3. Min undervisningsfilosofi (Teaching Statement)¶
Om denne undervisningsfilosofien¶
Denne teksten er skrevet til å leses som et selvstendig dokument, uavhengig av det øvrige innholdet i denne pedagogiske mappen. Dermed er det nok noe overlapp mellom denne teksten og de øvrige kapitlene. I tillegg er teksten skrevet på engelsk siden jeg ønsker å dele den med internasjonale kolleger.
Introduction¶
My approach to developing my teaching is based on the observation that education is its own science, and a conviction that we should take the results of that science seriously when we develop courses. Just as prior research informs us when we develop our own research, so should educational research inform us when we develop our own teaching.
Educational research is a vast field and there is a lot to learn. But, one can not use all there is to learn, one has to limit oneself and focus on a few areas at the time. For me, these areas are i) Cognitive Load Theory, ii) Formative Assessment and Peer instruction, and iii) Constructive Alignment. In the following, I will introduce what I have learned from these areas and describe how I have implemented this in my teaching. I will focus on a new course that I have developed in the period since 2017, BIOS1100 – Introduction to computational models for Biosciences. When developing, and improving the course, I increasingly tried to incorporate what I have learned, and am learning, from educational science. This will be described in the following three sections.
Cognitive load theory and the need to manage cognitive load¶
Theoretical background¶
Knowing some of the research of how humans, and especially university students, learn can (and should) inform us on how to organise, plan and execute our teaching. The study of mental processes, including learning, is called Cognitive Psychology. Part of this field is concerned with cognitive load theory.
Learning theory starts from acknowledging that humans have two memory systems: working memory and long-term memory. Working memory is where new information is processed, and coupled to pre-existing information present in long-term memory. It is said that learning happens if this new information is transferred to long-term memory. While long-term memory can contain vast amounts of information, working memory is considered small, and has a capacity of about “seven plus or minus two” pieces of information [6]. Cognitive load theory is “a set of learning principles that deals with the optimal usage of the working memory” [7]. This theory, as defined in a recent review on the subject [4], “aims to explain how the information processing load induced by learning tasks can affect students’ ability to process new information and to construct knowledge in long-term memory”. The theory argues that the limited capacity of working memory severely restricts how much new information can be processed at any one time. When too much is asked from this working memory, there is a risk of overloading it, hampering learning. Overloading working memory inhibits the effective transfer of new knowledge to long-term memory, which is required for learning. It is argued that instructional methods need to take these limits into account.
What does this mean for BIOS1100?¶
Learning how to program is an important part of the teaching in BIOS1100. Learning to program is generally considered difficult [8] [9] [10]. Reducing cognitive load for students then becomes an important goal. I have always felt students can not learn programming from looking at a slide presentation of programming concepts, and then asking them to start programming themselves. I have experienced this approach myself at some time and it did not work for me, nor did it seem a useful approach. One reason for this is that this approach would result in a large cognitive load: students would be required to retrieve the factual knowledge presented during lecture and apply it to solve complex problems without any guidance on how to approach the problem. There is thus a need for alternative ways of teaching the fundamental building blocks of a programming language. One that is more based on Guided Instruction [11].
My main technique for reducing cognitive load when learning programming is called Participatory Live Coding.
Managing cognitive load in teaching programming: Participatory live coding
Participatory live coding is a guided instructional technique “in which a teacher or instructor writes and narrates code out loud as they teach and invites learners to join them by writing and executing the same code” [12]. The instructor reads what is being typed out loud, explaining the different elements and principles. Teacher and students all execute the commands or program, leading to an immediate evaluation of the results - hence the term ‘participatory’. Crucially, the session contains regular, often short, exercises, where students are asked to solve a small relevant problem on their own or in pairs or small groups.
I am convinced that participatory live coding can help reduce this cognitive load for students learning programming. This approach works better than lecturing about programming, or relying on students reading a textbook or compendium. What is taught is immediately applied and the execution of the program being written provides immediate feedback. This helps student couple programming code with its result. Students’ questions arising during the session can immediately be addressed. During this form of guided instruction, students are shown not only what to program and how each element works, but also how to program, i.e. how to go from a problem formulation to a working solution (i.e., the thinking process). It also slows the teacher down relative to using slides to show the concepts and code, giving students more time to actively engage with the material before moving on to the next concept. Interrupting the live coding with exercises enables immediate practice using the material.
