Introduction

A focus of discussion and a subject of research at many universities involves innovative digital technologies, future skills, and particularly topics in computer science (e.g., artificial intelligence, extended reality, and robotics) (Autorengruppe Offensive Digitale Schultransformation, 2020; Gadeib & Noller, 2021; Köller et al., 2022; Suessenbach et al., 2021). To remain competitive and drive innovation in these areas, talents in various science, technology, engineering, and mathematics (STEM) fields need to be fostered and retained in every country (acatech & Joachim-Herz-Stiftung, 2022; Anger et al., 2021; Statistik der Bundesagentur für Arbeit, 2019, 2020). This is especially true for academic professions (acatech & Joachim-Herz-Stiftung, 2022; Statistik der Bundesagentur für Arbeit, 2020). Therefore, numerous STEM support programs exist throughout Germany, which aim to inspire students in different age groups to pursue careers in mathematics, science, computer science, or technology (Brenning & Wolf, 2021; Nickolaus et al., 2018). The support programs mentioned in two reviews by Brenning and Wolf (2021) and Nickolaus et al. (2018) were focused on various target groups and learning content, including school activities and extracurricular educational settings, and were predominantly designed for short intervention periods. An overview study of the effects of extracurricular STEM programs found that short-term support programs often do not have any stable effects on the participants’ interest in STEM (Nickolaus et al., 2018). In the case of school-integrated support programs, positive short-term effects were found (Nickolaus et al., 2018). However, sustainable effects could not be measured for the STEM programs analyzed. Nevertheless, the authors of this overview study saw great potential in supporting students’ interest in STEM through school-integrated support programs (Nickolaus et al., 2018).

STEM support programs try to spark or foster students’ interest in STEM in one way or another. One approach holds that new ways of experiencing things are evoked in individuals through examining possible new situationally interesting STEM learning objects. This can lead to experiences that cause students to engage with objects from the mathematical, natural, and technical sciences again at a later date of their own accord, without external incentives and without externally controlled reasons to act. Such engagement can help to develop a subjectively meaningful relationship with STEM topics under certain circumstances. An internalization process can be set in motion in the context of career choices, at the end of which individuals can acquire a “genuine” dispositional interest in STEM professions (Holland, 1997; Krapp, 1992, 1998). There is only limited knowledge about the learning characteristics (e.g., STEM interests, vocational interests) of participants in long-term school-based STEM programs (Neher-Asylbekov & Wagner, 2022; Nickolaus et al., 2018), and a systematic description of the learning characteristics of students who voluntarily participate in STEM programs seems interesting. In the context of this research aim, using intra- and intergroup comparisons, we examined differences and similarities between groups of students who participated in one of three long-term STEM support programs and a control group (CG) regarding their actual interest. In this paper, we first discuss the theoretical background, then the state of research on STEM support programs and their effects (mainly in Germany, as the three support programs are in Germany and Austria), the methodology, and the results. The paper concludes with a discussion of the results and the limitations of the study and elaborates on possible implications for future STEM programs.

Background

In contemporary interest research, the term “interest” is defined as a construct that characterizes a meaningfully prominent relationship between a person and an object (Krapp, 1992). Two lines of research can be distinguished regarding the concept, which define “interest” differently. The first line considers interest as a personality-specific characteristic of the learner, e.g., a relatively stable trait-like preference for a certain field of knowledge or action (e.g., the STEM field). This type of interest is called individual or personal interest. Individual interest-oriented actions are regarded as current realizations of a general personality trait or a time-lasting attitude toward an object area. Personal interest is expressed, among other ways, by a tendency to engage with a particular object repeatedly, joyfully, and without an external cause. In this context, interest has a prominent subjective significance and forms an essential part of the self-concept. For example, interest in a specific subject (e.g., technology) can be understood as an individual interest. The second line of research considers interest as situational, which is not dependent on a dispositional preference for an object; it results from a (learning) situation or a (learning) object and engenders more intensive attention on the part of the individual (Krapp, 1992; Mitchell, 1993). This means that through “skillful” pedagogical preparation of the subject matter in support programs by teachers, a favorable motivation to learn can be generated in students.

Usually, both individual and situation-specific conditions can be triggering factors for the development of interest in support programs (Krapp, 1992). Regarding the long-term support of interest, Mitchell (1993) differentiated between two components: “catch” and “hold.” “Catch” components trigger situational interest, e.g., through the learning situation, and “hold” components hold the acquired situational interest (Krapp, 1992; Mitchell, 1993). Subsequently, individual interest can develop through various stimuli. According to Krapp (1992), this interest lasts over time and has a significant influence on learning.

Moreover, according to situated expectancy-value theory (SEVT), interest is a significant factor influencing career choice decisions (Eccles & Wigfield, 2020). Influencing factors of career choice decisions, according to SEVT, are the expectation of success and task values, further divided into attainment value, intrinsic value, utility value, and cost (Eccles & Wigfield, 2020; Wigfield & Eccles, 1992). Intrinsic values, as described in the previous paragraph, are linked to the construct of interest. Many studies have shown correlations between these factors (e.g., Nagengast et al., 2011). Additionally, gender aspects are integrated into SEVT, and many studies have shown that the career choices of female and male students differ (Bahr & Zinn, 2023; Eitemüller & Walpuski, 2018; Mauk, 2016; Nagengast et al., 2011; Neher-Asylbekov & Wagner, 2022; Wigfield & Eccles, 1992).

