Open Coding resulted in 117 different codes that differed quite a bit depending on the lecture, since different topics were discussed in different lectures. We did not want the codes to be too broad, since we could always categorize them together later on.

• The most frequent unique codes (codes that appeared more than 50 times) were:
  • students simply giving the instructor examples after the instructor asked the students to give some (119). This suggests that an instructor that tries to involve the students has a greater chance for the chat to be more active.
  • jokes by students (56). These jokes were mainly used to create a more friendly atmosphere and this results in the lectures feeling less stiff. In some cases this also caused some students to talk more in the chat.
  • emojis and emoticons (54). These were used to express different emotions, which is otherwise hard to achieve through text without writing a bigger message. This is essentially a short way to show how you feel about something.

Axial Coding resulted in 19 different codes. Some of these codes were similar, mainly because they differed by being a positive or a negative categorization (such as criticisms towards lecture or instructor compared to compliments towards lecture or instructor). This was important to examine possible differences and to see if the messages were leaning into a certain direction. We also wanted to see if there were categories that would've formed around a negative or positive code.

• Similar to the Open Coding, the two most frequent categories of Axial Coding were students answering their instructor (156, including students giving an example to the instructor or giving an affirmative or negative answer, often just with a yes or no in the chat) and small talk (154, including jokes by students, parting and greeting words or thanking other people).

Selective Coding resulted in 7 different codes. The by far biggest group here is for discussions that are about the lecture and between students or instructors with the goal of providing information with 318 messages that could be applied to this group. This is followed by discussions that are off-topic or not about the lecture and providing information for it with 235 messages. That last of the major groups is interacting with features that are suited or unique to a live chat with 114 messages.

Note: The category „Evaluating the quality of the lecture or instructor“ can be applied to 30 and not 31 messages. That one message came from a stream that was held on YouTube instead of Twitch that was at the time incorrectly included in the data. It included only one message, therefore only this category as the final representation was really affected. The mistake was corrected in the relevant files and the paper.