Introduction

The fields of computer science and software engineering require a strong understanding of data structures and algorithms. These concepts form the foundation for solving complex problems and are necessary for developing efficient and scalable software systems. Understanding algorithms and data structures pose a significant challenge for undergraduate computer science and software engineering students (Kitchin, 2017). Computer code, words, and individual pictures represent static descriptions, specific views, or instances of a dynamic process. This should be given special attention during the learning process because some students cannot translate static descriptions such as these into a dynamic process in their imagination (Kitzie, 2019). As a result, these students often have difficulty understanding the algorithm’s code. In addition to empirical and mathematical analysis, algorithm visualization (AV) is another technique used in teaching data structures and algorithms. Visualization is the process of data abstraction, operations, semantics of the algorithm software, and the creation of graphical content of the abstractions. Creating a visualization of the algorithm should make it easier to understand (Seaver, 2018; Rajala et al., 2007).

Research has demonstrated that visualization and animation can effectively enhance student learning outcomes in data structures courses. Osman and Elmusharaf (Osman & Elmusharaf, 2014) conducted an experiment to compare the performance of computer science and information technology students with and without a visualized learning environment. The study found that using the visualized learning environment positively impacted student understanding and performance, as reflected in their scores on the final exam. Nathasya et al. (2019) integrated algorithm visualization, presenting DS-PITON as a tool to help students understand data structure implementation. The study aimed to address the limitations of traditional algorithm visualization tools, which do not provide enough technical details. Their study showed that using the integrated DS-PITON tool improved students’ assessment scores and helped moderate-paced students complete assessments faster. Andreev et al. (2021) examined the use of visualization technologies in distance education and found that visualization improves the quality and effectiveness of training, and that visualization tools help to improve distance learning for students. However, they acknowledged the need for further research to explore other aspects of visualization in professional training. Perháč and Šimoňák (2022) developed a new algorithm visualization system designed to teach data structures and algorithms. Their study aimed to improve the usability, extensibility, availability, and functionality of existing systems. The newly-developed system provided most functionalities for data structure visualization and was evaluated as easy to understand and helpful in learning. The respondents requested the addition of new visualizations and functionalities, indicating the potential for further development in this area.

This paper explores the benefits of incorporating animations and visualizations in teaching and learning data structures and algorithms in computer science courses. We describe the case study performed during the fall of 2021 in a Data Structures course at SCEFootnote 1 College. For this study, we developed a dedicated interactive website tool tailored to support the Data Structures course at our institution. The tool encompasses all the data structures and algorithms taught in the course. We demonstrate the potential impact of using visual aids in the classroom, including willingness to use the tool to clarify complex concepts and ease of using the tool, and enhanced problem-solving skills.

Research objectives

The primary objective of this research was to design and implement a free online animation and visualization tool to complement the undergraduate Data Structures course and evaluate its impact on student comprehension and performance. The developed tool covers the data structures taught in the course, providing the visual capability to represent the algorithm execution process. We investigated the effect of this specialized tool on students’ comprehension and academic performance in the Data Structures course at the Sami Shamoon College of Engineering.

We formulated the following research questions:

  • RQ1. How do students evaluate the use of animation and imaging technology in learning data structures concepts?

  • RQ2. How do students accept the AVDS tool?

  • RQ3. How will the use of AVDS affect the understanding of algorithms and learning?

Based on the research questions, the following hypotheses were formulated:

  • H.1. Using animation and imaging technology in teaching data structures will be useful to students.

  • H.2.1. Perceived ease of use (PEOU) will have a positive impact on users’ attitudes toward using (ATU) the AVDS tool.

  • H.2.2. Perceived usefulness (PU) will positively influence users’ ATU the AVDS tool.

  • H.2.3. The PEOU will positively influence users’ PU towards the AVDS tool.

  • H.3.1. Using animation and imaging technology in teaching data structures will enhance students’ performance in solving exercises.

  • H.3.2. Students can complete more questions in a given amount of time (i.e., completion time will be shortened).

