Novice Programmers Talking about Projects: What Automated Text Analysis Reveals about Online Scratch Users' Comments

Abstract

In this paper we examine the possibilities of applying predictive analysis to users ' written communication via comments in an open-ended online social networking forum: Scratch.mit.edu. Scratch is primarily used by youth ages 8-16 years to program software like games, animations, and stories; their social interactions take place around commenting, remixing, and sharing computer programs (called projects). This exploratory work contributes to work in educational data mining by broadly describing and comparing comments about projects versus other topics in Scratch. Referencing communication accommodation theory, we found that user comments about projects exhibited different linguistic cues than other comments, and these cues were successfully used to classify comment topic. Further, results also suggest that project comments embody richer language than other comments. This suggests several future avenues for research on youth's online comments about programming and other technical projects that may reveal educational opportunities in creating and sharing projects.[16]. These websites include places where kids can share written stories or fanfiction (e.g., Fanfiction.net, Storybird.com), mods or adaptations of popular games (e.g., Little Big Planet, The Sims), and, as we focus on in this paper, computer programs that can take the form of video games, animations, stories, or art (e.g., Scratch, Kodu). Since the potential for learning when children design or make projects has been well documented for many years [7; 21], it is understandable why there is such excitement over this new phenomenon. Furthermore, sites such as Scratch are unstructured environments where engaged and authentic learning can occur, in contrast to constrained and structured learning environments such as cognitive tutors [4; 22].Yet while websites where people post usergenerated content (UGC) are increasing in number, we know relatively little about the quality of communication on these sites, especially on a large scale that can reveal trends across thousands of users. Further, we know even less about the quality of communication of young people on such websites. Given that children and youth are at different stages of development and that their language tends to differ from adults' language, it is important to study communication on websites dominated by them. Most importantly, communication itself is learning. Developing a better understanding of what children and youth are doing on these emerging types of websites will help us better grasp the opportunities in participating in such sites as well as the challenges to designing for richer communication on websites for children.In addition, there may be value to understanding communication specifically about projects that children create. Although the benefits to learning by making projects are widely discussed [7], sharing and discussing creations may contain particular educational value. Heath [18] argued that sharing

An Educational Coding Laboratory for Elementary Pre-service Teachers: A Qualitative Approach

“…Results are in line with those of other researches [25], [26], [27]; [28], [29], and showed that negative stereotypes of people who are not digital natives as avoidant of technology and incapable of its use are outdated.…”

“I have a tutorial for this”

“…In our earlier study [17], which examined 8,000 peer comments of online Scratch project, we found that only one third of comments (2,273) were about projects while the majority of comments (5,727) were about other topics such as social conversation or current events. A further linguistic analysis revealed that comments about Scratch programming projects also exhibited statistically significant differences in 14 linguistic cues that showed not only higher quantity of overall words, verbs, and modifiers but also suggested various qualitative differences.…”

“…We conducted three separate analyses of the 2,273 comments on projects previously identified [17]. First, we made a distinction between simple and more complex, constructive comments in the data set.…”

“…Out of this random sample of users, 18.5% (926) left comments on users' projects generating 36,802 comments out of which we randomly selected 8,000 to manually code into two primary categories: comments about projects and other comments, resulting in a 1:3 ratio [see 17]. Manual coding from an earlier analysis by Forsgren Velasquez et al [17] generated 2,273 comments on projects, which formed the data corpus for our analyses. We describe these earlier findings further in the results section.…”

“…To examine the language of online support, we build on a previous computational linguistic analysis of a random sample of comments [17]. We found that comments specifically concerning Scratch projects were more likely to have more sophisticated linguistic features than comments about other topics (social conversation, current events, etc.).…”