**4. Discussion**

This research explored the suitability of the RP method as a tool for understanding the knowledge gained by participants in an intensive two-week summer school course and to support faculty in reflecting on the course design and teaching. To our knowledge, this study is the first to apply the RP method to teaching about food systems. The results of the RP analysis gave detailed insights into the knowledge categories for which significant changes were observed as a result of the course. At the beginning of the course, there were nine sub-categories present in the majority of the RPs drawn by participants (indicated by presence in >50% of RPs). These were sub-categories representing key elements of the food value chain (namely agriculture, transport and storage, processing and packaging, retail, consumption and resource inputs) as well as farmers, consumers and some system interactions. This highlighted that the knowledge most participants had when they came into the summer school course was based on a rather simple concept of what a food system is, with little reference made to the intended outcomes of the system, the diversity of actors involved, the boundary conditions it is embedded in or the forces that shape it. Based on the post-course RPs, there was a substantial increase

in knowledge in nine sub-category areas, namely the three key food system outcomes (food and nutrition security, social wellbeing, and environmental sustainability), the four food system boundaries that influence how food systems function (environmental, social, political and economic boundaries), food waste and losses, and farm workers as actors in the system. Thus, using the RP method, the faculty was able to easily identify if one of the key learning objectives of the course, i.e., to explain the elements of a food systems, the desired outcomes and the challenges, was achieved.

There were also a large number of sub-categories for which there were no substantial di fferences in knowledge before and after the course. This included agriculture, which was depicted in 98% of all pre-course RPs and added in the post-course RPs by the remaining 2% of participants, or consumption, present in 88% of pre-course and 96% of post-course RPs. Further, some sub-categories were not often present in either the pre- or post-course RPs. These included a large number of the possible actors that could have been included, the depictions of scales and tradeo ffs as a part of the system, as well as special topics like labels. It remains unclear why these latter sub-categories were not included. It may be that the participants gained new knowledge on these aspects, however considered it less relevant for a general, i.e., hypothetical food system, or that they simply found it too di fficult to draw. Alternatively, it could be that these participants did not learn or retain any new knowledge on these topics. As outlined by Schneider and Stern [17], there are many factors that need to optimally interact for learning to occur, the knowledge a person has can only be directly assessed by this person and is only made sense of when interpreted in the light of prior knowledge. This means that to understand in detail why certain sub-categories were omitted we would need more detailed data from each individual, which is not possible to gather for all participants due to limited resources (namely, time and faculty). In this sense, the RP method allowed us to collect high-level information from all participants that indicated the knowledge at the forefront of each participant's cognition at the end of the course.

From the perspective of the course faculty, a number of sub-categories were more important to be conveyed during the teaching program than others. For example, the understanding of food system outcomes and boundary conditions as well as their interactions, was a key part of the food systems concept explored in the courses. The analysis of the RPs indicated a substantial lack of knowledge in all these areas before the course which were closed with the course, clearly indicating that the course was successful in building knowledge in these areas. Another sub-category that was very under-represented in the pre-course RPs was the diverse range of actors within the food system. The post-course RPs saw a substantial increase in participants including farm workers as actors, in addition to the farmers who were often already depicted in the pre-course RPs. However, there was a large number of other actors and institutions in the system which were still not widely depicted in the post-course RPs, despite the role of these actors being explored in the program and integrated into interactive activities like stakeholder mapping. Although it is not our expectation that the final post-course RPs include an exhaustive mapping of all actors, their absence in the post-course RPs helps faculty reflect on why this may be the case, and consider how this could be better embedded in the program. Regardless, the addition of farm workers was a very positive step considering that the majority of participants came from a natural sciences background and had not been exposed to many social science topics.

