Streaming Software Development: Accountability, Community, and Learning

As the Twitch platform evolves with additional categories of streams outside of video games, the categories of IRL, Just Chatting, and Educational have more traction with viewers [11]. A 2018 study focused on video game programming streams indicated that this form of live streaming has the possibility to be integrated into mainstream online learning environments [8]. Live streaming has some benefits over prerecorded and structured educational programming videos. Chen et al. found that live streaming has benefits over prerecorded videos due to less upfront planning and financial commitments, which gives streamers a lower barrier of entry to teaching [10]. The ease of getting onto the streaming platform means that it is accessible to those who may not have formal pedagogical training but can still share information and knowledge with others. Streamers are then not limited to traditional ways of becoming a professor or teacher. Prior research surrounding community and developers’ work has shown that streaming platforms benefit developers as a means of developer advocacy and providing insights into software development careers [12].

For viewers, streaming was found to be beneficial for those who preferred an “over the shoulder” style of learning, informational learning, and interest in seeing how others approach a problem set [10]. This style of informal learning that Twitch provides has also been shown to give learners the ability to personalize their learning goals and interests [13]. However, a detriment to viewers, the Twitch platform does not go any further to categorize streams. Viewers cannot search for specific styles of teaching or, in some cases, specific areas of interest, nor does the platform afford that type of interaction style for content creators [10]. The platform attracts those who are seeking information, even in the context of watching others stream video games, with information-seeking positively impacting usage (time spent watching content) [14, 15].

A 2017 experiment with educational streams found that learning performance for viewers of educational streams is roughly the same for being taught by a beginner or an expert [16]. This study also found that the personality of the streamer factors into learning performance - viewers who found streamers agreeable have a positive takeaway from new streamers [16]. These streamer attributes factor into how and why communities of viewers may form.

Hosting a stream on Twitch and other platforms like YouTube and Facebook Gaming is free; there is no cost to the streamer concerning hosting themselves on the platform, same as many prerecorded video hosting websites like YouTube and Vimeo. No upfront monetary costs make the platform accessible with low barriers to entry. In general, streamers are motivated by a need for companionship and a desire to socialize [17]. A 2019 study within low-viewer count programming education streams found that a significant motivation to stream was finding help with their programming and for socialization [9]. This same study also found that these streamers were in a transitional state in life that enabled them to stream regularly and gain access to a place to socialize or a community with similar interests [9].

One of the core elements of the live streaming platforms is the live chatting feature. During streams, a live chat continuously updates with messages from viewers to the streamer, streamer to the viewers, and amongst the viewers. These three types of interactions facilitate community development. Within well-established communities, mentorship is found to be three-fold, mirroring the types of interactions that the platform affords [8]. This indicates that interactions facilitate learning within these communities by the streamer teaching or working through a concept, a viewer pointing out a potential bug, and viewers explaining concepts to other users. Communities around a streamer manifest in a few ways with regulars and off-site communities due to live streaming platforms not having adequate or apparent features to facilitate discussion outside of a stream when it is not live. These communities can be found on Patreon¹¹1A crowdfunding platform to pay people for content they produce; https://www.patreon.com/, Discord, Reddit, and others.

While community and mutual benefit are positive aspects of streamer and viewer interactions, some downsides exist. Streamers have and continue to voice concerns over harassment, privacy, and moderation tools available to them [10]. Volunteer moderators (mods) can monitor chat for inappropriate, hateful comments or spam; however, this responsibility typically falls on the streamer to moderate their chats while live. This downside to the platform dissuades new streamers from choosing to start streaming or is the reason why they discontinue streaming. Many live streaming platforms have not addressed community concerns for better moderation, better ways for viewers to find the content they seek, nor provide better tools to enable creative streamers to engage their viewers in new, more meaningful ways [15]. Additionally, prior work lays the foundation for understanding that streamers, though they might know the potential downsides of harassment, will use the platform to overcome social anxiety and use the negative aspects of the platform as exposure therapy [18].

Limitations exist in prior studies in accounting for all genders for both streamers and viewers [8, 9]. Very recent research has more diverse demographics [10], but there still is a gap in representation. Based on our observations, less than 15% of streamers in software and game development are female, our work begins to reach a more representative sample with 20% of our participants identifying as female or non-binary but female-presenting.