The main drawback of using this technique is that it takes time to develop appropriate material and to teach it. Additionally, it does not scale too well as a single student with a problem that keeps them from following along and that takes some time to fix may hold up the entire group. Finally, students may think that they learn enough by simply following along. They should be made aware that to properly learn how to program they should practice, for example by doing exercises.
Participatory live coding in BIOS1100¶
The first edition of BIOS1100 (2017) relied on the students using the textbook for learning the programming concepts needed each week. Programming was not taught during lectures for reasons described above. Oppgaves were handed out during group work where students could practice applying programming to simple biological problems. Two-thirds into the semester it became clear that a significant group of students did not master the Python programming, and were in danger of failing the course. I then decided, rather than introducing more new Python material, instead to offer some extra teaching using Participatory Live Coding. I had learned this technique from being an instructor for Software Carpentry. Software Carpentry, now part of the global non-profit called The Carpentries, “teaches researchers the computing skills they need to get more done in less time and with less pain” and is mostly aimed at researchers at the PhD and postdoc level. Participatory live coding is the main method of teaching in the two-day workshops, and learning it is part of the training for becoming a certified instructor [13]. In 2017, in the last weeks of the course, I thus offered sessions re-teaching the Python material with Participatory Live Coding, with the effect that many students reported that they now finally understood it.
I had previously considered using Live Coding for BIOS1100, but felt it would not scale to such a large group. The experience in 2017, and the student’s feedback, convinced me that I had to find a way to adapt the technique of Participatory Live Coding to a course with 200 students, starting from the 2018 edition of BIOS1100. I decided that it should be taught in small groups, not with the entire group of students, and that I could not teach it all by myself. My solution was to develop a completely new set of teaching materials to teach Python programming in BIOS1100 using Participatory Live Coding, and train Teaching Assistants to be able to teach using this technique. As a certified Instructor Trainer for The Carpentries, I already had taught Participatory Live Coding to incoming instructors. I thus reused the material developed by The Carpentries [13] to train enough Teaching Assistants so that they could teach Python using PLC during the group sessions.
The results were that string from 2018, students had a much greater confidence in Python programming. To start learning programming using Live Coding helped them overcome the initial barrier (sometimes fear) of programming, and led to a feeling of mastery early on. Students really loved the Live Coding (‘samkoding’ in Norwegian), reporting they learned a lot from it. A drawback of this approach was that in 2018, much, sometimes all of the 4 hour group sessions were used doing Live Coding with the students. Too little time was left for students working on their own with exercises. A master student who studied the BIOS1100 students for his project that year concluded that students lacked good problem-solving skills because of this [14].
In other words, while in 2017, we did not help students enough with learning programming, in 2018 we helped them too much and did not challenge them enough to apply what they learned.
From 2019 onwards, the Participatory Live Coding was moved to voluntary sessions, two hours each week. Students new to Python were strongly encouraged to participate. In group sessions, organised activities led by Teaching Assistants were limited to the first two hours, which left two hours for students working on their own solving problems. During the first two hours, some Live Coding was done to further explain concepts, or for so-called worked examples. Worked examples “provide a full problem solution that learners must carefully study” [4], and are another technique for reducing cognitive load.
Participatory Live Coding was what made introducing Python programming to new Bioscience students possible. I believe this technique can be used in many settings where students who traditionally do not have to learn programming are introduced to it.
Formative Assessment¶
Theoretical background¶
It would be a mistake to assume that students have learned the thing you just presented to them. Formative assessment is concerned with informing both the teacher and the student about how much students understand about a topic, and discover any misunderstandings. This allows for addressing misunderstandings promptly, which helps learners to progress through the material. Formative assessment is not graded, although sometimes teachers make participation count towards being able to pass a course. Note that graded assignments are what is called part of a course’s summative assessment. Ideally, formative assessment can be done quickly and in an easy way for students and teachers to participate and see the results.
There are many forms for formative assessment, but a very common one is Multiple Choice Questions. A well-designed Multiple Choice Question poses a problem with one or more correct, and multiple incorrect answers, so-called distractors. Ideally, distractors should not be too obviously wrong, but rather be indicative of possible misconceptions.
