techniques to reverse-engineer the optimal training data.

Item Generation and Evaluation

Machine teaching algorithms can generate worked examples and test items for training human students and psychometric artificial intelligence systems. The algorithmic generation of items is referred to as automatic item generation. Items can be evaluated utilizing item response theory.

The Assessment and Development of Cognition

Beyond the measurement of intelligence, sequences of exercises can be designed to activate, strengthen, coordinate and integrate specific regions and processes of both varieties of cognition: natural and artificial. That which there is to be learned from the intersection of machine teaching and psychometric artificial intelligence will advance our understanding of the teaching of human students via personalized sequences of exercises and activities.

Conclusion

Artificial intelligence systems can be evaluated with respect to their performances on IQ tests. The training of such systems can be optimized via machine teaching techniques.