As software becomes increasingly important, so does the education of people who develop software. To provide a good and motivating education to learners, it is necessary to be aware of their preferences. To give insight into the preference of learners, we conducted a study with university students and professionals aimed at learning more about their preferences. Results from our survey- and focus group-based research show large similarities between both groups of learners, such as the preference for e-learning materials, as this provides them flexibility for scheduling their learning sessions. However, differences exist when it comes to the preferred scheduling of online and offline sessions. Our findings can help educators not only in designing courses for university students or professionals, but also in designing courses that can be used by both groups of learners.

We are pleased to introduce the 2024 36th International Conference on Software Engineering Education and Training (CSEE&T) held in Würzburg, Germany, from July 29thto August 1st, 2024.

Software inspections play an important part in ensuring the quality of software development.With the emergence of model-based development approaches, there is also a need for model inspections to ensure correctness of model-based artifacts.In practice, ad hoc inspections are regularly conducted, often by new and rather inexperienced colleagues, which are asked spontaneously to review an artifact of interest.The use of novices, such as trainees or student assistants, allows shorter review cycles at reduced costs.The quality of these ad hoc inspections is commonly attributed to different factors, often related to the reviewer.Increasing review time can be seen as an indicator that the reviewer takes the review serious.Furthermore, with more time spent, it can be assumed that more defects will be found.In this paper, we report the results of an experiment on ad hoc model inspections.Our results show that -contradictory to these assumptions and empirical findings from inspections of textual documents -the review time a reviewer decides to spend on a review has no significant influence on the effectiveness of ad hoc model inspections.

Requirements validation is an important aspect for ensuring high quality software. Commonly used are requirements inspections, where the specification is read from different persons assuming different roles or applying different reading techniques, partly accompanied by checklists. Actual defect detection with requirements inspection is costly, and defect detection rates must be considered low. Therefore, repeated validation is used or validation with multiple inspection groups - known as N-fold inspections. However, this does not only yield more defects found, but also more false positives. In this paper, we investigate how defect aggregation can be used to improve the overall quality of validation. Therefore, we conducted an experiment with 22 N-fold inspection groups consisting of four to five reviewers each. Results show that simple aggregation of all results leads to a number of false positives that can actually negatively impact the validation task, while the use of more tailored aggregation strategies can considerably improve the validation of requirements with N-fold inspections.

This paper discusses the challenges in evaluating the quality of conceptual models in educational settings. While automated grading techniques may work for simplistic modeling tasks, realistic modeling tasks that allow for a wide variety of solutions cannot be evaluated using automated techniques. However, the traditional approach of having instructors grade the exercises may not be feasible in larger courses. To address this issue, alternative approaches, such as educating students to assess the quality of their own solutions or using calibrated peer reviews, can be used. Therefore, it is crucial to identify the quality of feedback a student can deliver on their own. As a first step, this paper reports on the results of controlled experiments with 368 participants to investigate factors that influence students’ model comprehension and to identify ways to distinguish good student assessments from bad ones.

This position paper discusses the potential for using generative AIs like ChatGPT in software engineering education. Currently, discussions center around potential threats emerging from student's use of ChatGPT. For instance, generative AI will limit the usefulness of graded homework dramatically. However, there exist potential opportunities as well. For example, ChatGPT's ability to understand and generate human language allows providing personalized feedback to students, and can thus accompany current software engineering education approaches. This paper highlights the potential for enhancing software engineering education. The availability of generative AI will improve the individualization of education approaches. In addition, we discuss the need to adapt software engineering curricula to the changed profiles of software engineers. Moreover, we point out why it is important to provide guidance for using generative AI and, thus, integrate it in courses rather than accepting the unsupervised use by students, which can negatively impact the students' learning.

Abstract Model inspections are important to ensure high‐quality software and to satisfy legal obligations in model‐based engineering processes. As model‐based specifications are typically documented on type‐level, errors concerning the interactions between multiple system instances can go unnoticed. For collaborative cyber‐physical systems (CPS), a plethora of possible instance‐level configurations need to be taken into account. Therefore, we propose the definition of instance‐level review diagrams that show representative interactions of instance‐level configurations that help detect defects in the system specification. To evaluate the approach, we conducted a controlled experiment whose results indicate that instance‐level review diagrams have—compared with type‐level diagrams—important positive effects on reviewing processes for behavioral specifications of CPS. Specifically, the experiment provides empirical evidence that instance‐level review diagrams are significantly more expressive and effective than type‐level diagrams.