Additionally, while prior work assumes many of the motivations to live stream are the same as motivations for gaming streams or omits development streams from analysis, our work differs by directly investigating developers’ initial and continued motivations to live stream. This larger scope of why developers choose to stream contextualizes the educational benefits of development live streams, provides new data on the beneficial outcomes of live streaming development for streamers and viewers, and extends and confirms knowledge on the benefits of live streaming developer communities [8, 15].

Finally, this paper does not investigate the implications or motivations of financials or monetary gain for streamers, nor the motivations for viewers to give monetary gift donations to streamers. Prior work regarding monetary compensations for non-development streams, often on Twitch through “bits”, advertisement revenue, or another funding platform (e.g., Ko-fi, Patreon, Gumroad, Paypal), indicates that there may be a link between compensation and motivation for streamers [6, 7, 19]. Additionally, while we argue that streaming software and game development has a low up-front cost and low barrier to entry, recent work investigates investing in stream peripherals and equipment aids in audience engagement and interaction [20].

The goal of this work is to better understand the motivations, intentions, and perceptions of software and game development streamers, understand streamer and viewer interactions from the streamer’s perspective, the perceived learning expectations of streamers, streamers’ perceived learning outcomes for viewers, and design implications for live streaming platforms that could enable more people to utilize the platform for educational purposes. Therefore, we ask the following Research Questions (RQs):

• $RQ_{1}$: What are the motivations of developers that live stream?

• $RQ_{2}$: What are streamers’ perceptions of their viewers and how do they perceive their engagement with viewers?

• $RQ_{3}$: What are developers’ perceived beneficial outcomes of live streaming for themselves and for their viewers?

• $RQ_{4}$: What are the challenges developers have while live streaming?

The rationale behind $RQ_{1}$, $RQ_{2}$, and $RQ_{3}$ is to understand a holistic perspective of the streamer about themselves and their motivations, as well as their perceptions about the viewers of their stream. This information helps us to understand more about programming education and developer communities. From there, $RQ_{4}$ addresses challenges that can help influence platform design choices and connects back to the platform as an educational and community-forming media for those interested live streaming their programming.

We designed our study to elicit qualitative data through semi-structured interviews. We interviewed participants for 40-70 minutes about their experiences and thoughts on streaming, live coding practices, perceptions of their viewers, and thoughts on the live streaming platform. We started each interview by collecting demographic information and moved into a semi-structured and open-ended question-answer format. We asked the participants about their programming and educational experience, stream-specific questions, interactions with viewers, and then ended the interview with questions about their perceptions of their viewers and any community they may have formed through streaming.

The authors identified interview participants by browsing live streams on Twitch between the hours of 6 a.m. and 12 a.m. Eastern Standard Time (EST) for seven months, selecting streamers at random from who were at the top of the Software and Game Development category (most current viewers) to the bottom of the category streamers with low viewer counts (typically less than five viewers), noting down usernames and using the Twitch platform to identify preferred means of contact. These broad hours were chosen to reach international streamers and those who may not fall into typical popular streaming times²²2Popular streaming times vary, but typically fall between 3 a.m. EST and 11 a.m. EST, and again from 2 p.m. EST to 5 p.m. EST. [21, 22], and to sample from streams with varied viewership numbers. For each streamer, we would observe the live stream for 15-30 minutes to ensure that the content presented was primarily English-speaking, and relevant and on-topic to software or game development. During our observations, we would note down aspects of the stream’s presentation and experience (i.e., overlays, interactive elements), engagement of the audience, and streamers’ disposition. However, we do not take these elements into account in our analysis as they would fit better into further research into stream communities and nuanced research from viewer perspectives. We excluded streams where the streamer was not present (i.e., an empty chair with no “Be Right Back” screen), streams with non-software and game development-related content, non-English speaking streams, and overlaid content blocked the primary view of the stream content. We noted subscriber counts and current viewers directly from their Twitch page, and used TwitchTracker³³3A third-party Twitch streamer statistics tracker. https://twitchtracker.com/ to gather historical viewership averages.

We conducted and recorded interviews between February and September 2021 using Zoom and Discord voice calls, recorded with participant permission, and anonymized transcripts. We present a sample of interview questions in Table 2. No incentives were offered to participants to complete the interview. Interviews began with basic demographic information before moving into general streaming questions, questions specific to participants’ streams, and finally, questions about their interactions with chat.