Among the various factors influencing career choice decisions, vocational interests are crucial, according to Holland (1959, 1997) and Volodina et al. (2015). Instead of asking students about their interest in STEM subjects to indicate their interest in a field of study, it is also possible to ask students about their interest in specific tasks. Therefore, an instrument was developed to ask working people about their tasks, such that students’ vocational interests could be assessed in a different way, and each task is then related to an area. Based on the activities in different fields/working environments, different domains were created. Holland (1997) classified six areas: realistic (R, technical), investigative (I, research), artistic (A, creative), social (S, educational), enterprising (E, selling), and conventional (C, administrative). This resulted in a hexagonal model to visualize interest profiles (Holland, 1997). The model can be visualized in a circular circumplex structure (Holland, 1997; Junkuhn & Nagy, 2022). Prediger (1982) proposed two coordinate axes: People–Things for the x-axis, from Social to Realistic, and Data–Ideas for the y-axis, from Enterprising and Conventional to the middle of Artistic and Investigative. Figure 1 shows a circumplex representation adapted from Prediger (1982) as described above.

Fig. 1
figure 1

Circumplex representation of vocational interests with RIASEC using Prediger’s People–Things and Data–Ideas axes (Prediger, 1982)

Full size image

The RIASEC areas are located on the circle starting from 0° (Realistic) and proceed counterclockwise in 60° steps (investigative, artistic, social, enterprising, conventional). The localization of interest areas (e.g., investigative) can be a mix of two poles of Prediger’s axes, Ideas and Things (Junkuhn & Nagy, 2022).

The six RIASEC areas, in turn, can be assigned to concrete occupations and can be used to assess the vocational interests of students. The vocational interests in the six areas are according to the types of personalities in RIASEC theory (Holland, 1997). Hence, they represent stable personality traits. These types can be seen as individual interests. Therefore, in the context of this study, we considered both the STEM interests and vocational interests of students, which can be seen as individual interests.

The authors found that this background, while including multiple perspectives on students’ interests, provided a comprehensive picture of interests, vocational interests, and career choice decisions for STEM support programs. In the next section, we discuss the current state of research regarding the knowledge about STEM support programs (mainly in Germany).

The Current Study

On average, students’ interest in STEM subjects declines over the school years (Christidou, 2011; Hoffmann & Häussler, 1998; OECD, 2007; Sjøberg, 2000; Sjøberg & Schreiner, 2010; Tröbst et al., 2016). In order to counteract this development, various extracurricular STEM programs have been created (Brenning & Wolf, 2021; Hausamann, 2012; Neher-Asylbekov & Wagner, 2022; Nickolaus et al., 2018; Suviniitty & Clavert, 2020). A meta-analysis by Brenning and Wolf (2021) found no significant effects of out-of-school STEM support programs (higher education domain) on the career choices of female students. Assessing whether students liked a project or not cannot imply that the project contributed to their career choices (Brenning & Wolf, 2021). In an analysis of various studies on STEM support programs, Nickolaus et al. (2018) found that mostly short-term effects, if any, on the interest of students could be proven, although the participants in these programs evaluated them positively; long-term effects could not be proven in any of the studies considered. School-integrated programs are considered to have more potential to increase students’ interest in STEM due to short-term effects on constructs such as interest (Guderian, 2007; Nickolaus et al., 2018; Sahin, 2013; Sahin et al., 2014). In this context, afternoon clubs are linked to the content of regular lessons. Thus, theoretically, a cumulative effect of various measures and social influences should have an impact on students’ interest (Nickolaus et al., 2018; Sahin, 2013; Sahin et al., 2014), in accordance with SEVT (Eccles & Wigfield, 2020; Nagengast et al., 2011). In particular, mentoring programs should have a positive effect on students’ educational choices (Kuchynka et al., 2022; Nickolaus et al., 2018). Two literature reviews (Neher-Asylbekov & Wagner, 2022; Wang & Degol, 2013) noted that gender should be explored in future research as it may play an important role in one’s career identity or educational choices (Schelfhout et al., 2021; Stringer et al., 2019). Additionally, the influence of digitalization merits exploration (Neher-Asylbekov & Wagner, 2022).

The research on factors that influence K–12 students’ career choice decisions considers students’ subject interest and self-concept as important factors (Bahr & Zinn, 2023; Eitemüller & Walpuski, 2018; Jann & Hupka-Brunner, 2020; Mauk, 2016; Wang & Degol, 2013). The influence of parents, in the form of stereotypes and (parental) expectations, is also important to consider (Tegelbeckers et al., 2019; Wang & Degol, 2013). Furthermore, social influencers, career prospects, and teachers are important, as mentioned in a literature review by López et al. (2023).