Methodology

The study was carried out using a performance test as an objective measurement designed by the research team (Appendix A, Table 12). In addition, we applied the common standardized questionnaires, a Technology Acceptance Model (TAM) and System Usability Scale (SUS) questionnaires (Appendix A, Tables 10 and 11), which are acceptable for testing new technologies.

The content of the questionnaires and the order of their presentation were precisely determined in advance. The SUS and TAM questionnaires were translated from English by three people and then validated by three others. The questionnaires included closed-ended questions in which students needed to choose an answer from a pre-determined set of answers. The performance test included two types of questions: open-ended and partially closed-ended; the latter required written justification of the answer. The answers were coded according to the percentage of correctness from 0 to 100.

Participants

A survey was taken among second-year students enrolled in the Data Structures course at the SCE College, Beer Sheva, Israel. Of 162 students in the course, 78 showed interest in participating in the research study: 23 females (mean age 21, SD = 2.56) and 55 males (mean age = 21, SD =1.97). Since all the students study in the same department and have completed all pre-course requirements, we assume they are similar to each other in terms of their educational background and academic level, and they can be classified as a homogeneous population.

Questionnaires

The SUS is a standardized questionnaire used to measure the perceived usability of a system (Suharsih et al., 2021). The SUS score is calculated by obtaining responses from participants on a 10-item questionnaire. Every item contribution score ranged between 1 (strongly disagree) and 5 (strongly agree) (Brooke et al., 1996).

The TAM (Al-Emran and Shaalan, 2021; Agbo et al., 2022; Dimitrijević and Devedžić, 2021) questionnaire used in the study measures three key factors related to the acceptance of the AVDS tool:

  1. 1.

    Perceived usefulness (Davis et al., 1989): The extent to which individuals believe that using the AVDS tool will enhance their test performance and time efficiency.

  2. 2.

    Perceived ease of use (Davis et al., 1989): The extent to which a student’s use of AVDS tools is free of effort.

  3. 3.

    Attitude toward usage (Davis, 1993): The extent to which an individual’s attitude is favorably or unfavorably disposed toward using the AVDS tool and the influence of this attitude on the intention to use the tool.

These factors are measured to determine the level of acceptance and adoption of the AVDS tool and to provide insights into areas that need improvement.

Performance test

The performance test (see Appendix A, Table 12) helped researchers learn about the efficiency of the tool. For the test, we chose the AVL tree topic, a self-balancing binary search tree developed in 1963  (Adelson-Velskii & Landis, 1963). The AVL topic was taught in class a week before the experiment. The test consisted of six questions and lasted about 20 min. During the test, students in an experimental group used the developed Data Structures online tool, while the control group solved without that assistance. The AVDS tool helped the student solve the questions or verify the answers. At the time of the experiment, the system did not show the steps of the algorithm, but did provide a final answer only. To assess the students’ understanding of the solution, the assessment included more than just the final answer. It required students to explain the steps involved. For example, students needed to determine if a rotation was necessary during the insertion of a new node in the given AVL tree.

Data structures online tool

We found several publicly available AV tools: Data structure visualization, Visualgo, CS1332 Visualization Tool, Algorithm Visualizer, Vamonos, and Algorithm Wiki. However, the lack of several functionalities and features posed difficulties in utilizing them in our course and study. Some tools may not cover basic data structures, have restrictive licenses that prevent customization for education and research, lack required controls for navigating through data structures and algorithms, or do not cover some topics in enough detail. In addition, some tools do not provide the essential levels of freedom for students, such as zooming and panning to focus on specific data regions, the ability to easily export the visualized data, or to save the history for future reference. Another issue encountered is the unavailability of all necessary visualizations for our course within a single package. For these reasons, developing an alternative tool tailored to our needs seemed the best alternative.

