Overview of all projects

1. Lifestream for Learning (go to the project ...) Salim Doost and Yanchuan Li
2. Evaluation of peer-authored content (1-2 students) (go to the project ...)
3. Evaluate possible feedback given to students as they collaborate / argue with a visual argumentation system (interdisciplinary) (go to the project ...)
4. Improve an algorithm that identifies and evaluates clusters of student contributions within a collaborative learning tool (go to the project ...)
5. Create machine-learned classifiers for a collaborative learning tool, using an off-the-shelf data mining tool (go to the project ...)
6. Woz Experiments with erroneous examples (interdisciplinary) (go to the project ...)
7. Embedded ActiveMath Dictionary on the iPhone: (go to the project ...)
8. Smart Paste for ActiveMath Authoring (go to the project ...) Thorsten Hey and Yecheng Gu
9. ActiveMath Skinned for the iPhone (go to the project ...)
10. Online Platform Mashup-integration in ActiveMath (go to the project ...) Iulian Ursu
11. Create an ActiveMath book on content authoring (go to the project ...) Sabine Hunsicker, Joo-Eun Feit and Eva May
12. Video annotation of this lecture (go to the project ...) Ilona Komor and Aleksandra Pochron
13. E-Portfolios and CompetenceMatching (go to the project ...)
14. Medical Learning Objects (go to the project)Mohamed Tahoun
15. Medical Imaging in ActiveMath (go to the project) Ghazanfar Ali
16. Visualization of Wiki-Discussion (go to the project) Dominique Taffin

Title: Lifestream for Learning
Supervisor: Martin Homik
Assignees: Salim Doost and Yanchuan Li
Description: These days social platforms become incredibly popular. Not only, that they connect you with your friends, they also offer you loads of information about your friends' interests. Users love to share their life and they also love to follow their fellow students' activities. Facebook is the most popular international social platform in this area.
All applications in Facebook are connected with a base system which offers functionalities to issue user actions to lifestreams and hence to notify connected friends about these actions. This feature potentially offers a high benefit. Hence, let's use a Lifestream for Learning in ActiveMath.
The student group has to agree on a set of actions that should go into a life stream, for instance:

• Anton created a new discovery book on 'the average slope'.
• Anton successfully completed his book on 'the average slope'.
• Eva's competencies increased!
• Bert successfully solved a difficult exercise on 'binomials'. Want to try it, too? Go here.
• Anton added a public note titled 'This definition of the average slope is wrong'. Do you want to discuss it?
• Teacher X has uploaded a new book 'book title'

A further possibility are auto-generated news, like "10 students tried this exercise on 'slopes' but none succeeded yet. Do you want to give it a shot?"

To create a life stream, the following components are necessary:

• create a life stream newsfeed (Atom/RSS2.0) from the history/database and store it in a file. Use tools such as Rome, Propono, Abdera. Add proxy caches, server-side caches, and compression to reduce computer, network, and bandwidth usage.
• read in life stream newsfeed from file and display in a DIV on the main entry page.
• provide a RSS button to sign up for a newsfeed

Possible variations of the topic and follow-up activities are (to be discussed with the project supervisor):

• Devising a flexible architecture which allows to define/seperate life stream newsfeeds relative to user and to aggregate single newsfeeds to groupfeeds where groups can by defined by class, discipline, theory, competency, etc. Means/Tools for aggregation such as "Planet Tool" will be provided.
• Write a Facebook application that pulls the feed and displays the actions appropriately

Title: Evaluation of peer-authored content
Supervisor: PD.Dr. Erica Melis
Assignees: - (still available)
Description: We want to investigate the quality of peer-authored content and ways for its improvement. The project will produce/use a tool for socially rating the content for ActiveMath which is authored by the fellow students during the course. This includes the following tasks (for 1-2 students):

• make the peer authored content of the course available early enough, so that it can be rated as an activity of your fellow students during the course
• design the criteria to be rated
• set up tool that allows the other students to rate according to the criteria
• stimulate your fellow students (and potentially others) to do the rating
• compute the rating distribution and try to explain the results
• Write a report summarizing the results, in particular making suggestions on how quality and criteria could be improved.

Title: Evaluate possible feedback given to students as they collaborate / argue with a visual argumentation system
Supervisor: Dr. Bruce McLaren
Assignees: - (still available)
Description: One of the most important decisions to make in developing a tutoring system is how and when feedback should be given to students. A "wizard-of-oz" experiment is one way to test feedback, without actually implementing a real computer system to produce the feedback. In such an experiment, a human is given criteria for responding to students, watches as they use the system, and then responds in real time to actual student actions, emulating what a program might do. In this project the project team will evaluate potential tutorial feedback given to students as they argue with one another using a visual argumentation system (ARGUNAUT). The ARGUNAUT system includes a number of machine-learned classifiers that automatically evaluate individual and group contributions made by students and these classifiers will be used as the basis for wizard-of-oz feedback decisions. In this project, the team will:

• Learn how the ARGUNAUT system operates, i.e., experiment with the software and read relevant papers
• Review different tutorial feedback approaches, in particular, focusing on how and when feedback should be provided to students
• Devise a decision tree representation of how the ARGUNAUT machine-learned classifiers could lead to tutorial feedback given to students
• Design a simple wizard-of-oz experiment in which the decision tree will be used by a human "wizard" in providing feedback to students
• Conduct a small version of the wizard-of-oz experiment (with at least two groups of collaborating students), focusing on and evaluating how well the decision tree supports the wizard in providing feedback
• Write a report summarizing the results of the experiment, in particular making suggestions on how the decision tree could be used to, ultimately, support automated feedback