According to Volodina et al. (2015), the RIASEC scale is suitable for anticipating educational choices. According to Babarović et al. (2019), STEM professions are assigned to the Realistic (R) and Investigative (I) domains. Junkuhn and Nagy (2022) found that the general secondary schools in Baden-Württemberg, Germany, were not assumed to promote the vocational interests of K–12 students. The congruence hypothesis, according to them, holds that vocational schools promote the vocational interests of students in the respective areas (Junkuhn & Nagy, 2022). Few studies on whether school-integrated STEM support programs complement the regular school curriculum and promote vocational interests in the general education sector have been conducted (Nickolaus et al., 2018). Studies on STEM support programs for students with an existing interest in STEM or particularly high achievements in STEM could show no or limited effects on interest or other constructs (e.g., self-concept, motivation) (Schütte & Köller, 2015; Sumfleth & Henke, 2011). In their study, Dierks et al. (2016) reported a trend that high-achieving students have a greater interest in science than other students. This could indicate that STEM support programs that select students by grade already include high-achieving students interested in STEM. In their study, Blotnicky et al. (2018) showed that career choice decisions depend on, among other things, STEM career knowledge. In summary, according to the findings, only a few studies have looked at school-integrated long-term STEM support programs and the vocational interests of K–12 students in Germany. To expand the scientific discourse in the field of STEM interest research, this paper focuses on characterizing students’ individual interest in STEM across different age groups and different support programs. It would be expected from the research that the mean values of R and I for students already enrolled in STEM subjects at universities are high. Based on this objective, the paper aims to answer the following research questions:

  • RQ1: To what extent do students differ in their STEM interests and vocational interests in the realistic (R) and investigative (I) domains based on groups of students in three STEM support programs and a control group?

  • RQ2: To what extent do students differ in terms of the self-reported factors that influence their educational choices among groups of students in three STEM support programs compared to the control group?

  • RQ3: How can the RIASEC profiles of K–12 students in STEM support programs (IMP, Digi, Hari1) be compared with those of STEM students enrolled in a university (Hari2) and the control group?

Methodology

This aim of the study was to provide systematic descriptions of the vocational interest profiles of students participating in three school-based STEM programs. In addition, we sought consistent aspects of vocational interest that characterized the group of existing STEM students that could be identified among students in the three STEM programs. The students’ STEM and career interests were considered as possible factors influencing their STEM study/career choices (Holland, 1959; Prenzel & Drechsel, 1996). Three approaches to fostering students’ interest in STEM were considered in this study: the DIGISTEM school trial at the Otto-Hahn Gymnasium (OHG) in Nagold, Germany (Hamann, 2020); the profile subject computer science, mathematics, and physics (IMP) in Baden-Württemberg (Ministerium für Kultus, Jugend und Sport Baden-Württemberg, 2018); and the Dr. Hans Riegel Subject Prizes funding program for German and Austrian schools and universities (Alfter, n.d.). The data were collected in the process of scientifically monitoring these three programs. The selected programs could be compared because they use identical scales in the questionnaire instruments for their respective data collection. In addition, the composition of the sample made it possible to compare the interest profiles of students who decided to study STEM and those who decided on a STEM program at the end of middle school in grade 10 up to the Abitur (German A levels).

Descriptions of the Support Programs

The analysis was focused on two long-term school-integrated STEM programs and an in-depth STEM course. The latter was a course called informatics, mathematics, and physics (IMP) in Baden-Württemberg, Germany, which was offered as a four-hour compulsory elective subject from grades 8 to 10 as a supplement to standard mathematics and physics lessons (MKJS, 2018). The content areas of the computer science part are data and coding (e.g., efficient storage of data), algorithms (e.g., basic components and software projects), computers and networks (e.g., logic gates and the foundation of data transmission), information society, and data security (e.g., encryption algorithms) (MKJS, 2018). The content areas of the physics and mathematics part are connected to the computer science content areas. A detailed explanation of the content of the support program can be found in the Supplementary Material. In the following, we refer to it as support program 1 (IMP) with subgroup 1 (IMP).

For the second group (Digi), the authors considered students who opted for a support program in the context of digital media and STEM as part of a school trial at a secondary school in Baden-Württemberg. This program aims to promote students’ career and study choices at various levels. For this purpose, digital learning can be embedded more deeply in STEM lessons in grade 10. Tablets were meaningfully integrated into the STEM courses, serving as an incentive to engage students in the Digi program. At the course level, a basic STEM + course up to the final state exams (A levels) and a seminar course (engineering academy) in cooperation with local Youth Research Centres tie in with the secondary school curriculum. The program was supplemented by an extracurricular working group, in which students conduct workshops with lecturers from industry and research. Additionally, the program includes the possibility of internships at engineering and computer science companies as well as project days at universities and colleges (Hamann, 2020). We refer to this as support program 2 (Digi) with subgroup 2 (Digi).

The third group (Hari) included participants in a support program for students in upper secondary school who have written a subject-specific paper in the field of mathematics, natural sciences, computer science, or technology and have taken part in a school competition (Alfter, n.d.). After winning in their regional group, the participants received funding for several STEM workshops and networked with the cooperating university. We refer to this as support program 3 (Hari), with subgroup 3 (Hari1) as current participants and subgroup 4 (Hari2) as graduates.