We designed and implemented a free online animation and visualization tool that complements the undergraduate Data Structures course and provides an interactive learning experience on a deep level. The AVDS tool we developed is the one we utilized in our study. The AVDS tool covers a classic set of algorithms and data structures commonly studied in the undergraduate data structures course. The source code for the tool is available on the GitHub repository (https://github.com/genakogan/data-structures-21-sce.git), allowing others to build upon or modify it as they see fit. The AVDS tool can be especially helpful for students seeking to understand the code behind the visualization, as they can see precisely how the algorithms and data structures are implemented. While understanding the source code may require some technical skills, such as programming experience with JavaScript, HTML, and CSS, the GitHub repository can be a valuable resource for educators and students seeking to gain a deeper understanding of algorithms and data structures through visualization.

Experimental procedure

The experiment involved three stages: (1) Survey preparation, (2) Survey conduction, and (3) Data analysis. Below, we provide an overview of each, with Fig. 1 illustrating the experimental pipeline.

Fig. 1
figure 1

The experimental pipeline

Full size image

Stage 1: Survey preparation. The preparation consisted of the following steps:

  • 1.1: Preparing the teaching staff. The teaching assistants received an explanation of the technology (AVDS) and its use, as well as an explanation of the experimental procedure.

  • 1.2: Preparing the tool demonstration. A short video presenting the AVDS tool was recorded before the experiments. This video was available to the students before the beginning of the experiment (stage 2.1). The video demonstrated the AVDS tool described in Sect. 3.4, which students used during the survey.

  • 1.3: Random assignment of students to experimental and control groups, with 39 participants in each.

  • 1.4: AVL lecture, which was taught in class a week before the experiment.

Stage 2: Conducting the survey. The experiment took place during the practical session, the next week after the AVL tree topic was introduced in the lecture. The experiment was carried out as a sequence of the following steps (summarized in Table 1):

  • 2.1: The students in both groups were provided with an explanation and clarification about how to use the AVDS online tool.

  • 2.2: The students in each group completed a TAM questionnaire.

  • 2.3: The students took the test: the experimental group used the AVDS tool, and the control group took the test without the AVDS tool. The test duration was approximately 20 min.

  • 2.4: The students in the experimental group completed a TAM questionnaire a second time.

  • 2.5: The students in both groups completed the SUS questionnaire.

Stage 3: Data analysis. The data analysis included cross-referencing data within the experimental group and between the groups, analyzing the results, and drawing conclusions.

Table 1 Conducting the survey
Full size table

Experimental design and variables

To analyze the results, we used integrated experimental sampling (De Datta, 1978). Integrated experimental sampling enables analysts to perform comparisons between groups as well as within the groups. There are three components in experimental design used to establish cause-and-effect relationships between variables by controlling for extraneous variables and reducing bias (Gribbons & Herman, 1996):

  • X—Activation of the manipulation, in our case, using the AVDS tool.

  • O—Indicates an observation or measurement of the dependent variables: \(O_1, O_2, \ldots , O_7\).

  • R—Random division into groups.

The study involved several observations that are defined in Table 2, including a pre-test and a post-test TAM questionnaire, a performance test, and a SUS questionnaire. Table 3 shows the steps of the experimental design used to evaluate the performance of the AVDS system. Using random division (R), the study included two groups: the experimental group (with access to the AVDS tool) and the control group (without access to the AVDS). The participants in both groups were given a TAM questionnaire (\(O_1\), \(O_2\), \(O_5\)). The experimental group used the AVDS tool (X) during the performance test, while the control group did not (\(O_3\), \(O_4\)). Additionally, both groups were given a SUS questionnaire (\(O_6\), \(O_7\)).

Table 2 Definition of observations
Full size table
Table 3 AVDS system performance evaluation—a pre- and post-test design
Full size table

Independent Variables. The independent variables were: (1) Use of the AVDS tool; (2) Perceived usability of the AVDS tool; (3) Perceived usefulness of the AVDS tool; (4) Perceived ease of use of the AVDS tool; (5) Attitudes toward the use of the AVDS tool; (6) Experience using the AVDS; (7) Group assignment (experimental/control).

