ChatGPT has a functionality where users can provide custom instructions that it will take into account in each conversation, which allows it to provide more tailored responses to their prompts. Here, you can fill in information about you: for example, what you do as for work/study, and what your hobbies and interests are (do be careful with which information you provide).  

For example, if you use ChatGPT to generate recipes but do not eat meat, an instruction like this could be useful:  

Additionally, you can provide information about how you want ChatGPT to respond. This could include output length and tone.

Canvas is a interactive text and code interface in ChatGPT. When asking  for a text or code output, ChatGPT has the ability to answer you in a sperate window (the canvas) within the chat environment. The canvas can be triggered in multiple ways, for example by uploading a text or code document or explicitely asking for the canvas format (see the below example)

After generating a text or code, you can alter the contents in three ways: By continuing the chat, you can give ChatGPT instructions on where and what to improve. Secondly, you can use the pencil button in the bottom right corner. Lastly, you can change the text or code yourself by typing in the canvas the same way you would do in a code or text editor.

To make sure your input isn’t used for training, go to your profile > Settings > Data controls, and turn “Improve the model for everyone” off. Note that this is a good-faith agreement. Using sensitive data in ChatGPT is still strongly discouraged, even with this setting disabled.

ChatGPT has the functionality to share links to chats, allowing others to read your conversations. Do note that this is not a permanent form of documentation: if the original chat is deleted, the link will not work anymore. To share links to chats, click the top right “Share chat” button, and create and copy your link.

Alternatively, you can use browser extensions to export your ChatGPT chat to a pdf-document, which provide a more permanent form of documentation (which is more in line with the WUR guidelines). This is a more secure form of archiving of your chats. One example is "Save ChatGPT to PDF", which easily converts your chats to a PDF document. (Click here for Chrome download, click here for Firefox, click here for Edge). To use the extension, download the extension, and make sure it is turned on. Then, open ChatGPT to export your chat to a PDF document. Note that side-views such as the Canvas feature may not be visible in this output.

Since Spring 2024 free users of ChatGPT have also received access to GPTs. These are customized chatbots that have been developed by paid users or associated businesses. Each GPT has its own customized set of instructions on how to behave towards its user and can contain a set of supplemental documents to expand its training data with highly specialized information. They can also be linked to external sources such as web interfaces, databases, or proprietary features through API connections, greatly expanding the functionality of the GPT. You can browse through GPTs through the ‘GPT Store’ and open a chat with any of these customized bots. A link to a GPT may also be shared with you, for example one designed by your teachers, for use during a course.

Data files, such as *.csv, *.pdf, or *.xlsx files, can also be submitted to ChatGPT. Using the Python programming language these data files can then be analysed by ChatGPT and the information used to answer your question.

The quality of the response does depend on the formulation of your prompt, as well as the clarity and formatting of the data file submitted. The Python script is used internally by ChatGPT to analyse these files. The added benefit of this method is that Python script written by ChatGPT can also be used within the ChatGPT environment and be tested for basic bugs or coding errors. Hence the quality of ChatGPT for coding tasks is also greater with the use of the Code Interpreter. Recommended is to copy the code used for either data analysis or coding tasks, to ensure a clear record exists of the code used by ChatGPT. This way you can always adjust the code if it needs to be adapted to a new situation or verify the functionality of the code before applying it more broadly.