Table 2 provides an overview of the ways to argue for RBL depending on the epistemic value. The table shows first the specific type of knowledge that is valued and aimed at with which epistemic value. Then possible arguments in support of RBL are listed. Two arguments appear for each epistemic value, first an argument with a strong claim. However, as the strong claim can easily be doubted, a second argument with a weaker claim, often based on indirect impact of RBL, follows. In case of the scientific claims, the strong one would be (first argument): RBL enables a deeper understanding of research and thus also deeper knowledge (Huber, 2009). The weaker claim (second argument) would be: University is defined by the strong nexus of research and teaching (von Humboldt, 2010), as realized by RBL. For both arguments a university can exert epistemic authority or is the epistemic authority (as an institution). The kind of valued knowledge is scientifically verified knowledge, whether it is evidence-based or supported by mathematical proof.

The second epistemic value in the Cascade Model is professional, referring to highly qualified work in a risk-propense field (Mieg and Evetts, 2018). The strong professional claim would be: RBL helps develop professional skills and attitudes, as in Schön’s model of the “reflective practitioner” (Schön, 1983, 1987). The weaker claim would be: Today’s professional practice is inconceivable without the use of research results (be it medical research or market research); you can only understand research if you have done research yourself, e.g., through RBL (Mieg, 2020). The type of knowledge professionally valued is practical, generally regulated by standards controlled by a professional association outside of the university (Mieg and Evetts, 2018). Within university, the representation of a profession is a discipline. The distinction between professions and academic disciplines has historic roots. Three of the classical four founding “faculties” of the medieval European university are the classic professions of medicine, law, and theology (Shank, 2013). With the technical universities of the 19th century, the professions of engineers and architects were added. The fourth original faculty was philosophy (which had its origins in the faculty of arts). This, in principle, was the only faculty dedicated to knowledge itself, and it is from the faculty of philosophy that the entire spectrum of today’s - not purely professional - disciplines, from mathematics to ethnology, has emerged over time.

The third epistemic value is skilled, referring to knowledge incorporated in skills and competency. This view is more or less directly linked to the idea of education as an investment in human capital (Becker, 1964) – an investment that can be very specific, as in the case of experts (Mieg, 2001), or more general, the case of transferable skills and key competencies (European Commission, 2016) such as digital literacy (Mieg et al., 2024) or critical online reasoning skills (Molerov et al., 2020). At the political level, in the European Union, this value has been anchored in concepts such as lifelong learning, aiming “to enable each of us, without exception, to develop all our talents to the full and to realize our creative potential, including responsibility for our own lives and achievement of our personal aims” (Delors et al., 1996, p. 17). As Table 2 shows, a RBL argument with a strong claim would be: Competency in research (which can be acquired through RBL) is one of the key competencies in our modern and complex world (González-Pérez and Ramírez-Montoya, 2022). A weaker claim would be: RBL enhances knowledge and skills (Kleinschmit et al., 2023). For the universities, arguments referring to individual skills carry the risk that other educational institutions or charismatic educators such as Maria Montessori can be seen as equally good or more competent in educating relevant skills, e.g., in terms of social empathy or leadership capability.

The fourth epistemic value is social, that is the principled epistemic primacy of the social. In other words, knowledge is relevant as far as it serves society. The orientation of knowledge and its education to a specific set of social values can be called ideological (i.e., related to politically relevant ideas and corresponding beliefs and sets of values). Historically, there have been several ideologies of education that subsume university education, e.g., Marxist, nationalist, or religious. A RBL argument with a strong claim would be: RBL can have social benefits, e.g., as community-based or application-oriented research (Pawson et al., 2022). A common argument for RBL and undergraduate research with a weaker claim goes via critical thinking: Critical thinking - as acquired by RBL - makes us better citizens through critical thinking (Dewey, 1916; Glaser, 1985). Actually, there is broad evidence that we can educate critical thinking at universities (Abrami et al., 2015). However, there seems to be little scientific evidence that critical thinking really makes us better citizens. For the universities, any society-oriented argumentation carries the risk of being confronted with epistemic authorities that subscribe to specific ideologies, religious or political. Currently, there is a “post-truth discourse” (Lewandowsky et al., 2017), bringing forward counterevidence against research on climate change or vaccination. Lewandowsky et al. (2017) claim that this is not misinformation, but “an alternative epistemology that does not conform to conventional standards of evidence reporting” (p. 356). Thus, the knowledge that is valued from a social perspective has to be compatible with a particular social ideology that may appreciate scientific standards of evidence-finding or not.

The Cascade Model for justifying RBL describes belief systems. Academic controllability means that the university has epistemic authority over the beliefs involved. In particular, academic controllability refers to the power that the university has to implement the support of RBL. Academic controllability is highest when the university can set the teaching standards itself. Controllability decreases when the university is dependent on the specific cooperation of the disciplines. In Europe, strong professions such as medicine or law have traditionally had a strong influence on education through their disciplines or faculties at universities. However, arguments that refer to the university and to professions as epistemic authorities both belong, so to speak, to the realm of the university and can therefore be controlled by the university to a certain extent. In areas beyond academic and professional education, the university loses to the extent that its epistemic authority declines. Our assumption is that the university is more academically controllable over the claim to skilled knowledge than it is over social claims. The training of skills can be anchored in curricula. However, the university has limited control over social and democratic claims. Anyone claiming freedom of speech can completely challenge academic control. According to the Cascade Model, free speech is clearly a “social” epistemic value.

