Automated Assessment of Complex Programming Tasks Using SIETTE

This paper presents an innovative method to tackle the automatic evaluation of programming assignments with an approach based on well-founded assessment theories (Classical Test Theory (CTT) and Item Response Theory (IRT)) instead of heuristic assessment as in other systems. CTT and/or IRT are used to grade the results of different items of evidence obtained from students' results. The methodology consists of considering program proofs as items, calibrating them, and obtaining the score using CTT and/or IRT procedures. These procedures measure overall validity reliability as well as diagnose the quality of each proof (item). The evidence is obtained through program proofs. The SIETTE system collects and processes all data to calculate the student knowledge level. This innovative method for programming task evaluation makes it possible to deploy the whole artillery developed in this research field over the last few decades. To the best of our knowledge, this is a new and original contribution in the area of programming assessment.