Abstract One central aspect of software engineering research is the transfer of the proposed approaches into industrial practice. In the past, a number of technology transfer approaches and experiences from technology transfer projects in software engineering have already been reported. However, many researchers still struggle to get their research results noticed by practitioners. To investigate what is important to practitioners, we conducted a mixed‐methods study that provides us with reliable quantitative data as well as deeper insights from qualitative data. Our results show that there is a mismatch between industry professionals' needs and commonly proposed technology transfer approaches in the software engineering field. For instance, collaboration between industry and academia as well as participation in empirical evaluations is not deemed important from an industry point of view. In contrast, industry professionals emphasize the use of company‐specific pilot projects conducted by industry and the need for experts to be available in every phase of technology transfer.

Abstract Students’ experience is used in empirical software engineering research as well as in software engineering education to group students in either homogeneous or heterogeneous groups. To do so, students are commonly asked to self-rate their experience, as self-rated experience has been shown to be a good predictor for performance in programming tasks. Another experience-related measurement is participants’ confidence (i.e., how confident is the person that their given answer is correct). Hence, self-rated experience and confidence are used as selector or control variables throughout empirical software engineering research and software engineering education. In this paper, we analyze data from several student experiments conducted in the past years to investigate whether self-rated experience and confidence are also good predictors for students’ performance in model comprehension tasks. Our results show that while students can somewhat assess the correctness of a particular answer to one concrete question regarding a conceptual model (i.e., their confidence), their overall self-rated experience does not correlate with their actual performance. Hence, the use of the commonly used measurement of self-rated experience as a selector or control variable must be considered unreliable for model comprehension tasks.

Abstract Collaborative cyber-physical systems are capable of forming networks at runtime to achieve goals that are unachievable for individual systems. They do so by connecting to each other and exchanging information that helps them coordinate their behaviors to achieve shared goals. Their highly complex dependencies, however, are difficult to document using traditional goal modeling approaches. To help developers of collaborative cyber-physical systems leverage the advantages of goal modeling approaches, we developed a GRL-compliant extension to the popular iStar goal modeling language that takes the particularities of collaborative cyber-physical systems and their developers’ needs into account. In particular, our extension provides support for explicitly distinguishing between the goals of the individual collaborative cyber-physical systems and the network and for documenting various dependencies not only among the individual collaborative cyber-physical systems but also between the individual systems and the network. We provide abstract syntax, concrete syntax, and well-formedness rules for the extension. To illustrate the benefits of our extension for goal modeling of collaborative cyber-physical systems, we report on two case studies conducted in different industry domains.

In the field of safety-critical systems, manual reviews are important to ensure high-quality software and to satisfy legal obligations. When applying model-based engineering approaches, no longer are only textual requirements specifications or software code under review, but also model-based specification artifacts like behavioral requirements models. As such behavioral specifications are typically documented on a type-level, errors concerning the interactions between multiple system instances can go unnoticed in manual reviews. This is particularly the case when multiple system instances of the same system type are interacting during runtime, which is typical for cyber-physical systems where networks of cyber-physical systems form dynamically to fulfill an overall purpose. In this paper, we report on a controlled experiment whose results indicate that instance-level review diagrams have -- compared to type-level diagrams - important positive effects on reviewing processes for behavioral specifications of cyber-physical systems. Specifically, the experiment provides empirical evidence that instance-level review diagrams are significantly more expressive and effective than type-level diagrams.

Background: In model-based engineering models need to be regularly validated by manual assessment. Therefore, review processes have established. This commonly means that a visual inspection of the respective models is conducted. Aims: In this paper, we report a study that aims at investigating whether automatically generated review models can aid the manual review of model-based specifications. We investigate this for the case of embedded systems' functional design. Method: For that purpose, we compared the manual review of the functional design with the review of an automatically generated review model. In this paper, we report on a controlled experiment to compare effectiveness, efficiency, user confidence, and subjective supportiveness of both review artifacts. Results: The experiment results show that the use of the review model as review artifact for the functional design is significantly more effective, leads to a significantly higher user confidence in decision making and is valued as significantly more supportive than the review of the original functional design. Conclusions: Our experiment provides evidences that reviewing a generated review model instead of the original model-based specification of the functional design increases the quality of the reviews. Our findings also indicate that the use of generated review models have the potential to improve the review of model-based specifications in general.