In total, we sent 103 recruitment messages through email and Twitter direct messaging. We collected 15 responses for participation (response rate 14.5%). Of 15 participants, 2 were females, 12 were males, and one was non-binary. The age of participants ranged from 23 to 55 years old, with a mean age of 34 and median age of 31. Participants resided in the United States, Germany, the United Kingdom, Australia, or Canada. Most participants (11 of 15) indicated they were self-taught programmers. Interestingly, 9 of these 11 self-taught participants had indicated prior formal computer science or STEM-related education. Self-taught participants, like P1, noted they learned “development fundamentals” through work experience post graduation, same with P3 who entered the workforce as a developer before beginning his formal STEM education. Participants like P6, P11, P12, and P14, all noted that they used online resources and personal projects to teach themselves software development and other programming skills. Additionally, participants like P2, P5, P10 and P11 expressed interest in hobby programming before engaging in formal education. Programming experience ranged from 3 to over 25 years of experience; 1 participant had less than 5 years of experience, 4 participants had 5-10 years of experience, 3 had 10-15 years of experience, 2 had 15-20 years of experience, and 5 participants had more than 20 years of experience. Streaming experience ranged from less than a year to 8 years of experience, with five participants noted having streaming experience before starting their software and game development streams. Table 1 presents a breakdown of participant demographics with participant identifiers (PID) and stream statistics. Please note, we do not provide all demographics or stream statistics to protect participants’ anonymity.

The data that we collected includes the streamers’ programming background with years of experience and languages they commonly use on and off stream, educational background, work experience as a developer, and time they have spent streaming. Additional data we collected includes insights into streamers’ motivations to start and continue streaming, preparation for a stream (i.e., if preparing for a stream, how?), and notable successes and challenges of streaming as a whole. To understand the interactions a streamer has with chat, we collected data regarding the frequency of interaction with viewers, what a typical interaction with a viewer entails, how the streamer interprets chat suggestions for programming errors or bugs, and if they have any routine regulars to their streams they recognize. Taken together, streamers’ development backgrounds, streaming motivations, and interactions with chat members and viewers provide a holistic high-level understanding of how and why developers use live streaming.

We treat interview transcripts as qualitative data and analyzed them using thematic analysis and qualitative analysis guidelines outlined in McDonald et al. [23]. Using the main themes from the initial three interviews, we identified similar and additional themes in the following data and organized them into broader themes to identify any relationships and reoccurring themes between the participants and of live streaming development as a whole. During the data collection, the first author would revisit earlier interviews to assess and identify any new potential themes missed that were realized during the analysis of subsequent data and correlate consistent themes between the interviews.

Upon completion of the interviews, we reviewed the data and, with high-level themes, identified and organized sub-themes related to motivations to stream, audience members’ perceptions, skill improvement, and challenges of streaming. The authors met to iterate, discuss, and refine themes and sub-themes.

To understand developer motivations to stream ($RQ_{1}$), we asked streamers questions focused on why they chose to stream, how they became involved in streaming, and their continued motivations to live stream. Though programming experience, education, and years of streaming varied among the participants, they shared much of the same starting motivations, continue to share some of the same starting motivations, and have evolved as their streams have matured. We found that starting and continued motivations converge around five themes - (1) accountability, (2) self-education, (3) teaching and mentoring others, (4) visibility of the streamer’s work, and (5) finding a community with shared interests. Throughout this section, we discuss the themes and present quotes from our participants. Table 3 presents a high-level overview of each theme we identified for $RQ_{1}$ with top-level attributes of what the theme entails.

• “I was working on a side project. And basically, before doing that I was just sitting around not doing much. So streaming is essentially just a way to hold myself accountable to doing the work and makes it a bit more sociable.” -P10

• “Live programming on Twitch was a great form of accountability for learning new stuff. And so like, my initial motivation for doing this was basically, I knew I wanted to kind of more regularly do some practicing. And, you know, I didn’t really care how many people showed up, but this is going to be my, my public commitment to do actually working on some of this stuff.” -P8

We found streamers held personal accountability and responsibility to provide to the community they fostered while streaming. Streamers reported that the community they built motivates them to hold and keep a schedule out of respect for their followers, saying:

• “I feel a little bit of a responsibility as well. I’ve got kind of, yeah, like a community of people who, you know, they expect on, you know, during my stream schedule, hours, they expect me to be there. And, and I expect to be there. So, it’s a bit of a contract at this point.” -P7

• “So if you build a community and just keep a consistent schedule, that’s super helpful. And both, you know, boosting morale for everybody. And, you know, accountability.” -P14

• “Sometimes if I get stumped, rather than Google, people will just chime into chat and say try this. And that interaction that back and forth is, is really, you know, it’s almost like being in a room of people. And they’re just, you’re shouting out, I’ll try this, click here, type this. And, and I like that. I like that. It makes it much more like a pair programming type of thing.” -P11

P6, P9, and P10 also shared experiences where they learned from their audience who used the text chat to explain a concept, offer a possible solution, or point out a bug in the code.