Moreover, this change in the RPs clearly demonstrated a knowledge gain by the participants, which would have potentially gone unnoticed to the faculty without the employment of the RP method. Assessment methods in higher education are often 'feedout' (rather than 'feedback') in nature, in that the main objective is to grade people at the end of the course to act as a performance indicator for faculty and future employers or admission departments [18,19]. Grades from such assessment methods, typically in the form of exams or end of course assignments, often do not reflect the amount of learning occurring among the people during a course [18]. The RPs could be considered an alternative to this, a type of Classroom Assessment Technique (CAT), which is carried out for the purpose of learning and not for grading [20]. In this manner, the RP o ffers a tool to gain insight into what people are learning and to collect information in a manner that can be personalized ye<sup>t</sup> broadly interpreted. Importantly, it has the nature of a pretest/posttest, which helps establish knowledge gain in the context of prior knowledge [21].

Our experience with the RP has highlighted it is a method that is aligned with several of the ten cornerstone findings from cognitive research on learning outlined by Schneider and Stern [17]. Firstly, it allows participants to create the new knowledge structures themselves, as they are completely free to draw and amend their diagram how they choose. Secondly, it allows the participants to make sense of new information by linking it to prior knowledge. By amending their pre-course RPs, they are making sense of new information by interpreting it against their prior knowledge. The RP also helps participants to create relationships between the individual pieces of knowledge they acquire in the course, which is a requirement for successful learning [17]. Constructing the RPs requires the drawer to break down a complex system into its component parts, which can also help support the learning process. Thus, although we did not directly collect data from the perspective of the participants regarding if and how the RPs supported their reflection on their knowledge gained, considered from the cognitive perspective of learning, the method has the additional role of supporting their learning process through reflection. This is supported by a substantial amount of anecdotal evidence from participants, in the form of verbal feedback to faculty, that they found this tool a very useful method to gauge their own learning and to capture the new knowledge they gained.

The insights from the analysis of the RPs has also informed the course design in terms of how participant reflection should be integrated into the course. In the last couple of courses, we thus have increased the number of reflective sessions with the participants and begun to trial a number of di fferent formats to help them reflect on what they are learning at more regular intervals. We do this through a number of methods, for example daily journaling, daily plenary discussions on "What are we learning?", back reflection sessions, and reflective artwork or creativity, which allow participants to reflect more broadly on their overall experience. This includes not only the knowledge participants are gaining, but also the skills they are building, values and attitudes they are developing as well as capturing anything that is coming up for them emotionally and psychologically during this intensive learning experience. We now see an opportunity to better link these intermittent activities to the final RP activity, so that this acts as a recap and culmination of what is being learned, rather than as a stand alone activity. Reflecting on the RP as a form of CAT, we identified that it could also be valuable to do the updates to the RPs at regular intervals during the course. This would o ffer an additional reflection tool for the participants during the course, and would allow faculty to evaluate these assessments and adapt the program in real time based on any gaps that are identified. This could also result in participants having a "richer" RP at the end of the course that they could refer back to.

To date, the RPs have been collected by faculty at the end of each course to be analyzed. Participants had the opportunity to take a photo of their picture if they wished. In the future, it could be beneficial to email a copy of the picture to each participant at a certain timepoint after the end of the course. In this way, the picture can act as a reminder of what they learned during the course, but participants could also be encouraged to use it as a template to continually add to as they learn more about the system. In this sense, it would help to better communicate that there is no "end point" per se to a RP, and the learning process. The picture can continually be iterated and used as a tool to capture new knowledge beyond the end of the course. In this way, the RP supports the understanding of knowledge as an interconnected structure that is continually expanding [5].

Through a structured analysis of the RPs, we were able to better understand the key areas where participants gained knowledge, and this helped us as faculty to reflect on which parts of the course were delivering the intended learning, and which parts needed further development. Interestingly, the comprehensive analysis of the RPs across three cohorts reinforced the information that faculty had picked up from informally reviewing the RPs at the time they were drawn. This experience encourages us to continue using the RPs without a comprehensive coding and analysis as a way to gauge the knowledge of the participants as they come into the course, and to understand what knowledge was gained or where there were gaps. As a process, it was helpful for faculty to see, very broadly, what messages came across, where there was room for improvement and how that can be designed into the course and associated reflection sessions.

Interestingly, the analysis indicated that the summer school cohort, gender, or field of study had no significant impact on the number of additional sub-categories in the post-course compared to the pre-course pictures. This showed that despite di fferences in courses and participant backgrounds, using our course design principles [13], it was possible to create a learning environment and course program that benefited all participants in terms of increasing their knowledge of food systems.

