Synthetic tutor : profiling students and mass-customizing learning processes dynamically in design scripting education

Abstract

Artificial intelligence is substituting human intelligence and robots are replacing human workers. Instead of settling for this competitive relationship between humans and machines, this thesis proposes a novel framework in which humans and machines work together to solve the complex problems of design-scripting education, problems which humans or machines alone cannot easily solve. In design education, there are few clear guides and pedagogies that can effectively teach students with diverse educational and professional backgrounds, some of who may need individualized tutoring. This thesis specifically explores applications of artificial intelligence (machine learning and computer vision algorithms) in which humans and machines mutually improve their learning performance. Humans can increase a machine's performance by providing training-data sets that can be a foundation for intelligent decision-making. Machines, on the other hand, can improve humans' learning performance by analyzing human study patterns and providing mass-customized instructions. This thesis illustrates that the developed Synthetic Tutor provides novice students with architectural precedents by analyzing their drawings and documents and effectively teaches these students introductory computer programming skills in the context of architectural design. Therefore, this human-machine collaboration has proven an effective framework to solve these ill-structured problems.