Additionally, we found a subset of streamers in both groups who wanted to be a mentor and advocate for developers. One participant, P4, made teaching and mentoring others his main motivation for streaming, saying they wanted to offer “free teaching.” This desire came as many schools moved online due to the pandemic and used it to give back to people who might be missing out on in-person opportunities or were taking the time at home to learn something new. Having a background in education made the transition easy, and they translated their work into a multi-day workshop where people could come and learn as though it was a real class setting. They (P4) also streamed weekly for different development concepts while offering the multi-day workshop/class once or twice a year to make what they do “open and freely accessible, and not locked away, behind paywalls” as they felt other online educational content did.

Other participants, with secondary motivations for teaching and mentoring others, wanted to share the knowledge they have because they think they are doing something unique, have years of experience as a developer and want to share with the next generation of developers, or they want to teach others while teaching themselves something new. P11, an experienced software developer with more than 20 years of experience, explained that he believed many development tutorials leave out the challenges and struggles a developer faces when trying to accomplish something and explained his process of streaming as:

• “If it took me a long time to do something, why not share that? So other people don’t have to spend a long time to do it. That’s one of the things that I think I do well on my stream. … Here are all the problems I have, here are all the potholes that I almost fell into, versus being something polished and clean. [By] sharing step one through step four, with the potholes and what you ran into between each step, I think it helps a lot of folks.” -P11

Much like P11, P3, with many years of development experience, explained they are passionate about teaching others and will take questions asked on stream, address them live, and then turn them into short YouTube videos for small concepts and items that could help others who potentially missed a stream. P5 also reported that educating others is a motivation because it is a feel-good moment for him when “you know, you did a good deed, you’re helping out the next version of people” and seeing them go on and have successes that they bring back and attribute to the streamer.

To teach and mentor, we found that 14 participants adopted think-aloud techniques to give viewers a stream of conscious thought process and walk them through what the streamer is typically thinking at a high level, unless deliberately prompted by a chat member to explain what is going on with more detail. P7 explained his thought process as “being able to frame a bunch of information in a way where someone at the end of it knows how to do something” because he felt it is both beneficial for his development processes and others.

• “Twitch drives a lot of my content. And I post that content on YouTube. And in other places, and the community, to me that the, my finances and my revenue are based around my community, and that community lives on a bunch of different platforms, but I spend most of my time on Twitch.” -P7

For the sole software developer, P8, who indicated visibility as a secondary motivator, they advocated for a niche programming language and stream to help others understand the language and its role in certain types of development. They also streamed for visibility and advocated inclusively for LGBTQ+ and female representation in software development.

We found for five participants that being at home drove them to seek out social interaction, community building, and engaging with other developers online through live streaming. P5, who resides in Redmond, WA, home to many technology companies and influential software developers, wanted connections with developers outside of his area, feeling like they might bring new ideas to his chat and wanted to connect with “a new audience.”

While some sought to create a community for others, one formed around them and became a motivator to stream once established. P3, who reported an established reputation in the Python community-at-large, notes “I’m going to be doing this work anyway. So I might as well do it while I’m interacting with people,” indicating that he would be working on development anyways and would rather have an outlet to create and interact with others while doing it, and later stated that “the community aspect is a big reason that I do it.” Still for others, like P4, whose primary motivation was to teach others by streaming, felt “like I’m giving something back to the community. And people seem to appreciate this. And that keeps my motivation up. And so it becomes like a kind of reciprocal thing.” P7 echos many of the same feelings and sees the community he has built not just as a community but as “friends. And not only do I want to see them and interact with them, and keep them updated on what I’m working on.” He also reiterated the feeling of reciprocal giving and the “contract” he has made with his friends and audience. P9 also emphasized seeing their audience as friends and that streaming is the only way they get to “see their friends.”