Dependent Variables. Dependent variables included: (1) Evaluation of the usability of the AVDS tool; (2) Assessment of the usefulness of the AVDS tool; (3) Acceptance of the AVDS tool; (4) Performance in solving exercises; (5) Response time to complete questions.

Analysis and findings

The results of the questionnaire surveys, based on 78 student responses, were analyzed using a combination of statistical tools, including the SPSS 28Footnote 2 analysis software, IBM SPSS AMOS 27 module for structural equation modeling (SEM) analysis (Bahcivan et al., 2019; Blunch, 2012; Guo et al., 2022; Yildiz Durak, 2019), and KaleidaGraphFootnote 3 for graph construction.

System usability scale score—(H.1.)

The SUS scores were converted to percentile ranks and adjective ratings (such as "very good" or "poor"). The combination of percentile ranks and adjective ratings can provide a more precise picture of the usability of a product or system (Bangor et al., 2009; Brooke, 2013).

The percentile ranks of both groups are presented in Fig. 2 (left-experimental group, right-control group). The mean score of the experimental group was 79.17, which is considered "Excellent," while the control group had a score of 69.23, which is considered "Good."

These results suggest that the AVDS tool was perceived as more usable by the experimental than the control group, with higher levels of ease of use and user perception of effectiveness and efficiency.

Fig. 2
figure 2

Percentile ranks: left—experimental group, right—control group

Full size image

Technology acceptance model—(H2.1., H2.2. and H2.3.)

The study used a two-measurement model with three constructs: PU, PEOU, and ATU, depicted in Fig. 3. The TAM constructs are measured using multiple items, each of which measures a different aspect of the construct (see Appendix A, Table 10 for details).

Fig. 3
figure 3

The measurement model for the TAM questionnaire

Full size image

The validity and reliability of these measures are evaluated to improve confidence in the results and quality of these models:

  1. 1.

    First model with AVDS before (\(O_1\)) and after (\(O_5\)) the performance test (experimental group).

  2. 2.

    Second model without AVDS before (\(O_2\)) performance test (control group) and with AVDS after (\(O_5\)) performance test (experimental group).

This improves confidence in the results and quality of the models, ensuring accurate reflection of intended constructs and consistent results, thereby enhancing credibility and providing overall confidence in the model.

Improving and validating the model

The results of a Cronbach’s alpha reliability coefficient, shown in Table 4, indicate the internal consistency of the scale by measuring the degree of interrelatedness among the questionnaire items. The table displays the corrected item-total correlation for each item, and the Cronbach’s alpha result if the item is deleted from the scale. It suggests that the PEOU2 and PEOU4 items measure a similar construct and explain the high R-squared value. Removing PEOU2 from the model shown in Fig. 3 decreased the R-squared value of the attitude toward usage variable below one. Overall, the results demonstrate high internal consistency and reliability of the TAM questionnaire items.

Table 4 Items statistics
Full size table

Adding a double-curved arrow between two error terms can be a potential solution for a Heywood case (Wang et al., 2021). A Heywood case refers to a scenario in the SEM model where the estimated covariance matrix of latent variables produces unfeasible or unreasonable results, such as negative variances or correlations greater than 1.0. In Fig. 4, the solid double-curved arrow is used to represent a correlated error in the first model, and the dashed double-curved arrow is used for the second model. This indicates that the error terms share a common source of variance, which could be due to similar conceptual content. This approach can help improve the model’s fit and address issues with model identification.

Fig. 4
figure 4

Improved TAM models

Full size image

The results of model validation demonstrate a moderate to good model fit to the data, with reasonable fit indices (see Appendix B, Table 13), including RMSEA and RMR, which indicate a close model fit and a good fit, respectively. The factor loadings of the indicators (Appendix B, Table 14 ) were found to be greater than 0.4, demonstrating the construct validity of the instrument. The convergent validity was good, with AVE values greater than 0.5 (Appendix B, Table 15 ). The discriminant validity was satisfactory, with the exception of one case where the construct PEOU had poor discriminant validity to ATU (see Appendix B, Table 16 ). Finally, the constructs had acceptable consistency, as shown by the composite reliability values, which were greater than or equal to 0.6 (Appendix B, Table 17).