The Cascade Model for justifying RBL imples a simple cascade approach for the argumentation of the question “Why integrate RBL in your teaching?” Start with scientific research as the core reference and the epistemic authority of the university. If this is not convincing, use the professional claim, then the claim of skilled knowledge, and then the social claim. This sequence goes hand in hand with a decreasing academic controllability by the university. We speak of a “cascade” model because the original argument of the link between research and teaching, which can be used to argue in favor of RBL, is broadened and at the same time diluted by a reference to additional values such as skilled or social.

To link the Cascade Model to the broader context of research on epistemic cognition, Table 2 lists possible epistemic interventions. As mentioned above, we found that epistemic cognition in general and epistemological beliefs in particular play a role in RBL (Berndt et al., 2021). Epistemological beliefs, commonly defined as “beliefs about the nature of knowledge and knowing” (Conley et al., 2004, p. 187), influence students’ learning processes (e.g., Bromme et al., 2008), in particular they affect critical thinking (Muis et al., 2021). A key finding is the transition from “absolutism” (science as facts) to “multiplism” (different legitimate views of knowledge) and to “evaluativism” (focus on the process and evaluation of knowledge) that guides the understanding of research by students (Klopp et al., 2023; Kuhn and Weinstock, 2002). There is only scarce research on epistemic cognition or epistemological beliefs of university teachers (e.g., Mataka et al., 2019). The epistemic interventions listed in Table 2 refer to the Apt-AIR framework by Chinn et al. (2020). These interventions are designed to help students perform fully apt “in a world with deep epistemic disagreements” (p. 171), e.g., in “post-truth” discussions about climate change or vaccination. We relate these interventions to discussions with university teachers about why they should support RBL (labeled as by Chinn et al., 2020):

This variable is nominal and has four values: scientific, professional, skilled, and social. The scoring was based on the answers to the interview question: “Do you see any added value in research-based learning, and if so, what is it?” (see Appendix A).

We assumed that some respondents would make a significant reference to an epistemic value in order to be able to answer the question about added value, e.g., “from the doctor’s point of view” (epistemic value: professional) or “if we want to teach research methods at the university” (epistemic value: scientific) etc. Examples from the interviews:

This dichotomous variable has the two values: yes (RBL is implemented) and no (RBL is not implemented). The scoring was based on the answers to the interview question: “Do you implement research-based learning in your courses? In particular, do students conduct their own research as part of your courses?”

Psychology & Education: When respondents made it clear that they allow students to conduct their own research activities in the regular course, RBL had the value “yes.” Otherwise, “no.”

Medicine: The reformed medical curriculum at the LMU includes a 12-week compulsory scientific project that each student must pursue individually (similar to the research thesis in Bachelor’s programs such as psychology or biology). In addition, the medical curriculum at LMU Munich and throughout Germany follows a national regulation with little room for RBL activities in the required courses. However, some medical faculty reported efforts to support students’ research activities in other ways, such as through problem-based learning or offering voluntary research internships. If respondents implemented more than one of the four RBL substitutes in their courses (i.e., problem-based learning; support for doctoral studies; evidence-based learning; “feedback through evaluation” to support active learning), RBL had a value of “yes.” Otherwise, “no.”

We can also calculate interrater reliability across all codes. To do this, we use the implicit ranking of epistemic values in the Cascade model and set: scientific = 4, professional = 3, skilled = 2, social = 1. The values reflect distances between the codes: “scientific” is therefore further away from “skilled” than “professional.” Differences in coding are more substantial depending on the distances. If there was uncertainty within a coding team and two codes were proposed, the mean value is taken for their codes. The interrater reliability of two teams, measured with Kendall’s tau, is 0.645 (p < 0.01). This can be considered a very good value. Even if there was not always complete agreement, the codings are close together.

Table 3 shows the distribution of matched codings according to epistemic value. The low number of references to “professional” (only 4 out of 26) is somewhat surprising, considering how important the reference to the profession seems to be, at least for many students. The result becomes somewhat more relative when we look at the categorizations by the two teams before integration (Table 3, last row). Then “professional” has the second most mentions (25), only “scientific” has more (35). In our interview context, however, “professional” does not seem to provide an easily assignable epistemic value.

In order to perform the calculation, we must also include the values for justifications with ambiguous coding (no matches). To do this, we create a new variable with a simple ranking of the epistemic values as “up” and “down.” If the categorizations from both teams are “professional” or “scientific,” then this should mean “up” (the new variable is assigned the value 3). If both categorizations are “skilled” or “social,” then “down” (the new variable is assigned the value 1). If the categorizations of the two teams would intersect, i.e., sometimes “up” and sometimes “down,” then they fall into a category called “middle” (the new variable is assigned the value 2). This proxy variable turns the four categories of the Cascade Model into three (up, middle, down), which run in the same direction as the ranking in the Cascade Model. The proxy variable correlates significantly with implementing RBL (Spearman’s rho = 0.286, p < 0.05). In other words, the “higher” the epistemic value, the more likely RBL seems to be implemented. The effect is there, albeit very weak.

If we consider all codings (mean values of all categorizations for a case)¹, we can expand the group of cases with “skilled” justification from 6 to 8. RBL is not supported in a single case. This is statistically significant (8 “skilled” cases vs. all other 32 cases, Fisher’s exact test, p < 0.05). This finding contradicts the common practice of trying to promote RBL by referring to the acquisition of necessary skills (with “skilled” as epistemic value), including our own rationale for the RBL research network (Figure 1).

We should also note that there were differences between disciplines, and that the RBL measure was already a compromise. In the narrow sense - truly independent student research - RBL was only supported in Psychology & Education. In Medicine we find approximate forms of RBL, e.g., problem-based learning. Interestingly, only one interviewee from medicine linked RBL to the aforementioned 12-week mandatory scientific project at LMU Munich Medicine.