Community feedback is an important part of streaming. Whether it is pointing out bugs in code, feedback on game mechanics or art style, and providing suggestions, streamers welcomed the interaction and the opportunity to “rubber duck⁴⁴4A method of debugging and problem-solving technique where the developer thinks aloud to an inanimate or or abstracted object and explains line by line what they are doing with code; http://lists.ethernal.org/oldarchives/cantlug-0211/msg00174.html” with the community. This feedback often leads to self-education and is expanded upon in Section 5.3 Improving Development Skills [24].

To better understand the audience of development streams, we asked streamers about their perceptions of their viewers and what they notice and generalize about them. Additionally, we asked streamers their perceptions of stream engagement activities from viewers like backseat coding⁵⁵5Backseating is a phenomenon where the person or people not in control tell those who are what to do and how to do it, with and without prompting. Backseat coding and backseat coders are those who tell someone who is programming what or how to do something within their code. and contributions to open-source software (OSS) or the streamer’s own projects.

Many participants commented on perceiving most of their viewers and followers to be younger than them, from ages 13-18, and that they were relatively inexperienced coders. Others reported they had a mix of age ranges and experience levels, even noting instances when chat members with more experience chime in with bug fixes or alternative solutions to a problem. Participants who believed they had a younger audience shared that they had instances of a viewer thanking them for teaching them about aspects of software development, being a mentor, and inspiring them to pursue programming. P5, noted that he gravitates toward mentoring younger people on his streams who ask novice questions and inquire about what professional development entails:

• “Most of these high school kids, they’re looking for something, you know … helpful, fulfilling [to] them. And they have all the time but they don’t really have necessarily, somebody to spend it with, and sometimes they just would rather spend it with me, which is okay. … And so if I can present them a little bit of my world, of some of the things that at least I know fairly well enough, that maybe I can inspire them to do the same sort of things.” -P5

Another participant (P7), who also noticed younger viewers wanted to have a one-on-one with him, felt these interactions might disrupt his regular streaming habits but would engage carefully to “never discourage anyone from contributing” to the stream. We found P11 had a similar experience with a viewer who was “brand new to programming” and participated in many of the streams and is now a full-time developer with the help of P11’s streams, in addition to other self-teaching resources. This viewer kept in contact with P11 and actively participates in streams. While P11 does not believe that this viewer learned everything from him, he believes he enabled this individual to explore software development and gave him the tools he needed to succeed. We found that the developer advocate for a niche computing interest (P8), spoke about a regular viewer who had only a “general interest” in the field and began participating in her streams. This viewer ended up working for a large technology firm in this niche interest a few years later and felt like she gave this person “the context and general understanding of the field” over specifics of a programming language.

While not everyone could definitively share their experiences of viewers’ successes, all participants shared instances of viewers’ backseat coding and perspectives of how viewers engage with the stream. Many streamers who started small would actively engage with each viewer in chat if they spoke up or asked a question. Engagement from viewers materialized as backseat coding, where viewers would ask why the streamer is implementing something in a particular way, offering solutions as to how they would approach a problem, and pointing out potential problems in the code. Streamers noted these types of interactions were helpful and productive more often than not. Backseat coding offered streamers an easy way to interact with their viewers and would facilitate additional discussion amongst viewers.

• “I had a gentleman that contributes to the [Arch Linux] kernel. He was in my stream, way more knowledgeable about the topic matter than I was. And we basically just became best friends. We’re in it to learn together.” -P5

• “There was somebody that I know from the community and she chimed in, and she said, why are you doing it that way? And then it ended up being that culminated into about three weeks of us doing a back and forth development on this.” -P11

• “I do make mistakes, because I’m human. And [viewer-name] is pretty good about pointing out mistakes in that case. So like, when I’m working on something that’s easy enough for people to follow, they’ll be, you know, pointing out typos, syntax errors, that sort of thing.”-P3

• “You can do things 20 different ways. And I do it one way, and somebody suggested, you could also do it another way. And that, to me, were always good suggestions, but it was never that somebody is really coming into critique the way I’m doing things.” -P4

Participants indicated that engagement with viewers, through backseat coding or prompting discussion is the best means of growing and continuing to grow their community. For streams with less engagement, P11 would actively call out viewers using chat features if they saw people were watching but disengaged:

• “I constantly call out especially if it’s quiet. I always go to the chat window and I click on the drop down and to see users in chat, because you can see users in chat. And and so I bring that up, and I’m like, are there really people there because they’re being really quiet. And I’ll see that there are people there. So then I’ll just go and say, suppose say something quite typical like ‘Hey, thanks for coming. I’m glad to see there are some of you out there, please say hi.’ You know, ‘let me know if you have any questions.’ And sometimes people will chime in.” -P11