Although the three support programs had different areas of focus in terms of content and implementation, all three were characterized by (1) intended support for STEM interest, (2) longer duration, and (3) school-based organization. An overview is given in Table 1.

Table 1 Descriptions of students in support programs
Full size table

The students in the three support programs engaged in different STEM activities related to the Investigative domain. Hari students wrote a scientific paper, Digi students used a tablet for their STEM research, and IMP students dealt with topics such as simulation and numerical methods. Hence, the authors assumed that the median values for many of these students in the Investigative domain would be high.

Samples and Procedure

The overall sample (N = 1170) was composed of five subgroups of the support programs. In addition to these, two subgroups were included to enable intergroup comparisons on the interest profiles of STEM students (subgroup 4, Hari2) and students who were considered to be average (subgroup 5, control group). Besides the surveyed groups, the maximum number of possible students in the programs was assessed based on voluntary participation.

  • Subgroup 1 (IMP): The data of the students in support program 1 (IMP) were collected at 31 general secondary schools in Baden-Württemberg from November 2021 to July 2022. A total of 371 10th grade students (m = 258, w = 96, d = 13, NA = 4) (age: M = 16.17, SD = 0.69) in this program took part in the survey. The response rate was approximately 30% for all schools with the IMP program.

  • Subgroup 2 (Digi): The data of the students in support program 2 (Digi) were collected at the OHG Nagold in Baden-Württemberg from October 2019 to October 2022. A total of 54 10th grade students (m = 22, w = 31, d = 1) (age: M = 15.74, SD = 0.48) participated in the survey. This was a full survey.

  • Subgroup 3 (Hari1): The data of the students in support program 3 (Hari) were collected from January to June 2019. A total of 199 students (m = 81, w = 118) (age: M = 17.27, SD = 1.09) participated in the survey. The response rate was approximately 22.5% of all competition participants.

  • Subgroup 4 (Hari2): A total of 382 awardees of the Dr. Hans Riegel prize (Hari) (m = 211, w = 170, NA = 1) (age: M = 20.19, SD = 2.36) took part in the study. The response rate was approximately 33%. The participants had been awarded prizes by various universities between 2010 and 2018 and were enrolled in a university.

  • Subgroup 5 (CG): A total of 164 10th grade students (m = 62, w = 89, d = 13) (age: M = 15.88, SD = 0.54) comprised the control group. It is important to note that all 10th grade students who were not enrolled in the Digi program were included in this group. No student was excluded. Since participation in the Digi program was limited, not all high-performing students were in the Digi subgroup. Instead, they had average profile subjects (natural science and technology, Spanish, Latin, physical education) and represented an “average” student sample compared to other 10th grade students from Germany. The data were collected in the same period as the Digi sample at the OHG in Nagold, Baden-Württemberg (October 2019 to October 2022).

The data for IMP, Hari1, and Hari2 were collected in two online surveys. The data for Digi and CG were collected with a paper questionnaire. All surveys were anonymous and voluntary, as explained to the schools, parents, and students in the cover letter and right before the survey. We excluded students who categorized themselves as “other” on the binary gender question, as the sample was too small to compare it with others.

Repeated testing was done to compare the STEM and vocational interests in the Realistic and Investigative domains (six variables) for the control group (CG) with every other group (e.g., CG with IMP, CG with Digi), so overall the total number of comparisons was 24. The Bonferroni corrected (Dunn, 1961) significance level was \(\alpha =0.00208\) and was used for all tests; only results below that value was considered significant (one-sided test).

Instruments

The present study used several constructs. Subject interest (based on the interest construct according to Krapp (1992)) was assessed with a 5-point rating scale from 1 = very uninteresting to 5 = very interesting in response to the prompt: “For each subject, please indicate how interesting you find this subject.”

The factors influencing educational choices (adapted from Eitemüller and Walpuski (2018) and Mauk (2016)) were assessed with a 10-point rating scale from 1 = not important to 10 = very important. The factors were based on SEVT (Eccles & Wigfield, 2020). Students were asked to evaluate the following factors in terms of significance to their career or study aspirations: school lessons, people (such as parents, educators, and friends), extracurricular programs (such as the programs mentioned above), interest and motivation, self-concept, sociological influences (e.g., gender, stereotypes, perceptions), career perspectives, content of the lessons, and place of study or work. Table 2 presents the descriptive values of the items.

Table 2 Test quality criteria of factors influencing educational choices
Full size table

Vocational interests were assessed using the scale by Mokhonko (2016) adapted from the AIST-R (Bergmann & Eder, 2005), according to Holland (1959, 1997) and Jörin et al. (2006). As the first evaluation (Hari) was done using that scale, the same instrument was used in subsequent studies so that the vocational interests of students in the different programs can be compared. Example items from each category are presented in Table 3.