Title: Improve an algorithm that identifies and evaluates clusters of student contributions within a collaborative learning tool
Supervisor: Dr. Bruce McLaren

Assignees: - (still available)
Description: The ARGUNAUT system is designed to support a teacher as he or she monitors groups of students as they discuss and argue a contentious topic. The students argue using an online tool within ARGUNAUT called Digalo, in which contributions made by individual students on their own computers are brought together in a shared workspace and are visible to all students, similar to a chat, except in graphical form. The ARGUNAUT system provides a variety of feedback mechanisms to help the teacher monitor these discussions, such as whether students stay on topic and whether they answer one another's arguments and claims. One of the most important tools in the ARGUNAUT arsenal is Detection Of Clusters by Example (or DOCE), an algorithm designed to find clusters of important contributions, such as "chain of argumentation." DOCE is very useful, but it has at least one critical design flaw: It always returns a pre-set number of clusters, even if some of those clusters are not particularly similar to the given example. What it ideally should do is return only the clusters that are really similar to the given example, regardless of whether that is less or more than the pre-set number and according to a threshold of similarity. However, in its current design, it is not possible to objectively define such a threshold in DOCE and thus some code extensions are necessary. Therefore, in this project, the group of students will:

• Review and evaluate the DOCE algorithm, reading the masters thesis and papers on which it is based and reviewing the code
• Design and develop an extension to the algorithm that will allow it to vary the number of similar clusters it finds, according to whether clusters meet some basic similarity criteria
• Design and run experiments to test whether the implemented extension really improves the algorithm, with respect to returning only relevant and similar clusters
• Write a report summarizing the changes made to DOCE and the results of the experiment

Title: Create machine-learned classifiers for a collaborative learning tool, using an off-the-shelf data mining tool
Supervisor: Dr. Bruce McLaren
Assignees: - (still available)
Description: The ARGUNAUT system is designed to support a teacher as he or she monitors groups of students as they discuss and argue a contentious topic. The students argue using an online tool within ARGUNAUT called Digalo, in which contributions made by individual students on their own computers are brought together in a shared workspace and are visible to all students, similar to a chat, except in graphical form. The ARGUNAUT system provides a variety of feedback mechanisms to help the teacher monitor these discussions, such as whether students stay on topic and whether they answer one another's arguments and claims. Some of the feedback mechanisms are created using machine learning, in which the content of past discussions is used by a machine learning tool to learn new behavior classifiers. While a number of these classifiers are already available in ARGUNAUT, we believe more could be created, ones that may be even more effective in helping teachers as they monitor discussions. Thus, the primary work of this project is:

• Review the existing machine-learned classifiers in ARGUNAUT to understand their purpose
• Learn how to use RapidMiner, a freely available and easy-to-use machine learning tool to support the development of classifiers
• Devise at least three new categories of classifiers, ones that can clearly be relevant to teachers as they monitor discussions
• Annotate pre-existing map data with the new categories, having at least two members of the team do the annotations
• Calculate inter-rater reliability between the two annotaters
• Run machine-learning experiments to create classifiers for the new categories, attempting to get classifiers that perform at a Kappa level of 0.65 or above
• Write a report describing your new categories, annotation process and inter-rater reliability and experimental results

Title: Woz Experiments with erroneous examples (interdisciplinary)
Supervisor: PD Dr. Erica Melis
Assignees: - (still available)
Description: In traditional schooling, mathematics errors are punished and avoided although learning from errors could be beneficial. We want to test whether, when and how such benefits of learning from errors others have made can be achieved through presenting erroneous examples for which the student has to detect the error and then correct it/solve problem correctly.

For this purpose, a wizard-of-oz (Woz) experiment will be conducted. That is, instead of actually implementing the technology that can give feedback to the learner or adapt to her needs, a (hidden) human tutor provides the feedback, selects the exercises, examples or erroneous examples, etc for the learner. This human has to follow certain rules that have been preset for the experiment in order to test which conditions work well but has some freedom of actions.

As the actual learning environment you can use ActiveMath or your own input-help system.

The work for this project includes:

1. design worked examples, exercises and erroneous examples for specific competencies and misconceptions
2. design / adopt small questionnaire for motivation, interest, learning goal
3. implementation of Woz environment
4. invent at least 2 hypotheses to be tested in Woz (help of supervisor)
5. run Woz experiments including pre- and post-tests possibly questionnaire (with at least 10 students), document all learner and tutor actions
6. report

Title: Embedded Dictionary on the iPhone
Supervisor: Martin Homik
Assignees: - (still available)
Description: This project is an alternative project. You don't need an iPhone, but it would be great, if you have one. You definetely need an Apple Computer with iPhone SDK2 installed. If you have both, you are the perfect candidate.

ActiveMath has a dictionary component for searching math items. In this project, you should implement a native iPhone app which serves as a portable math dictionary. The library should be extracted from ActiveMath and synched with the iPhone. Identify key design questions such as "What kind of information do you need from the knowledge base, what information is not necessary. Do you need links to exercises? Keep your iPhone KB tight/slim.