This thesis primarily reports on an action research project that has been conducted on a course in theoretical computer science (TCS). The course is called Algorithms, data structures, and complexity (ADC) and is given at KTH Royal Institute of Technology in Stockholm, Sweden.

The ADC course is an introduction to TCS, but resembles and succeeds courses introducing programming, system development best practices, problem solving, proving, and logic. Requiring the completion of four programming projects, the course can easily be perceived as a programming course by the students. Most previous research in computer science education has been on programming and introductory courses.

The focus of the thesis work has been to understand what subject matter is particularly difficult to students. In three action research cycles, the course has been studied and improved to alleviate the discovered difficulties. We also discuss how the course design may color students’ perceptions of what TCS is. Most of the results are descriptive.

Additionally, automated assessment has been introduced in the ADC course as well as in introductory courses for non-CS majors. Automated assessment is appreciated by the students and is directing their attention to the importance of program correctness. A drawback is that the exercises in their current form are not likely to encourage students to take responsibility for program correctness.

The most difficult tasks of the course are related to proving correctness, solving complex dynamic programming problems, and to reductions. A certain confusion regarding the epistemology, tools and discourse of the ADC course and of TCS in general can be glimpsed in the way difficulties manifest themselves. Possible consequences of viewing the highly mathematical problems and tools of ADC in more practical, programming, perspective, are discussed. It is likely that teachers could explicitly address more of the nature and discourse of TCS in order to reduce confusion among the students, for instance regarding the use of such words and constructs as “problem”, “verify a solution”, and “proof sketch”.

One of the tools used to study difficulties was self-efficacy surveys. No correlation was found between the self-efficacy beliefs and the graded performance on the course. Further investigation of this is beyond the scope of this thesis, but may be done with tasks corresponding more closely and exclusively to each self-efficacy item.

Didactics is an additional way for a professional to understand his or her subject. Didactics is concerned with the teaching and learning of something, and hence sheds light on that “something” from an angle that sometimes is not reflected on by its professionals. Reflecting on didactical aspects of TCS can enrichen the understanding of the subject itself, which is one goal with this work.