Hypotheses testing

The chi-square for the structural model was calculated before and after applying the weight constraints to the research hypotheses (Al-Aulamie, 2013). Table 5 shows the chi-square, degrees of freedom, and p-value for two structural models, before and after applying weight constraints, and the difference between them. The results indicate that the constrained models have a higher chi-square and lower p-value than the unconstrained models, suggesting that the relationships among the variables differ between cases at the model level.

Table 5 Chi-square difference
Full size table

After establishing a significant difference between the cases presented in Table 5, the next step is identifying the hypotheses that may account for these differences. Identification is accomplished by performing the weight constraints method separately on each hypothesis in two models.

For the first model, the first hypothesis (H2.1) was verified using the relationship between PEOU and ATU (see Table 6). The t-values calculated for PEOU and ATU are equal to 3.480 and 5.502, respectively, for before (\(O_1\)) and after (\(O_2\)) using AVDS. Both t-values are greater than the Z-score which is equal to 1.977. Thus, the relationship between variables related to PEOU of technology with an ATU is acceptable because it gives a positive and significant influence, meaning that hypothesis H2.1 can be proven acceptable. The standardized coefficient showed that to be the case after using AVDS, resulting in a stronger relationship (\(\beta\) = 0.755, p < .01) than before using AVDS (\(\beta\) = 0.752, p < .01).

In the second hypothesis (H2.2), by examining the relationship between PU and ATU, the resulting t-value of 2.567, which represents the result before (\(O_1\)) using AVDS, and 2.224 after (\(O_2\)) using AVDS, is stated to be greater than 1.977. It can be concluded that the relationship between variables on the construct of PU of AVDS technology is proven to have a positive and significant influence on ATU, which means that H2.2 is proven to be accepted. The standardized coefficient showed that the case before using AVDS has a stronger relationship (\(\beta\) = 0.345, p = .010) than after using AVDS (\(\beta\) = 0.274, p = .026).

The third hypothesis (H2.3) was tested using the relationship between PEOU and PU variables and produced a t-value of 4.665 > 1.977 after using AVDS. At the same time, the case before using AVDS showed that the t-value is less than the Z-score (t = 1.829, Z-score = 1.977; p = .067). Thus, the relationship between variables regarding the PEOU of technology on the PU has a positive and significant influence, meaning that H2.3 is proven to be accepted only after using AVDS (\(\beta\) = 0.726, p < .01).

Table 6 Significantly different hypotheses over cases \(O_1\) and \(O_5\)
Full size table

The structural model test revealed that the variance in the dependent variable PU that was explained by the independent variable PEOU, differed between \(O_1\) (12%) and \(O_5\) (53%). Additionally, the predictors PU and PEOU accounted for 98% of the variation in ATU for \(O_1\) and 95% for \(O_5\), according to the R square value, which means that this model is rational, although other unknown factors may impact dependent variables.

For the second model (see Table 7) hypothesis H2.1 was verified using the relationship between PEOU and ATU. The t-values calculated for PEOU and ATU equal 3.790 and 5.586, respectively, for the control group (case \(O_2\)) and experimental group (case \(O_5\)). Both t-values are greater than the Z-score, which is equal to 1.977. Thus, the relationship between variables related to the PEOU of technology with an ATU is acceptable because it gives a positive and significant influence, meaning that hypothesis H2.1 can be proven acceptable. A standardized coefficient showed that the control group (case \(O_2\)) had stronger relationship (\(\beta\) = 0.753, p < .01) than the experimental group (\(\beta\) = 0.740, p < .01) (case \(O_5\)).