Some participants whose streams have gradually changed noted the different types of viewers and how they engage with the stream - those who are there for the technical aspects and those there to consume a general type of content. We found that streamers must balance their interactions with each type of viewer in different ways not to alienate or bore other audience members while still trying to keep both engaged and entertained by their content. P12 shared his experience with changing the content of his stream:

• “I saw the audience changed when I switched away from my MMO [gaming streams]. So some people were there from that, some people were there for the game development. When I started to do the WordPress alternative [streams], the audience changed. So just deduction says some people were there for the game development, some people were there for other stuff. And you can tell that other people have just been there for all of it, like somebody has been subbing for, like, nine months now. And, like, they’re just buying me a coffee, right?” -P12

P7 discussed how even though their content changes because of the aspects of game development, they were remembering usernames in chat and having respect for the viewers who are retained over extended periods of time.

• “I’ll still remember them when they come in. And those people who are contributing in that sense, either whether it’s like, just supporting me supporting the project from wanting to be here kind of thing or inputting ideas, those people I kind of always have time for, and I always have like a pocket of my brain that I can like, yeah, I can always address them and they respect the amount of time they take up as well.” -P7

Viewer retention is an essential aspect for many participants as it helps gauge the growth and health of their community. Additionally, retention of viewers provides positive feedback about their stream if people return consistently over time. For those that return, streamers echoed the sentiment of P7 that recurring viewers have an investment in what the streamer is providing.

We asked participants questions about their successes encountered while streaming and how those events affected them personally and within their streams. As many used streaming to hold themselves accountable or learn new concepts, the main benefits we observed surrounded improved programming or development abilities. Additionally, to understand this benefit for streamers, we asked participants if they perceived viewers experiencing programming skill improvements by watching their live streams.

Another participant (P14), did something similar with a built-in development day for “Feedback Friday” where he took feedback from game testers and tried out changes or implemented features that audience members and game beta testers requested. Another participant (P12), after several years of streaming, felt that they would finally be employable in the language they chose to learn on stream. Others who used accountability as their start motivation have continued to stream regularly over several years.

We found several participants who did not expect to learn from chat members but ultimately found themselves in the situation where they had. P7 noted, and P10 echoed the same sentiment:

• “I don’t have any expectation that I’ll learn from the community, but it has proven to be the case where I have done. I mean, I’ve definitely improved as a programmer, like, a lot.” -P7

• “People in chat have been very knowledgeable about the thing that I’m doing and have been able to help me.” -P10

We found P11 referring to his interactions and learning from the chat as “like pair-programming⁶⁶6A development method where programmers work together at one computer, often where one developer writes the code and the other observes and reviews the code for mistakes, bugs, or gives suggestions. https://en.wikipedia.org/wiki/Pair_programming”, especially when faced with a challenge that viewers have potential solutions for or experiences with the same issue. When questioned about his feelings of viewer learning, P13 expresses that it was “more so [himself]” learning from chat members and that the roles were reversed.

For some, like P3 and P13, viewers shared their thoughts with streamers like “your content has helped me become a better software engineer, I have learned techniques that I would not have been able to learn elsewhere just from watching you code” or simple expressions in chat like “I understand that now.” Expressing to the streamers that viewers think they have become better programmers through watching their streams. P6 shared that viewers expressed “positive feedback” with her when they were able to learn something from her stream and apply it in their own work. P7 had a viewer come to him for game critiques and feedback on their own game they created.

Additionally, P2 shared that one of his viewers and long-time community members began to help others in their Discord’s designated help channel, as well as contributing to Twitch bots on the streamer’s channel. While not distinctly measurable, most streamers felt they positively impacted others, whether that was helping learn a new approach to a problem or providing motivation and “drive to keep going on their [viewer’s] projects.”

To understand the challenges streamers faced, we asked streamers directly about the challenges they have faced with streaming and while streaming. We found challenges faced by streamers fell into interpersonal or platform-related challenges. To gain deeper insights, we asked questions about what led up to interpersonal friction and what live streaming platforms could do to improve or lessen the impact of challenges.