Table 3 Example items of applied RIASEC scale
Full size table

An overview of the test quality criteria for vocational interests can be found in Table 4. Cronbach’s α values for internal consistency were acceptable to good (Taber, 2018). As we used tested and validated scales, we used Cronbach’s α. The data were analyzed with SPSS 28.0.0.0 (IBM Corp., 2021) and R 4.3.1 (R Core Team, n.d.) as well as the tidyverse (Wickham et al., 2019), car (Fox & Weisberg, 2019), rstatix (Kassambara, 2023), holland (Heine & Hartmann, 2021), and caret (Kuhn, 2008) packages. If a number reported for different variables in the results section differs from the overall sample size, then some answers are missing. The authors excluded no students from the analysis.

Table 4 Test quality criteria of applied RIASEC scales
Full size table

Interest Profiles

The vocational interests of students across all six facets were modeled as interest profiles following the methodology outlined by Junkuhn and Nagy (2022). Three coefficients were computed based on the six RIASEC variables for students in the subgroups IMP, Digi, and Hari1, as well as the STEM group (Hari2), non-STEM group (Hari2), and control group. These coefficients were derived using the methodologies of Junkuhn and Nagy (2022) and Gurtman (2009). One of the coefficients represents the profile level of each student, while the other two coefficients represent the position of the interest profile within the RIASEC circumplex, determined by the coordinate weights on the People–Things and Data–Ideas axes, using the z-standardized RIASEC variables. The formulas, as well as their explanations, can be found in Junkuhn and Nagy (2022).

Adopting the methodology outlined by Junkuhn and Nagy (2022), which incorporates Gurtman’s (2009) approach, ensured a robust and comprehensive analysis of vocational interests in this study. This methodology enabled the localization of vocational interest profiles across various student subgroups within the circumplex model, making it an appropriate approach for the current investigation.

Missing Data

For Hari1 and Hari2, there were no missing data for the RIASEC scale. To calculate the amount and category of missing data for the remaining sample, Little’s MCAR test was used, assuming data missing not at random (MNAR) (Schlomer et al., 2010). The results show that 0.08% were missing (\({\chi }^{2}\) = 529.62, p = 0.130 (df = 494)). Thus, the dataset could be analyzed (Allison, 2001).

Results

Characterization of Students in Different STEM Programs (RQ1)

The authors assumed that students who chose various STEM programs had unique STEM interests. Table 5 presents an overview of STEM subject interest for the individual subgroups. Students in IMP had the highest interest in mathematics, physics, and computer science. This aligned with expectations, given that students chose the 4-h IMP program. The median values for Digi suggest an equivalent level of interest in STEM subjects. This observation reflects the integration of the program within the STEM subject context, offering an additional extracurricular learning opportunity with a focus on computer science and digitization. Hari1 recorded the highest median scores for mathematics, computer science, and physics among all subgroups. The median for biology was also notably high compared to those for other subgroups. Considering that this support program was a competition for which students actively applied and were selected by professors, the highest values for subject interest overall appeared plausible. With the highest possible values in biology and physics and the second-highest values in other STEM subjects, Hari2 showed only slight differences from Hari1 in terms of STEM interest. Thus, at the culmination of the support programs, students exhibited high STEM interest. Additionally, it was assumed that the samples included students with diverse vocational interest profiles, as indicated by the sometimes large interquartile range (IQR).

Table 5 Scores for STEM subject interest, vocational interest, and R and I among students in subgroups
Full size table

According to the current research and the theoretical background, it is considered advantageous for vocational choices in the STEM context if the values for interest among students participating in STEM support programs are high in the Investigative and Realistic domains (Babarović et al., 2019). For the subgroups IMP and Hari2, interest in these domains was higher compared to CG (Fig. 2). Additionally, the median values among female students in this study were higher for Enterprising compared to other domains, while the median values among male students were higher for Enterprising than Conventional, Social, and Artistic domains. CG demonstrated a higher interest in the Social and Artistic domains but a lower interest in Investigative and Realistic. The median values among male students enrolled in university programs (Hari2) were the highest for the Investigative domain (Fig. 2). Conversely, the median values among male students in CG were the highest for the realistic domain. Gender-based analysis revealed that values among male students in both support programs and CG were higher for the Realistic and Investigative domains, whereas the values among female students were higher for the Artistic and Social domains. Consequently, an analysis of differences in STEM and vocational interests was conducted by separating the groups based on gender. As reported in the “Methodology” section, all tests were conducted with the correct alpha level.

Fig. 2
figure 2

Vocational interests of different groups

Full size image

The separately calculated differences in STEM and vocational interests in R and I among female students in the support programs and CG can be found in table in the Appendix. To better visualize the results, the Likert scale for CG is shown in Fig. 3, and the scales for all four groups are shown in Fig. 4. Significant differences between all groups (IMP, Digi, Hari1, Hari2) and the control group (CG) were found for the subject mathematics, computer science, and physics, and vocational interests in the Investigation domain (Appendix). In the subject biology, the group differences between CG and Digi and between CG and IMP were not significant; significant differences were found for Hari1 (p < 0.001) and Hari2 (p = 0.0015). These differences could be explained by the school’s influence on CG students. The interests of female students in all three support programs in this study did not statistically differ from those of the control group in the realistic domain (Appendix). However, the interests of female students differed for the investigative domain compared to CG, with medium and large effect sizes (see Fig. 2) (Fritz et al., 2012; Tomczak & Tomczak, 2014). Furthermore, medium and large effect sizes were found among female students in the support programs with a higher interest in mathematics, computer science, and physics (Figs. 3 and 4). The analysis of group differences between female students in the different support programs showed no differences in the STEM interests of those in IMP compared to Digi. The median values among female IMP students were significant lower for investigative compared to female Hari1 students (U = 3136, − 4.39, p < 0.001, r = 0.31). The analysis also showed significantly higher median values among female students in Hari2 compared to IMP for Investigative (U = 3748, Z =  − 5.75, p < 0.001, r = 0.37), physics (U = 3213, Z =  − 3.76, p < 0.001, r = 0.27), and biology (U = 3013, Z =  − 4.88, p < 0.001, r = 0.35). Female Digi students did not differ from Hari1 students in their interest in STEM subjects.