Formative assessment combined with Peer Instruction
Originally created by Eric Mazur at Harvard [15], Peer Instruction is an evidence-based method where students are actively discussing the material amongst themselves based on prompts provided by the teacher. By asking students to reflect on course material together in their own words, a student that just understood the material is able to explain it in a way that matches better a student that almost understands the material, than the way a teacher would explain it. Thus, *Peers* are *Instructing* themselves.Often, Peer Instruction is combined with formative assessment through Multiple Choice Questions. There are different approaches on how to do this, but commonly, a Multiple Choice Question is posed and students consider the different answers for themselves. They then vote for the answer they think is the correct one. The tally of votes is shown, without revealing the correct answer. If there is a spread of answers among the correct one and one or more of the distractors, students are asked to discuss the question in pairs or small groups. They then vote again (individually, not as a group). Often, the results of the second voting show many more students converging on the correct answer. If needed or desired, the teacher can go over the different answers and explain how they are right or wrong.
Peer instruction has also been studied in the context of programming, with positive results on student learning [16].
What does this mean for BIOS1100?¶
Also in BIOS1100 there is a need to investigate student learning and check for misunderstandings. This can be partly achieved by Teaching Assistants helping students during group work, and by studying the obligatory assignments students hand in during the course. But, also the technique of Peer Instruction through Multiple Choice Questions is an ideal addition to this.
Teaching Assistants are a vital resource for student learning. In BIOS1100, they have the most direct contact with students (during the 4 hours group sessions and the non-compulsory live coding sessions). This means that they have a lot of experience that could inform me as a teacher about student progress and misunderstandings. The challenge is then how to ensure this information reaches me as a teacher, in other words, how to implement formative assessment through Teaching Assistants.
Practical implementation¶
I use the following Formative Assessment techniques in BIOS1100:
Multiple Choice Questions and peer Instruction
I have developed a set of around hundred Multiple Choice Questions for BIOS1100. Some of these are based on common misconceptions I found in the scientific literature. Often, when I observe students displaying a misconception, or are told about one by the Teaching Assistants, I use that as inspiration for writing new ones.
I have used Multiple Choice Questions, through the online Student Response System Mentimeter, during lectures and group work. I always combine this with Peer Instruction, using the approach described above. I regularly observe a mix between questions that are ‘too simple’ (a large majority of students get it right at the first try) and questions that reveal a misunderstanding in a significant proportion of students, that then gets largely resolved in the group discussion. Students and Teaching Assistants really like the “Menti’s”, as they are fondly called. It is an easy form of Active Learning that helps create a dynamic group session or lecture. A drawback is that executing Multiple Choice Questions takes time. I usually use no more than 4 questions concurrently, and those easily take up half an hour or more. Ideally, I would be able to always see the tally of votes for all questions to be able to filter out the ones that are too easy for next time, but is it challenging to collect that data from all Teaching Assistants.
It is fairly straightforward to adopt Multiple Choice Questions to an online teaching setting, provided the platform used allows students to work in small groups (for example, in so-called Breakout Rooms that tools like Zoom offer). In my experience, this works best if the students in these groups know each other from before, otherwise it is much more challenging to get them to discuss the question.
Obligatory assignments
In BIOS1100, students hand in 5 obligatory assignments spread evenly throughout the course. These are meant to ensure students work with the material throughout the course. The assignments are deliberately modeled after exam problems, so as to help students prepare for the exam. These “Oblig’s” are graded pass/fail by the Teaching Assistants, and students have to pass 80% of them to be able to take the exam. When students fail on their first attempt for an assignment, they get two additional chances.
In the first edition of BIOS1100, not only were there 11 assignments, I intended these to be only used for summative assessment. I had attended a presentation where it was argued that one should not try to combine formative with summative assessment as assignments can not effectively serve these two purposes. So I instructed the Teaching Assistants to grade them without leaving any comments. It quickly became clear that it made much more sense to give students feedback on their assignment, regardless of whether they had passed or not. For students, having the option to hand in some of the work they are doing, and receiving constructive comments on it, is very useful for their learning. Using the obligatory assignments for this is really a very good way for all students to receive such feedback, and one of the few ways to offer this to all students throughout the course. I thus concluded that I deliberately wanted to use the obligatory assignments for both formative and summative assessment. From then on, I asked Teaching Assistants to give feedback to all delivered assignments.
Ideally, I as a teacher would also study the deliveries to distill common misunderstandings, in other words, use them as a proper formative assessment tool for myself. Unfortunately, there has as of yet not been enough time during the course to do this.
Constructive Alignment¶
Theoretical background¶
Constructive Alignment is concerned with aligning the learning activities with the intended learning outcomes. Following Biggs [17], we start with determining:
The desired learning outcomes, these are the objectives
How to measure whether desired learning has been achieved, i.e., assessment
Appropriate teaching and learning activities that engage students in a way that leads to learning
Biggs calls this Constructive Alignment, which is an aligned system of instruction where “the objectives define what we should be teaching; how we should be teaching it; and how we could know how well students have learned it” [5].