While not a result of harassment, some streamers adjusted their stream as they grew. These adjustments were not an easy transition when gaining new viewers and followers. P7 noted that one of the biggest challenges in his streaming career has been adjusting to increased audience size and having more “backseat coders.” Interactions scaled from few to many as viewership increased and the drive to interact with all of them was no longer feasible, saying:

• “The impulse I have is to is to engage directly with all of them, because that’s how I’ve been doing it this whole time. And just from a productivity standpoint, it can just go right off the rails and it actually makes it harder to watch as well. People don’t enjoy the streamer getting like sidelined over and over again, trying to address individual concerns or repeating concerns because like, people just jump in, they’ve just started watching and they haven’t got the context of the last 20 minutes.” -P7

Additionally, we found that P3 echoed the same experience of his stream and that his development-to-interaction ratio weighted more toward interactions as his stream grew. He indicated as his community and stream grew, he now “spends probably 20 to 30% of my stream actually writing code, and then the remainder of the stream addressing questions from chat or interacting.”

Since beginning our study, Twitch revised categories by separating Software and Game Development (S&GD) out of S&T into their own category with the same moniker. Many migrated to the new category but expressed hesitation because they built their platform in a different category, and viewers may overlook a new category to search in, as P11, a software developer, noted:

• “I’m glad they finally put a programming category. But by the same token, nobody’s switching to it because it would hurt them. It would make them less discoverable on a platform that’s already terrible for discovery.” -P11

Still, with the new category, some game developers expressed that they would like game development to be separate from software development. They felt that while the new category was closer to what they needed, however still felt unseen and less likely to be discovered because game development synthesizes programming and art, as P7, a game developer, explained:

• “A lot of the time, I’m working on something that doesn’t actually fit into science and technology, it might be concept art, or pixel art or something like that. That becomes more creative. That category doesn’t exist anymore. Now, it’s like we fit into IRL, or under Science and Technology. There’s no way to actually browse for Game Dev, outside of searching for the tag.” -P7

Streaming things like concept art and character modeling felt out of place to game developers in the S&GD category, but felt if they stream in the Art category, the replacement for Creative category, they would not be as discoverable or lose current audience members when they switched back to programming aspects of the games. The mismatch between appropriate categories and lack of tags for discoverability left game developers feeling it was impossible for them to grow and maintain an audience. Additionally, other platforms like YouTube Live and Facebook Gaming have similarly difficult ways of finding live channels related to software and game development, with tag searching generating prerecorded or irrelevant content.

Streamers who use live streaming as an accountability token to practice and learn new development skills found that they could improve their skills over time (the largest beneficial outcome we found in $RQ_{3}$.) Prior research shows think-aloud as a form of learning, and verbally walking through a piece of code or through the thought processes of figuring out a bug potentially benefits both streamer and viewers [25, 26]. Streamers noted they can walk through an issue and use chat as a “rubber duck” to problem solve and receive feedback on their projects, as well as use the audience as a distributed pair-programming or mob programming entity [27]. Using the audience in these ways and thinking aloud through a problem can be faster than searching through traditional forum posts. This may lead to a more creative solution and skill improvements for the streamer and viewers who participate in the problem-solving process. Think aloud and “rubber ducking” during streams may also afford the opportunity for streamers and viewers to show off their own development knowledge and be a fulfilling experience if their solutions correct the issue at hand.

Self-education appears to be a strong motivator for live streaming that provides beneficial impacts for developers. Streamers using live streaming as a tool to prove themselves as capable was effective for many. We believe that using live streams as a tool for continued education allows developers to be well-rounded and provides a breakup from the monotony of full-time careers. Additionally, streaming provides a real-time feedback loop for developers and viewers alike that is more instantaneous and self-gratifying than relying on forums or trial-and-error programming approaches. We believe this feedback cycle may be strengthened as streamers gain followers, as viewers who might be familiar with the stream might be more eager or willing to speak up and provide feedback.

Pivotal moments of change in motivation to live stream were gradual as stream viewership increased. As more viewers showed up, it was inevitable that streamers needed to start being more conscious of how they interact with their audience and adopting moderators for chat interactions. A rise in viewership can take out the intimacy of one-on-one discussions that gained followers and jeopardize the sense of community streamers fostered. Though a rise in viewership may pose interpersonal challenges, we believe that once over the initial change that increased viewership strengthens community bonds and provides additional motivation for streamers.