Fig. 3
figure 3

STEM interests of female students in the control group

Full size image
Fig. 4
figure 4

STEM interests of female students in different groups

Full size image

The analysis of differences among male students in the support programs compared to the CG (appendix) was similar to the one for female students. The interests of male students in the support programs did not differ from those of the CG regarding in the realistic domain (Fig. 2) (Appendix). Only small effect sizes were found for the differences between male IMP students in their interest in mathematics, physics, and biology compared to CG students. Hari2 students differed from CG students in their interest in mathematics, computer science, physics, and investigative. Digi, Hari1, and Hari2 students differed from CG students in the investigative domain. Both differences were significant (Appendix). Significantly lower interest in investigative (U = 5491, Z =  − 5.05, p < 0.001, r = 0.28), physics (U = 5010, Z =  − 4.15, p < 0.001, r = 0.21), and biology (U = 2512, Z =  − 3.69, p < 0.001, r = 22) was found for male IMP students compared to Hari1 students. Male IMP students had a lower interest in investigative (r = 0.49), mathematics (r = 0.24), computer science (r = 0.29), physics (r = 0.33), and biology (r = 0.27) than Hari2 students. All differences were significant (Appendix). The interests of male Digi students did not differ from those of male Hari1 students. The interests of male Hari1 students did not differ from those of male Hari2 students.

Due to the somewhat different sample sizes, the group differences were calculated with the Mann–Whitney U test (Kühnel & Krebs, 2001; Mann & Whitney, 1947; Stonehouse & Forrester, 1998). For the effect sizes for non-parametric testing (Mann–Whitney U test), the effect size r was calculated (Fritz et al., 2012; Tomczak & Tomczak, 2014). The results show that the program groups could be distinguished from CG with respect to their individual STEM interest. In addition to subject interest, students’ vocational interest was found to be important for their career choices. For clarity, the vocational interests of the entire sample are shown in Fig. 2. After reviewing the STEM differences in Figs. 3, 4, 5 and 6, it seemed appropriate to split the analysis by gender due to the different distributions.

Fig. 5
figure 5

STEM interests of male students in the control group

Full size image
Fig. 6
figure 6

STEM interests of male students in different groups

Full size image

Factors Influencing the Educational Choices of Students (RQ2)

In order to find differences between the factors influencing students’ educational choices regarding different programs or subjects, various factors were considered (Fig. 7). The focus of the original analysis of Hari1 did not include factors influencing educational choices. Hence, only the subgroups IMP, Digi, Hari2, and CG were analyzed. The subgroups were plotted on the y-axis. A boxplot was constructed to visualize the 10 factors influencing educational choices according to the students’ ratings. In line with the state of research, interest (interest and own motivation, Fig. 7) and self-concept (self-confidence in one’s own abilities) were found to be the factors with the biggest influence (Eccles, 2009; Eitemüller & Walpuski, 2018; Mauk, 2016). The content of school subjects (different and new topics offered by the IMP program, described in the data section and the Supplementary Materials) was another influencing factor for students in IMP and Hari2 compared to CG in this sample (content of the lessons, Fig. 7). Compared to the control group, experience with extracurricular programs (e.g., Digi, Hari) had a greater influence on the educational choices of students in other groups. A significant difference between Digi and CG was found with respect to this characteristic (U = 871, p < 0.001, Z =  − 4.37, r =  − 0.41). A difference was also found between Hari2 and CG (U = 4611, p < 0.001, Z =  − 6.42, r =  − 0.33) as well as IMP due to the influence of the content (U = 5631, p < 0.001, Z =  − 6.47, r =  − 0.32). No gender differences were found among IMP students regarding the factors influencing educational choices (Bahr & Zinn, 2023). Furthermore, no gender differences were found regarding the factors influencing educational choices. Other factors, such as the place of study or work, were similar for the different subgroups. Sociological influences such as gender, stereotypes, and attitudes were also similar for all subgroups and were quite low compared to other factors influencing educational choices. Career perspective was considered an important influencing factor among students, aligning with the premise that the students chose a support program or not because of their vocational interests.

