Sunday, 22 December 2013

Computer Science: An Approach to Teaching and Learning

In 2012 we decided to completely review our Computer Science BSc. We are in the process of delivering the First Year of the degree at the moment. Our First Year aims are to:
  • Support each student to develop an appreciation of the key topics of the discipline through practical problem led holistic sessions that aim to reflect the way that CS occurs in the real world.
  • Integrate programming throughout the year and to use programming as a basis to engage with as many of the foundational aspects as possible.
  • Invert the locus of control for teaching and learning by allowing students to dictate the pace at which learning outcomes are demonstrated.
In order to support the First Year we made two decisions. The first was to use Racket as the programming technology. There were various reasons for this including the support Racket provides in terms of the Dr Racket tool, referential transparency, interaction through REPL, and the ability of Lisp-based languages to have a close association with foundational concepts such as sets, functions and data structures with the minimum of boiler-plate.

The second decision relates to course coherence and assessment. Under normal circumstances in the UK, a course is decomposed into a number of modules each of which  has its own syllabus and assessment processes. This imposes fire-walls around the different aspects of the overall year and makes it difficult to let the learning process naturally spread without over-assessment. For example, a First Year will normally contain modules for Architecture, Programming, Fundamentals etc. Topics such as graphs, events, or binary representation can occur as important features in any or all of these modules. If module apartheid is imposed then modules can tend to be overstuffed or there is a danger that topics might be missed. Furthermore, passing the year tends to require a minimum success in a range of assessments, usually 40% across the modules (typically with some minimum threshold in each). This raises the question: what does a student who has achieved 40% actually know? 40% of what

Our plan for the First Year is to introduce a sequence of examples, challenges, mini-projects, case-studies that provide an opportunity for multiple topics to be introduced and investigated and thereby  provide the opportunity for students to demonstrate the acquisition of knowledge and skills. Each of the key CS topics might occur multiple times throughout the year. A student is free (within reason) to choose how to do certain challenges and thereby when to demonstrate the learning.

Of course such a free-wheeling approach introduces a significant overhead regarding managing the learning portfolio of each student. This led to the second decision: the CS First Year is supported by a tool that manages Student Observable Behaviours (SOBs). The key topics are decomposed into a collection of SOBs that can be measured by a member of academic staff interacting with a student in a number of ways including labs, group sessions, observing presentations, or one-to-one tutorials. 

SOBs are tagged as: threshold, typical, or excellent. All essential foundational topics in CS are covered by threshold-level SOBs and to pass the year a student must have demonstrated all threshold SOBs. This ensures that the 40% problem described above does not arise: all student who pass the year have acquired a minimum knowledge of CS for all topics. The typical-level SOBs might introduce more in-depth or specialist knowledge and skills. Excellent-level SOBs provide an opportunity for students to demonstrate advanced standing through activities such as mini-projects.

Our SOB tool provides a platform for managing the student portfolios, but also has the benefit of providing a number of reports for both the students and the staff. Students can see their performance to date against what is expected (SOBs have demonstration time-windows that identify where the teaching team have planned opportunities for demonstration, although students are not tied to these) and against the rest of the cohort. Our experience to date is that this real-time reporting provides a beneficial incentive to maximise performance through competition within the cohort.

A screen-shot of the tool showing the current progress of students against threshold-level SOBs is shown below. The black line shows the expected progress. Actual progress is better than expected although most are around the 12-18 threshold SOB mark which is about right. The fact that there are quite a few ahead of the game is consistent with our experience that the approach has improved student engagement over previous versions of the CS First Year.

Here is a screen-shot of the typical and excellent level SOBs:

The year is organised as a sequence of 3 blocks each of which has a different academic team. Each team organises labs and workshops around the SOB framework and sets a single holistic block-challenge. The first challenge uses an Arduino connected to Racket to build a collection of traffic lights. The second challenge is to use data structures in Racket to build a simple dungeon game. The third challenge is to build a robot using Racket and a Raspberry Pi.

The block 2 challenge handout is shown below. It uses a series of simple games implemented in Racket: dungeon_rooms.rktmoving_about.rktmonsters_attack.rktcommand_language.rktitems.rkt.

Although we are only half-way through the first run of our new design, it seems to be working. Student engagement is very good with many students determined to 'get ahead of the game' by demonstrating SOBs as early as possible. Although many of the excellent SOBs are designed to be very challenging, as shown above some students are keen to try them. The use of Racket has gone well, although the small number of textbooks for Racket, and the dry nature of the on-line documentation, is a problem for introductory-level students who want to go further through self-study. However, in practice this appears to be a very minor issue. 

The SOB-tool is working better than we could have expected. It has been developed in-house but is not tied to CS or any course in particular. We would be happy to chat with anyone who might be interested in using the SOB-tool or finding out more details about how the CS First Year has been designed and delivered.