The implementation and analysis of the RPs also allowed the identification of a number of limitations of this method as a tool to determine knowledge gained over the course. Firstly, by adding to the picture that was drawn at the beginning of the course, the participants were structurally constrained to the framing of the system as they originally understood it. Although they were able to completely change it, the time restraints and the point in time when it was undertaken (at the very end of the course) made it less likely that they did so since this activity coincided with the wrapping up of an intense two-week course and the process of closing a social dynamic that had been built. This means it was di fficult time to keep the attention of the participants and, as faculty, we often found ourselves constrained by time pressure to ge<sup>t</sup> through the full day of closing the course. This has meant there was never adequate time to discuss the changes to the pictures in grea<sup>t</sup> depth, which would provide additional data to support the interpretation of the pictures. From this insight we could identify a need in future courses to allocate and structure time at the end of the course to do a collective reflection on lessons learned with the group, based on their individual pictures. Further, we see the RPs o ffering a good opportunity to discuss more openly with participants around questions of worldview, how we know what we know, and how any situation is perceived di fferently by di fferent actors. Time for participants to sit together and compare their pictures and discuss these questions could potentially be valuable at the end of the course. In order to reintroduce the original conceptualization of the RPs as a participatory process, we could also have participants collectively draw a new RP at the end of the course, using it as an opportunity to explore knowledge, understanding, perspectives, values and blind spots, as well as to reflect on how they know what they have learned by attending the course. This more nuanced discussion may also give faculty additional information which can help in the interpretation of the RPs drawn by participants.

As has been extensively written about by Bell, Berg and Morse [6], the interpretation and sense making of RPs by a third party is inherently challenging. We took an approach of coding categories based on one of the key learning objectives of our summer school, thus identifying changes against a predetermined set of themes. This helped to obtain a high-level overview. However, this might mean that we lost the nuanced information in each individual picture. It could be interesting in the future to apply a form of content analysis called 'Educative Interpretation' proposed by Bell, Berg and Morse [6], where emergen<sup>t</sup> themes are identified from the diagrams themselves. Either way, there is substantial risk of misinterpretation, subjectivity or bias on the part of the coder, even when grea<sup>t</sup> care is taken to avoid these things. Essentially, we are still interpreting the creative work of others which is always challenging without extensive dialogue with the creator of the work.

Generally speaking, we have found the RP method to be an e ffective and simple way to assess the knowledge gained by participants as a result of the educational experience they have during our two-week course. Although it has a number of limitations, it has allowed us to obtain a high-level overview of what knowledge participants have when they come into the course, what knowledge they have when they leave, and to identify any gaps where we may need to adapt the content and methods in the course. Additionally, this method is valuable as a reflective tool for participants to consider and structure what they are learning from a content perspective. This needs to be supplemented with additional reflective methods to allow the participants to consider learnings beyond only knowledge to include skills, attitudes, values and their own role in creating change. It would be valuable for future research to further explore the e fficacy of this method as a tool for reflection for participants' learning in and about other complex systems. In our experience, the RP method is a simple tool that works

very well to support both reflective teaching and reflective learning in food systems education and we recommend its use in the sustainability education space.

**Author Contributions:** This study was conceptualized by N.B. andM.G.; M.G. was responsible for the methodology and the implementation of the method during the summer schools; M.G. and A.K.G. carried out the data analysis; M.G. and A.K.G. wrote the first draft; M.G., A.K.G. and N.B. finalized the paper; M.G. and A.K.G. were responsible for project administration; N.B. and M.G. acquired the funding.

**Funding:** This research was funded by the Swiss National Science Foundation through the project Delivering Food Security on Limited Land within the NRP 68 and FACCE JPI (grant number 40FA40\_160406).

**Acknowledgments:** We thank Aimee Shreck for her contribution to and discussion on the original design of the study. Special thanks go to all summer school participants for their grea<sup>t</sup> motivation and engagement.

**Conflicts of Interest:** The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.
