An intelligent scoring method for a creative thinking test

A well-known test for assessing creativity is the Torrance Test of Creative Thinking Figural (TTCT-F). It includes scores for fluency, flexibility, originality, elaboration and abstractness of titles in the figural version. These assessment measures are highly subjective and largely dependent on the assessors’ analysis and knowledge. Also, the evaluation requires much effort in efficiency and cost. Thus, this article presents an efficient way for evaluating creative thinking that is both consistent and meaningful, particularly on the originality scale which refers to the rarity of ideas. The originality scales are calculated using a sparse coding-based scale-invariant feature transform (ScSIFT) algorithm. The proposed method is evaluated using the TTCT-F administered to 202 students. Sparse dictionary learning, sparse query image processing and picture matching algorithms were employed to process student responses. The effectiveness of this proposed method was evaluated by comparing it to a manual assessment of TTCT-F by expert judgments. The results revealed that the proposed method is as accurate as expert judgments. However, the proposed method saves more time and is more objective.