User:Hall1467/CHI 2017 Report

Event name

Conference on Human Factors in Computing Systems (CHI) 2017

Participant Connections

I attended to present work that I had done in the context of OpenStreetMap. Contributor freedom has helped OpenStreetMap experience success (just as it has helped in Wikipedia). However, structured data in OpenStreetMap needs to be highly standardized for it to be useful and so I explored this "freedom versus standardization" tension through an interview study of OSM contributors.

Outcome

There was some interesting work related to wikis/peer production (and also more generally):

• I attended a talk on a domain-specific programming language which can be used to help personalize news articles online, here's the paper . The presenter gave examples of personalized map content and statistics in news articles based on where someone lives. This work raised a question in my mind: has displaying or highlighting personalized content been considered in Wikipedia? For example, say the article about the United States has some content related to the state of Minnesota or say, related to Minneapolis (totally hypothetical examples). The content about these regions could possibly be highlighted in some way for the sake of somebody who is interested in them. If I were reading a big article on some general topic, I think I'd be quite interested in having personalized content highlighted (e.g. regarding, say, the state I live in). I'd think it may help me better understand the role the content that I'm interested in, plays in the concept represented in the (general) article; it would help me relate with the content more. Having said this, I could also see article personalization raising concerns about bias and neutral point of view depending upon what sorts of personalization are allowed.

• The talk directly prior to mine, here's the paper, was on the topic of how "crowd diversity" (specifically "contribution diversity" and "experience diversity") affects "crowd performance" (measured by article quality) in Wikipedia. They defined "contribution diversity" as the variability in contributions to an article. Sometimes, contributors to an article all provide roughly equal numbers of edits to it. This would mean there is low contribution diversity for this article. Other times, a small number of editors do most edits, this would mean there is high contribution diversity for this article. They found high contribution diversity results in better crowd performance. "Experience diversity" is the variability in contributor experience for a certain article. They found that lower diversity results in better crowd performance.

• Another paper, here it is, discussed how Wikipedia contributors' motivations change over time. Two different types of motivations to contribute, instrumental and non-instrumental, were defined in previous work and used in this work. These two types correspond to extrinsic and intrinsic motivations respectively. Over time, motivations shift from non-instrumental to instrumental. A contributor may start in the community and want to have fun, but they might stay for social status that they build. The study took place over a 6 month period with contributors who were newcomers at the beginning of that period. It's interesting (but maybe not surprising) how those who stick around tend to grow more social. As is mentioned in the paper, this work provides more reason to help newcomers feel socially involved in the community.

• Here's some work related to structured data. They did a mixed methods (interviews and some quantitative data analysis) study that sought to understand the process of searching for useful structured data. Many of the interviewees used Google search when looking for a structured dataset (e.g. for data tables). The authors argued that the search process undertaken when doing a Google search is not as well-suited for structured data as it is for unstructured data. They also discovered that certain attributes of a given dataset were valuable to potential consumers in order to understand how to use the datasets and whether the dataset was appropriate. For example, information related to its relevance as well as information related to how past users cleaned the data. In terms of implications for design, the authors focused on how those providing structured datasets can better describe their data in order for consumers to better understand if it fulfills their needs. There may be implications from this work regarding how Wikimedia publishes its data dumps -- this could be investigated further. Interestingly, they talked about Google Knowledge Graph but no mention of Wikidata.

• Here's another work on Wikipedia. This work showed how editors in certain Wikipedia languages were more or less likely to hold discussions on talk pages prior to edits. There were large fluctuations of talk page posts across wikis. Hebrew had the highest average number of posts with 69% more than English Wikipedia, Portuguese had the least with 23% less than English Wikipedia.

Finances

Not applicable

Anything else

Not applicable