[Background] Systematic literature reviews and maps have become well-established research methods in software engineering research. Of the three commonly suggested and used search methods: manual search, database search, or snowball search; systematic literature reviews and maps typically employ one or a combination of two or three of those as their search strategy. As systematic literature reviews and maps raise a claim to result in a representative set of relevant papers for a certain area of investigation, it is of importance to understand the impact the search strategy has on achieving this goal. [Aim] This paper contributes a study to compare all three search methods. This study aims at providing evidence as to what advantages and disadvantages of these three search methods are. [Method] We conducted three systematic literature reviews on the same topic, which affects multiple software engineering related disciplines, using different search methods, while keeping other parameters like inclusion and exclusion criteria consistent among all three reviews. [Results] Our results show a similar effectiveness for snowball and database search and the highest efficiency for database searches. However, our literature reviews led to three barely overlapping sets of papers, which in turn led to distinct impressions of the same field. [Conclusion] Our results show that the use of a single search method can lead to a set of included papers, which misrepresents the research field under investigation. Hence, particularly when conducting literature reviews that affect different software engineering sub-disciplines and related disciplines, researchers should not just rely on the single most effective and/or efficient search method.

Cyber-physical systems interact closely at runtime with their operational context, i.e., other systems, users, and external entities. Therefore, eliciting and documenting interactions with the operational context allows documenting rationales for design decisions, which is essential to define a system architecture that suits the system’s purpose. For cyber-physical systems acting in system networks, this task becomes challenging since these networks can change at runtime, because individual systems can leave or join a network spontaneously. Hence, it becomes impossible to predict at design time the operational contexts at runtime. To aid context analysis and specification of such systems, this paper contributes a solution approach where the context of a system is not defined based on a static system boundary but depends on the specific intended use. We propose a view-centric perception of the operational context to leverage the functionality of external systems. This allows differentiating between functionality offered by the cyber-physical system network as a whole and the functionality provided by each individual system. An evaluation with an industrial case example and industry collaboration shows that this approach allows making suitable assumptions of the operational context at design time to guide architecture decisions.

Abstract Introduction Technology transfer aims at supporting the transfer of results from software engineering research from academia to industrial application. Objective This paper reports on the current state of technology transfer in software engineering. Method We conducted a systematic literature review, in which we investigated 3070 papers. We identified in total 70 relevant papers, which were subject of a detailed analysis. Results Many different approaches are proposed to foster technology transfer in software engineering. The majority of these approaches suggest direct collaboration between industry and academia or teaching new technologies in industrial training or university education. In addition, a considerable number of experience reports on technology transfer exist. Hence, a multitude of best practices, success stories, and lessons learned is reported. Among others, empirical evidence, maturity, and adaptability of the technology seem important preconditions for successful transfer, while social and organizational factors seem important barriers to successful technology transfer. Conclusion Our findings can aid software engineering researchers in determining how best to support the transfer of their research results into practice. Furthermore, analysis of the literature also revealed that no reports exist on the combination of various technology transfer approaches, which could increase advantages of existing approaches while reducing their disadvantages.

Educational approaches for computer science proposing the use of complete online courses or traditional courses employing some kind of online material have received much attention recently. The integration of online materials into traditional courses or the replacement of entire courses offer huge possibilities, including increased teaching quality and better study and work alignment. However, researchers and teachers also identified some drawbacks of using online material, including the lack of interaction between students and teachers, and the need to discuss and provide feedback of the students' exercise results. A solution for providing such feedback are automated assessment tools which can generate feedback. However, these tools are not applicable in all situations, e.g. for providing feedback to conceptual modeling exercises. In this paper, we report on the design and implementation of an online course for teaching conceptual modeling. In this course, we use explicitly ambiguous exercises and sketch multiple solutions in brief whiteboard-style videos, thus enabling students to assess their own solutions. Evaluation results show that the proposed approach is able to fulfill students' educational needs.

Cyber-physical systems are highly connected context sensitive systems that form networks. Within these cyber-physical system-networks, behavior emerges from the interplay of the connected systems that cannot be attributed to a single system. Verifying single system behavior as well as the resulting emergent behavior of the system-network the single systems contribute to, is challenging as the intended behavior differs between the different cyber-physical system-networks the single system takes part in. It can even differ between two almost identical cyber-physical system-networks, which, for example, only differ by one system. To ensure correct behavior, requirements engineering for cyber-physical systems must cope with the identification and documentation of the cyber-physical system's dynamic context, i.e. the different system-networks the system takes part in (e.g., a system-network of vehicles forming a platoon on a highway) as well as the context situations these system-networks can encounter (e.g., road work leading to the need for lane shifts). This paper contributes a solution idea for automated support in identifying relevant system-networks the system will have to interact with and for verifying the cyber-physical system under development against these relevant system-networks.