How can unplugged approach facilitate novice students’ understanding of computational thinking? An exploratory study from a Nigerian university

•This study investigates the impact of unplugged activities on students' computational thinking (CT) skills and reported how the intervention in the context of a developing country can improve novices’ problem-solving.•A mixed-method research approach and a quasi-experimental design were applied in this study to explore how the CT unplugged can be integrated into the classroom.•Findings from this study suggests that CT concept improve students’ problem-solving skills, however, further investigation is required to complement this finding.•This study also showcased how to explore natural language processing technique such as topic modelling to gain insights into students’ learning process.•The study provides relevant recommendations and implications including how to demonstrate CT unplugged activities to students, and to explore NLP in investigating learners’ learning outcomes. Computational thinking (CT) has been relevant in teaching STEM courses, and educators are incorporating both plugged and unplugged CT activities in their modules to prepare new entrants for advanced knowledge. However, additional empirical data are needed to understand how an unplugged activities approach can lead to an improved understanding of CT concepts. Therefore, this study investigates the impact of unplugged activities on novice students' CT skills and unravels the characteristics of students' CT competency from performing unplugged activities. The experimental aspect of this study was conducted with a cohort of 210 students participating in an introductory programming course at a University in Nigeria. A mixed-method research approach and a quasi-experimental design were applied in this study to understand how the unplugged approach in teaching CT could help students improve their CT competency. A paired-sample t-test of familiarity score for each CT concept showed a significant increase in the score before and after the intervention. Furthermore, Latent Dirichlet Allocation (LDA) was used for topic modeling of the text describing students’ solutions. Topic modeling allows for more insights into the steps taken by the students to solve CT problems. Only 27.62 % of students' solution descriptions contained CT concepts, 56.90 % of the students solved the puzzle with an optimal score, and 72.41 % of students who applied CT concepts of problem abstraction and decomposition were able to solve the puzzle correctly. Indeed, CT unplugged activities significantly increased the number