Hypothesis H2.2 was checked out using the relationship between PU and ATU variables. In this case, it was found that the t-value for the experimental group is equal to 2.500, which is greater than the Z-score. At the same time, a comparison of the t-value with the Z-score carried out for the control group showed that, in that case, the t-value was less than the Z-score (t = 1.471, Z-score = 1.977; p = .141). Therefore, the relationship between variables on the construct of PU is proven to have a positive and significant influence on ATU, which means that H2.2 is proven to be accepted only for the experimental group (\(\beta\) = 0.302, p = .012) (case \(O_5\)).

Hypothesis H2.3 was tested using the relationship between PEOU and PU variables. The t-value of 5.618 > 1.977 (Z-score) for the control group (case \(O_2\)) and 4.415 > 1.977 (Z-score) for the experimental group (case \(O_5\)). Thus, the relationship between variables has a positive and significant influence meaning that H2.3 from both relations can be accepted. A standardized coefficient showed that the control group (case \(O_2\)) had a stronger relationship (\(\beta\) = 0.831, p < .01) than the experimental group (case \(O_5\)) (\(\beta\) = 0.696, p < .01).

Table 7 Significantly different hypotheses over cases \(O_2\) and \(O_5\)
Full size table

Additionally, the results of the structural model test showed that the impact of the independent variable PEOU on the dependent variable PU differed between cases \(O_2\) (12%) and \(O_5\) (53%). The predictors PU and PEOU were found to explain 95% of the variation in ATU for both cases, indicating that the model has a high level of explanatory power. However, it is important to note that the model may not account for other unknown factors that could also impact the dependent variables.

Multiple linear regression for performance—(H3.1. and H3.2.)

A significant difference existed in the test scores of the experimental and control groups (see Fig. 5). For the experimental group, the test scores were considerably higher: more than \(75\%\) of students who got grades from 80 to 100 belonged to the group that used the AVDS tool. The findings also indicate uniform distribution among the students that used the AVDS tool: 37 students out of 39 received grades above 80. Furthermore, students in the experimental group demonstrated significantly shorter response times.

Fig. 5
figure 5

Test and time performance

Full size image

Additionally, we used multiple linear regression to predict the performance test (Table 8). The model is developed for the following case: control group \(O_2\) without AVDS (i.e., before the test) and experimental group \(O_5\) with AVDS (i.e., after the test ), which answers the TAM questionnaire after using the tool. The performance test is based on four significant main effects: PU1 (p \(\le\) .027), ATU3 (p \(\le\) .004), PEOU3 (p \(\le\) .030), and group with (after) or without AVDS (before) (p \(\le\) .001). A significant regression equation was found (F(4.73) = 13.728, p <.001), with an adjusted \(R^2\) of 0.398. The predicted test performance of participants is equal to 0.424 + 9.431 (ATU3) − 4.415 (PU1) − 5.602 (PEOU3) + 32.428 (before and after). For the first main effect, higher perceived flexible use leads to a higher performance test (\(\beta\) = 9.431, SE = 3.137). In addition, it can be concluded that for the experimental group that used AVDS, the higher perceived improvement performance (\(\beta\) = \(-\)4.415, SE = 1.959) resulted in lower test performance. Similar to the previous findings, higher perceived easiness to the question PEOU3 (\(\beta\) = \(-\) 5.602, SE = 2.527) led to lower test performance. Generally, the test performance was higher for the experimental group (\(\beta\) = 32.428, SE = 4.9).

Table 8 Multiple regression for performance test
Full size table

The second multiple linear regression was calculated to predict performance time (Table 9). The model was developed for the groups described previously in the first regression. The performance time is based on three significant effects: (a) PEOU2 (p \(\le\) .008), (b) PEOU3 (p \(\le\) .035), and (c) group with (after) AVDS or without (before) AVDS (p \(\le\) .001). A significant regression equation was found: (F(3,74) = 15.700, p <.001), with an adjusted \(R^2\) of 0.364. The predicted performance test of participants is equal to 15.700–1.066 (PEOU2) + 0.756 (PEOU3) − 4.707 (before and after). For the first main effect, for the experimental group that used AVDS, the PEOU2 (\(\beta\) = \(-\) 1.066, SE = 0.388) led to lower performance time results. Moreover, the higher PEOU3 (\(\beta\) = 0.76, SE = 0.353) resulted in higher time performance. Mostly, the performance time decreased for the experimental group (ES = \(-\) 4.71, SE = 0.8).