Fig. 7
figure 7

Factors influencing educational choices of different subgroups

Full size image

Interest Profiles of the Different Subgroups (RQ3)

Figure 8 shows a visualization of the interest profiles among the subgroups IMP, Digi, and Hari1 (students in support programs) and Hari2 (former students in support programs), split into students who were enrolled in a STEM subject (STEM) and students who did not chose a STEM course at a university (non-STEM) as well as students in the control group (CG). The CG was placed in the center, as it more or less represents “average” students. Then, the mean values for the interest profiles of the subgroups were calculated according to the methods of Junkuhn and Nagy (2022) and Gurtman (2009). The interest profiles of students who did not choose a STEM course in university tended to be close to those for the CG with an orientation toward the Artistic and Social domains (Fig. 8). This is in line with expectations. They were also more oriented toward People on the People–Things axis.

Fig. 8
figure 8

Visualization of mean RIASEC interest profiles of students in different subgroups

Full size image

The interest profiles of students in the IMP, Digi, and Hari1 support programs were oriented toward the Ideas and Things axes as well as the Investigative domain. The interest profiles of STEM students, in line with RIASEC theory, were oriented toward Ideas and Things as well. Their interest profiles were closest to Investigative. The interest profiles of students in IMP were the closest to those of STEM students.

The interest profiles of subgroups were split by gender, as the different RIASEC interests of students in this study varied between female and male students. Figure 9 shows the interest profiles of female students in each subgroup.

Fig. 9
figure 9

Visualization of RIASEC interest profiles of female students in different subgroups

Full size image

At the center of the calculations were the female students in the control group. Female non-STEM students enrolled in university were located at the biggest distance from the realistic domain and Things. The interest profiles of female students in the support programs (Digi, Hari1, IMP) tended toward Investigative and were on the Ideas and Things axis. The interest profiles of female STEM students were strongly oriented toward Investigative. Female IMP students were located the closest to enrolled female STEM students.

The interest profiles of male students are visualized in Fig. 10.

Fig. 10
figure 10

Visualization of RIASEC interest profiles of male students in different subgroups

Full size image

In contrast to female students in the support programs (Digi, Hari1, IMP), the mean values for the interest profiles of male students were located close to those of the male students in the control group (center of Fig. 10). All interest profiles of male students in the support programs and male STEM students tended toward Things on Prediger’s axis and were located in the Ideas and Things area. The interest profiles of male non-STEM tended toward Enterprising and the People and Data area. Overall, the interest profiles of male students in the support programs (Digi, Hari1, IMP) were close together.

Discussion

The median interest in mathematics, computer science, and physics among IMP students was high, as expected after choosing the profile subject (Table 4). This was similar among students in Digi, with the addition of a notable interest in biology. Likewise, students in Hari1 demonstrated high or even higher interest in STEM subjects. Notably, students in the Digi and Hari support programs demonstrated significantly higher interest in the Investigative domain than the average students in the control group (CG). Students in the Digi program received talks and internships from companies. IMP students, due to the interdisciplinary nature of the subject and vocational orientation being included in the curriculum, could receive some guidance regarding their vocational orientation. Students in the Hari program had contact with STEM university mentors. Therefore, it can be inferred that these three programs sustained (or perhaps even cultivated) the vocational orientation of the students (RQ1). In examining the vocational interests of students in Digi, a broader perspective across different domains was apparent. Interest in the artistic and investigative domains stood out for female students compared to other vocational interests among the subgroups (Fig. 2). Similarly, an interest in the artistic domain was higher for male students compared to other subgroups.

The factors influencing educational choices were consistent with the state of research outlined in the results (RQ2). The indication by IMP students that the content of their lessons contributed to their educational choices supports the notion that the profile subject could stimulate or maintain their STEM interests. The statistically significant difference in the influence of extracurricular programs on educational choices between students in Digi and Hari1 compared to the control group suggests that these programs may have had an influence. However, given the small size of the control group, this result should be validated with a larger sample in future studies. Gender differences in the factors influencing educational choices were observed, but due to the small sample size or the small effect size and variations across support programs, no general conclusions can be drawn.

To address RQ3, we computed the mean values for interest profiles of all subgroups and positioned them on the circumplex model (Fig. 8). The findings reveal that the mean values for the interest profiles of students in the support programs differed from those of the control group and students who did not pursue a STEM course in university. The highest mean value was found for STEM students in the Investigative domain, situated in the region of Ideas and Things, according to Prediger’s (1982) classification.

This study emphasizes the importance of examining gender differences in support programs, as suggested by literature reviews (Neher-Asylbekov & Wagner, 2022; Wang & Degol, 2013). The STEM interests of students in the support programs varied between female and male students (Figs. 3, 4, 5 and 6). Consistent with existing research (Bahr & Zinn, 2023; Jann & Hupka-Brunner, 2020; Nickolaus et al., 2018; Sjøberg & Schreiner, 2010), male students showed a greater interest in physics, while female students exhibited a stronger interest in biology. Among the overall large sample of 1170 students, the interest profiles of those in the support programs and STEM tended toward the area between Ideas and Things. Thus, the support programs were effective for their intended groups and perhaps even bolstered or stabilized the STEM interest profiles of the participating students.

