refer to AI_Integration_ReadMe

Author: koelibri

README.md

@@ -17,6 +17,7 @@ The programming courses offered on openHPI include practical programming exercis
 2. The platform allows executing the code on the server and streams the output back to the learner’s web browser. Together with the customized view, learners are supported to get up with the programming in no time instead of taking care of the development environment. Learners do not need to download and install any compiler or runtime. Thus, MOOC instructors minimize technical help requests about correct machine setup and can focus forum posts on the content provided in the course.
 3. CodeOcean includes unit tests to provide feedback for learners and score their code. A unit test is defined as a program that either runs the learner’s code in a pre-defined way and compares the provided result with an expectation or the unit test parses the student’s source code and matches it against an exercise-defined string. While the code of the unit tests are hidden, learners can run the unit tests at any time and get instant feedback whether they passed or failed. If the unit tests fail the result is shown together with an error message defined by the MOOC instructors. On the one hand, this feedback helps people to help themselves and provides learners with a hint of their mistake. On the other hand, the automated scoring using unit tests is required to indicate progress for the learners. In the context of a MOOC with thousands of active learners, a manual review by the instructors is not feasible and peer-review of source code has not been implemented in CodeOcean so far.
 4. In CodeOcean, learners can ask questions about their program directly within the platform and in context of their current program. Usually, MOOC platforms provide a forum to discuss questions. While this concept also works great for source code in general outside of a MOOC (cf. [StackOverflow](https://stackoverflow.com)), it is an additional barrier for novices to summarize their problem externally. To understand the problem, contextual information is generally of help for others to provide the current solution. When using a dedicated forum, learners are required to provide as much information as necessary to reproduce the issue which beginners might find difficult to identify. As a result, they might copy too few or too much information. In addition, early iterations of the Java courses showed that learners did not format their source code appropriate in forum posts (but as plain text), making it difficult to read. With _Request for Comments_, CodeOcean provides a built-in feature to ask a question in the context of an exercise, thus lowering the barriers to get help. CodeOcean presents the learner’s source code and error message together with the question to fellow students and allows them to add a comment specifically to one line of code. Hence, the previously described issue is solved with a dedicated forum.
+5. To dynamically improve the static feedback provided by the unit tests and to encourage peer discussion of RFCs, CodeOceans provides a [ChatGPT integration](docs/AI_INTEGRATION_README.md). 
 
 CodeOcean is mainly used in the context of MOOCs (such as those offered on openHPI) and has been used by more than 120,000 users as of August 2024. CodeOcean is a stand-alone tool implementing the [Learning Tools Interoperability (LTI)](https://www.imsglobal.org/activity/learning-tools-interoperability) standard to be used in various learning scenarios. By offering an LTI interface, it is accessible from MOOC providers as well as other Learning Management Systems (LMS), such as Canvas, Moodle, or dBildungscloud. CodeOcean itself cannot be used directly by learners or other users than the MOOCs instructors or administrators.