Table 9 Multiple regression for performance time
Full size table

Limitations of the study

This study has limitations. Firstly, the small sample size may be unreliable, as RMR, NFI, GFI, RMSEA, and regression coefficients among variables are often done by Maximum Likelihood (Rao et al., 2011), which assumes normal distribution among the indicator variables. As a general rule, sample size requirements for SEM are often much larger than for other types of statistical analyses, such as multiple regression. A minimum sample size of 200 to 300 observations is often recommended for SEM, but the specific requirements will depend on the specific model and the data being analyzed. (Wolf et al., 2013). Secondly, this study may not fully capture the periodicity of AVDS usage. Therefore, the results of this study should be viewed as preliminary evidence for examining the relationship between students and their intention to use the AVDS system. Thirdly, the development of the site was based mainly on the course staff requirements, and was evaluated by the students’ course. Although, results show that most of the participants found the system with high usability, ∼ 1% of the participants in both groups rated usability as ’worst imaginable’. Therefore, the next phase of the development should take into account the requirements of the pilot outcomes and the students’ inputs.

Discussion

This research aimed to examine the efficacy of the AVDS tool in improving students’ learning and understanding of data structures and algorithms. The study found that the AVDS tool had a positive impact on students’ perception of usability, usefulness, and acceptance of the tool. The results showed that the AVDS tool was acceptable and effective for students’ learning.

The SUS score of the AVDS tool for the control group was 69.23 (in the "Good" category); for the experimental group, the score was 79.17 (in the "Excellent" category). This indicates that the AVDS tool is good and acceptable.

Using the TAM theory, the study explored the factors influencing the adoption of the AVDS tool by students. The findings suggested that students’ perception of the AVDS tool’s ease of use significantly affected their attitudes toward using the tool. If students found the tool easy to use, they were more likely to consider it useful and have a positive attitude toward integrating technology into learning. The study also showed that the AVDS tool improved students’ performance in terms of grades and completion time, such that \(\sim 95\%\) of them achieved grades above 80. These findings suggest that the developed AVDS tool was an effective learning medium that was easy for users to learn, understand, and operate. Overall, the study’s findings can be used to enhance understanding of user acceptance of the AVDS tool and lead to more successful adoption. The results of our study align with previous research (Andreev et al., 2021; Nathasya et al., 2019; Osman & Elmusharaf, 2014; Perháč & Šimoňák, 2022) which demonstrated that using visualized learning environments positively impacts student understanding, performance, and assessment completion times.

The application can help promote the learning of the Data Structures and Algorithms topics in the undergraduate computer science and software engineering courses. Beyond the common functionalities, the proposed tool provides valuable analytics for the teaching staff, e.g., usage reports for each topic. These reports can be utilized to prioritize more challenging topics during the course.

Conclusions and future work

This study investigated the effects of the AVDS tool, using a wide range of variables including TAM, SUS, and the students’ performance. According to the SUS results, we found that the experimental group perceived the usability of AVDS to be "Excellent", while the control group estimated it only as "Good".

Similarly, TAM findings showed that the experience with AVDS led to positive intentions to use the tool and higher levels of ease of use, effectiveness, and efficiency. Moreover, there is evidence that the use of AVDS elevated the overall learning experience and achievement levels. The students in the experimental group obtained higher and more uniform scores than those in the control group, and it took them less time to solve tasks than the control group.

At present, our team is actively enhancing the application by improving its functionalities and addressing usability issues identified in the current study. Recently we conducted a second additional pilot to evaluate the effectiveness of the additional developments and now we are analyzing its outcomes. The current version of the website may be accessed at https://ds-sce.sce-fpm.com/.