Among the overall large sample of 1170 students, the interest profiles of female and male students were different for those in the support program and IMP and former participants of the Hari program mean. The distance between the interest profiles of female students in the support programs (Digi, Hari, IMP) and female students in the control group was further that between male students in the support programs and male students in the control group. This finding is in line with the other findings. Therefore, there is a need for STEM support programs to attract a more diverse group of students (Neher-Asylbekov & Wagner, 2022; Wang & Degol, 2013). Means to achieve this could include using selection criteria such as grades in order to admit more female students into the Digi program. Furthermore, other selection criteria, such as suggestions from parents, teachers, school principals, and peers, which influence students when choosing a program, should be carefully considered when promoting support programs, as a significantly larger number of male students still choose STEM subjects in higher grades and at university (Bahr & Zinn, 2023; Happe et al., 2021; Lang et al., 2015; Schwarze, 2022). As these factors can influence the career decisions of students (Ertl et al., 2014; Rocker Yoel & Dori, 2022), it seems necessary to minimize (conscious or unconscious) gatekeeping processes in order to attract a more diverse group of students to participate in STEM support programs (Eccles et al., 1998; Nissen et al., 2009; Vieback et al., 2020). Furthermore, the literature states that the context in which STEM subjects are taught should be changed to include topics such as climate change, health, and the benefits of technology for everyday life in order to attract more female students (Bahr & Zinn, 2023; Elster, 2007; Pedersen et al., 2021; Otto et al., 2017).

Limitations

Limitations are evident in this study, primarily due to the lack of a longitudinal or pre-post analysis of the programs. Therefore, it is not possible to establish a causal attribution between STEM programs and actual changes in vocational interests. Moreover, it can be assumed that, in addition to the intervention of the support programs, other factors could moderate vocational interests (e.g., peer group, subject teaching, other social factors) (Gossen & Ivey, 2023). The response rate was below 30%, so a major limitation was possible bias in the responses (self-selection). Another limitation was the comparability of the programs and the subgroups. The size of the Digi group was much smaller than the rest and they came from only one school. Hence, effects on the school level might be reflected by this subgroup (and mistakenly be interpreted as effects of the program). On the one hand, the Digi group was a group of students with high grades in science selected from the total number of students in the 10th grade at one school. On the other hand, it was assumed that not all high-performing students in the 10th grade could be included in the Digi group, as the program limited the number of participants. This assumption was further supported by the teachers. Therefore, the CG represents an average group of students overall with the aforementioned limitation. Another limitation was that not all genders are represented in this study since the sample size of students categorizing themselves as “other” was too small. The comparison between the RIASEC profiles of STEM students and K–12 students in the support programs made it possible to determine whether the programs attract students with specific interest profiles. However, we cannot account for students dropping out of their initial course of study. Furthermore, the mean values for the interest profiles of subgroups were calculated, showing a tendency but not diversity. However, this was provided in the descriptive analysis for RQ1.

Conclusion

Despite the limitations, it can be noted that the results indicate that the support programs mainly reached students with a vocational interest in STEM professions. At the same time, STEM support programs and elective STEM subjects at school make it possible for students to choose fairly early on (IMP starts at grade 8, around age 14), therefore deepening their understanding of STEM topics and maybe even their interest in STEM. Thus, the authors consider that the results of this study are in line with the theory of interest and its “catch” and “hold” components according to Mitchell (1993) and Krapp (1992). Students in these programs received an incentive (a tablet) and chose an additional STEM subject (Digi) or their preferred profile subject (IMP), or delved deeper into a STEM topic of their choice (Hari). The promised STEM topics could be categorized as a “catch” for students who may have had an interest in STEM before. By participating in these programs and further exploring various STEM topics, the authors assume that students might have acquired an individual STEM interest (“hold”). One research avenue could be to verify that assumption with a longitudinal analysis of the development of students’ vocational interests.

After analyzing the results and splitting the groups by gender, one main finding of this study is that there was more variation in the mean values for the interest profiles between female students in the support programs and female students in the control group than between male students in the support programs and male students in the control group. Thus, one hypothesis could be that the support programs were more successful at fostering the vocational interests of female students than male students.

All three support programs ran over multiple years, and the majority students involved had a high STEM interests and vocational interests in R and I. These results, supported by the studies of Gossen and Ivey (2023), Guderian (2007), Neher-Asylbekov and Wagner (2022), Nickolaus et al. (2018), Sahin (2013), and Sahin et al. (2014), indicate the importance of offering support programs on a regular basis. In one way or another, all three support programs were integrated into the school or the curriculum. Since mentoring programs (Kuchynka et al., 2022; Nickolaus et al., 2018) and knowledge about careers and studies (Blotnicky et al., 2018) are important factors influencing career choices, and many of the Hari1 students tend toward Ideas and Things, it could be beneficial to include STEM universities in STEM support programs as a way for students to obtain more knowledge about the programs and courses they can choose and have university mentors supporting them.

Despite the limitations of this study, the comparison of STEM interests, vocational interests, and factors influencing the educational choices of students in different programs represents a valuable contribution to the field.

Future research could delve deeper into how different pedagogical approaches foster the STEM interests of students and what actions (workshops, mentoring, writing scientific papers, funding, etc.) are the most effective at fostering students’ STEM interests, and analyze gender-specific differences in those approaches and actions.