What does this mean for BIOS1100?¶
Following Biggs’ approach, we should start by determining our objectives. The learning outcomes of BIOS1100 are described on the course homepage (English). But since these are deliberately kept short, there is a need to expand on them. The next step would be to determine the summative assessments, in other words, make exam questions and any obligatory assignments before the course starts. In practice, this is not often done. Finally, we need to design appropriate activities that help new students learn the mix of biology, mathematical modelling and programming that BIOS1100 aims to teach.
Practical implementation¶
Jupyter Notebooks and JupyterHub
To align how we teach programming In BIOS1100, we use Jupyter Notebooks for all course activities. Jupyter Notebooks are a form of ‘computational’ notebooks, combining text, media, programming code and the ability to execute that code and include the results of running it in the same notebook. Teaching materials, including exercises, are delivered as Jupyter Notebooks. even the textbook written for the course is made available as Jupyter Notebooks. During Live Coding sessions and group work, students as well as teachers and Teaching Assistants, do all their programming in them. Obligatory assignments are also handed out, and handed in and graded, as notebooks. When students, assistants and teacher all are seeing and commenting code in the same environment, discussing problems and helping each other, this reduces the extra cognitive load of switching programming environments. We use a cloud-based server provided by the university, called JupyterHub, to provide students with this programming environment. An additional benefit of this programming environment is that it saves the students from installing software on their own laptop: as long as they have a working internet connection they can log in (using university credentials) to the server and start working.
Exam
Although we used Jupyter Notebooks from the start of BIOS1100, in the first two years of the course, there was one component that did not use these notebooks. During the exam the students had to use a different environment which did not allow them to test and run their code and programs. The four hour exam was done in a UiO’s digital exam environment Inspera, which did not have a technical solution for running programming code. This restriction was initially a deliberate choice. In dialogues with my colleagues who teach beginner programming courses at the Department of Informatics, I became convinced that not being able to run code during the exam is of benefit for the students. Informatics students until recently even handed in their programs during the exam on paper. The idea from not being able to execute code during the exam is that it prevents students from getting stuck with a relatively minor error (in the syntax, for example), and having to spend a lot of time fixing it - which can be hard. I thus decided to follow the same strategy, the motivation for which I also explicitly explained to the students.
However, in dialogues with students it became increasingly clear to me that not being able to test and run programs they write during the exam caused a lot of stress for the students. They had not experienced this way of working during the entire semester. I also learned more and more about the benefits of Constructive Alignment, and concluded that the exam format led to a major mis-alignment in the course. I decided that the situation needed to change.
Thus, in the fall of 2019, with help from the university’s IT department and the faculty’s Inspera team, it became possible, for the first time for any digital exam at the university of Oslo, for students to submit exam assignments in the form of Jupyter Notebooks. This means that the students could now also run and test their code before handing it in. The result was a much increased Constructive Alignment between teaching and examination methods. Students were better able to show the level of comprehension of the programming part of the course.
In 2020, due to the Corona pandemic, the exam was changed to a 4-hour digital home exam, but otherwise organised as before. Incidentally, this aligned the exam situation even more than the 2019 exam, as it allowed students to use all available resources, as they are used to when working with (obligatory) assignments.
Conclusion¶
After several years of continued development, I am very satisfied with BIOS1100. I feel I have found the right format for this course, with lectures, Participatory Live Coding sessions and group sessions. The material written for the course, the set of problem exercises and Multiple Choice Questions are of sufficient quality and amount.
Along the way, educational theory has informed me for the many choices I as a teacher had to make. It has made me choose methods to lower cognitive load, for example by successfully scaling up Participatory Live Coding. It has shown me the importance of formative assessment, how there are different ways to be informed about student progress and how collecting this information can guide course adjustment immediately, or in between course editions. Finally, it has led to a much better alignment of the way students are exposed to programming, and work with it during the course, with the summative assessment, by making the exam situation as similar as possible to the rest of the activities in the course.
Students also report more satisfaction with the course now than in the beginning. In the student evaluation of 2019, for the first time some expressed achieving a feeling of mastery (‘mestringsfølelse’) in the open questions of the evaluation questionnaire.
I plan to continue to explore educational science to make adjustments in my teaching. I am convinced that using it as a basis for course design is the most fruitful way forward.