One participant (P2) spoke about the accountability of streaming was not only to benefit himself and improve his skills and make progress on projects, but also to the benefit of others who watch him. We speculate that accountability for the streamer, who takes a leadership role for their audience, can produce accountability and learn within their sphere of influence [31, 32]. Providing accountability for others in a streamer’s sphere of influence may help viewers commit to themselves, provide self-discipline, or help form habits that influence the development skills of those around the streamer.

Self-education, wanting to learn outside of a formal class or academic media, might influence whether or not they choose to stream. Prior work has shown that self-directed learning for software developers has been seen as essential for creativity and contributing to OSS and is often practiced by those who are employed in a technology company [33, 34]. The majority (73%) of our participants identified themselves as self-taught developers, even though they received formal education in computer science or technology-related field, and found that streaming reinforces streamers’ self-education habits and desires. P2 indicated that streaming is his hobby and that he uses streaming to keep his skills to a higher standard, and P12 wanting to learn a new language to be employable in it. We believe that self-education might also be an influence to stream for some, as the act of putting themselves out there to be watched while learning something new, immersing themselves, being vulnerable to making mistakes, and walking through the solutions and roadblocks can be challenging, yet rewarding. We believe this to be both a motivation factor for streamers and have directly beneficial skill improvement implications for viewers wanting to learn by watching someone else, shown to be beneficial in prior research [35, 36, 37].

For the majority of the participants, community factors into their motivations and reasons to continue to stream. The communities surrounding streams are social, technical, and informational in nature. These aspects appear to lend themselves to being a core element of those who use live streaming as a learning platform for software and game development, and have been found to form similarly within Twitch gaming communities [17]. Prior work also shows that developers who video blog (vlog) end up forming small communities where viewers come together in shared experiences over the videos developers share [12]. Streamers found that once they have viewers and ultimately a community, they were more motivated and eager to stream and looked forward to socializing with their audience. These interactions appear to be symbiotic with streamers receiving socialization and help from their audience, and their audience receiving socialization and some aspect of learning a new concept or exposure to something new. Community also played a part in information distribution for OSS, as P1 indicated, and as prior research demonstrates this same importance of community in OSS [38, 39]. Development live streams may provide a unique approach to information distribution to community-developed applications and a way to update non-technical and technical community members alike with new features and future development work for an application or release cycle.

Streamers looked forward to, befriended, mentored, and watched their audience learn new ideas and programming concepts ($RQ_{2}$) Additionally, the majority of our participants stated they wanted to be developer advocates. By streaming, they want others to see their problem-solving processes, approach to new and unique challenges, and breaking stereotypes of who developers are and why everyone can learn to be a developer. Prior research shows that actions like vlogging can break down these stereotypes and foster healthy developer communities [12]. We observed that development streams and streamers, through altruism, giving back knowledge, and finding fulfillment in sharing knowledge may also work towards breaking down stereotypes and encouraging community participation. Contrary to views of Twitch having unruly and toxic audience members, seen in gaming streams, we observed that most streamers’ audiences are present with genuine and positive intentions [40]. These genuine and positive intentions could be a product of the stream content, but also due to smaller audiences where streamers and viewers have the chance to interact at a conversational pace. However, from our observations, we note that streaming style may change as streams grow in audience numbers. Additional audience members may take concentration and priority away from development to focus more on viewers’ needs and discussions, which could lead to streamers feeling like they do not accomplish what they set out to work on. However, most participants felt okay with this shift of priorities as it fills a social need for them, but have expressed that sometimes it can be frustrating not to complete the task they set out to in a reasonable amount of time.

We observed that streamers’ perceptions of audience members are overwhelmingly positive, and positive interactions are often an essential part of the stream. Streamer and audience relationships in the context of S&GD are symbiotic social structures where the exchange of information flows and social and communal bonds are built. By forming and promoting positive, healthy communities, we speculate that software development streams form a positive feedback loop for streamers and viewers. This feedback loop promotes maintaining or learning new software development skills and positive attitudes toward development. Our work contributes to a deeper understanding of the relationships in developer communities, social interactions, and habits of software developers.

However, for participants that did face harassment issues while streaming, it is not an uncommon occurrence on streaming platforms or within development circles as seen in prior work [41, 42, 10]. We believe that the continued fostering of a positive community, developer advocacy through the representation of all types of streamers and stream types, and streamer desires to mentor could change this trend within software and game development communities as a whole [43]. Additionally, the inclusion of moderators has shown to be beneficial in alleviating stream challenges, which several of our participants have implemented by recruiting friends or community volunteers [15].