Joint Pilot for Smart Content and Courseware

Reflection

Course improvements

Timi Seppälä made improvements to the A+ platform, including upgrades to the way the platform handles course lifespans from drafts to running courses to partially active courses to archived courses, and the access privileges of various groups during those stages. He also made various minor improvements to the A+ user interface. Working primarily for CS-A1110 Programming 1 and the corresponding MOOC course, he made numerous course-specific improvements such as partially reimplementing an assignment that relies on an external maps API (necessary due to a change in licencing terms), overseeing the transition to using the Slack discussion forum and producing associated instructional materials, and implementing a new visual appearance for the course’s ebook.

CS-A1130 Applications of Computing (Tietotekniikka sovelluksissa) and CS-A1141 Data Structures and Algorithms Y (Tietorakenteet ja algoritmit Y) courses were developed further by Aliisa Pietilä, Juho Pekkinen, Riku Ala-Laurinaho, Miki Tolonen, and Ari Korhonen. CS-A1130 has gradually turned from pure manually graded exercises to allow more and more automatic assessment. This is an ongoing process that continues in the future. One of the challenges here is that the course utilizes many programming languages such Python, MATLAB, LabView, and Javascript. All of which require implementing and producing their own scalable automatic grading back ends. In addition, developing understandable automatic feedback is time consuming process that need to be improved in the forthcoming years.  CS-A1141, on the other hand, already had full pack of automatically assessed exercises. Thus, here the focus was on streamlining and eliminating overlapping work from the processes, improving the automatic feedback provided by the back ends, utilizing novel tools for developing new smart content, and improving usability.

In CS-A1140 Data Structures and Algorithms, Markus Arlander (under supervision of Tommi Junttila) developed and implemented new programming assignments (including the student code packages and the graders) for the course, e.g. a debugging assignment for the recap round that helps the students to refresh and improve their debugging skills (which will be essential in the following rounds when they most likely have to find out why their own algorithms do not function properly) and an assignment for the hashing round asking the students to implement a rolling hash function and the Rabin-Karp string search algorithm in a rather efficient way so that it is competitive against the string search algorithms in the Java and Scala libraries. Some of the developed assignments have been in “production use” in the Autumn 2017 course instance (to give an estimate of the student volume, 206 students have reached the grade 1 or higher level by Dec 5).

In course CS-A1150, Jonatan Lehtinen implemented (under supervision of Kerttu Pollari-Malmi) automatic assessment of relational algebra queries. Automatic assessment reduces the workload of the teacher, but also enhances learning by giving the student immediate feedback. When the system was introduced in a seminar in August, the teacher of the corresponding databases course in Tampere University of Technology was also interested in using the new system in his course. Jonatan Lehtinen also transferred the Python programming problems of course CS-A1111 from old Greengoblin system to A+ system and tested having the course text material and exercise problems interleaved on the same platform. However, the final deployment of the new system was decided to be postponed to Summer2018 course to make sure that everything works without problems in the course of 700 enrolled students.

Teemu Pudas has implemented declarative visualizion interface for the Artificial Intelligence (CS-E4800) and Declarative Programing (CS-E3220) courses. The idea is that the content of an image can be specified in a declarative way, i.e., the required graphical elements are given a logical specification which may involve complex conditions on which elements should appear and where they should be placed. Technically, such specifications yield a collection of graphical elements (such as text, lines, circles, rectangular areas etc) along with style parameters (such as color, line style, etc). Any collection of graphical images can be translated into some well-known graphical format (such as tikz under LaTeX or svg) and converted into a concrete graphical image. The current formats of interest are pdf and svg because the production of web-based materials is of main interest. For future extensions, it will be easy to extend the approach by new graphical elements and style parameters supported by the tikz format, for instance. The interface has been defined by the needs of existing use cases which go back to various kinds of images drawn for the purposes of the courses involved. In addition to the declarative visualization approach, Teemu Pudas has implemented mechanisms for the automated evaluation of exercises in the AI and DP courses.

The Artificial Intelligence course (CS-E4800) is the first course utilizing active elements that have been implemented by Roosa Piitulainen in the context of A+ system. The idea to embed external computations as part of web-based course material (like chapters in A+). Active elements enable the incorporation of various kinds of tools at students disposal. Each computation is based on a number of input fields and an output field. For the moment, the supported input fields are text fields and menu-based fields based on fixed inputs. Once all inputs have been defined by the student, the associated external computation is launched. Technically, this amounts to the evaluation of an A+ exercise, but the computation of the respective output (rather than points and feedback) is of interest. Currently, the forms of output supported by active elements are plaintext, html, and svg images. Input fields can be shared and output fields can be freely placed in the material (in contrast with traditional A+ exercises which conform to a more fixed format). To give a concrete use case, the AI course embeds a Sudoku solver in the material: given the hints of a Sudoku puzzle as input, the respective output fields presents the solution as an svg image. The image is updated whenever the input is changed.