Creating Technology for Social Change

Researching Love and Thanks on Wikipedia: CrowdCamp Hackathon Report

Change favors the prepared,” Louis Pasteur once famously noted in a lecture on the nature of scientific observation. The best academic events create moments of highly likely inspiration, and the luckiest ones bring that inspiration into action. That happened for Emily Harburg and me this weekend at CrowdCamp, a two day intensive hackathon on crowdsourcing and social computing research.

CrowdCamp brought together over a dozen gradstudents from across the US to share ideas, brainstorm research projects, and actually implement & test our ideas. The day started out with a fascinating range of creative exercises. We grouped into teams and used design cards to imagine new technology designs in 90 second brain-jams (design something that helps you forget your friends; design something that migrant workers could use to connect across borders, etc). Next, we did a rapid speed pitch session, where rotating groups of four researchers took 90 seconds to talk to others about their passions and find common interests. By the end of the morning and our second round of coffee, the room had grouped into common themes around crowdsourcing education and accessibility. Within these academic MegaZords, the idea went, small groups of Research Rangers could focus on a project that combined, like fire, air, earth, and water to summon a Captain Planet of common research.

During the speed pitch session, I got to meet Emily Harburg, a PhD student in Technology and Social Behavior at Northwestern and alum of Disney Research. Emily and her colleague Mike Greenberg had been working to study gratitude and expressions of appreciation in peer production communities like Wikipedia and WikiHow. I was floored. Over the last two years, I’ve also been slowly building a thread of work looking at gratitude, thanks, and acknowledgment online. Both of us, it turns out, had recently started to archive data from Wikipedia. Emily was archiving WikiLove, and I was collecting data on the Thanks feature. Meeting each other was a major serendipity moment. Over the next two days, we combined our ideas and made huge progress towards a research to study the role that thanks and appreciation play in creative communities online.

Love, Thanks, and Motivations Online

In recent years, designers of cooperative social technologies have been creating technologies to support peer thanks and appreciation, in hopes of improving performance and retaining newcomers by fostering supportive communities. I’ve written elsewhere about the wide range of gratitude technologies used in companies and online communities. On Wikipedia, interpersonal appreciation takes the form of two systems: WikiThanks and WikiLove, which are designed to foster supportive encouragement among contributors.

On Wikipedia and other online communities, researchers have long been interested in the motivations of contributors, asking if treating people better might have an effect their experience and participation. Researchers have looked at the effect of positive and negative feedback on Wikipedia, the effect of upvotes on participation levels and polarization, and the effect of quality control systems on newcomer retention, and the link between participation levels and different kinds of motivations (altruism, self-interest, reciprocity, reputation, etc).

Appreciation on Wikipedia takes two forms: WikiLove and Thanks. WikiLove is a message of appreciation sent between any two people with a Wikipedia account. To send WikiLove, navigate to someone’s User page, click the heart button, select the type of WikiLove you want to send, add a message, and send the note. The receiver will receive a notification via email and the Wikipedia notification system. The WikiLove you created will appear on their User Talk page:

Wikipedia Thanks are more closely associated with individual edits that individual users. To send a thanks, view the edit history of an article, find an edit you especially like, and click the “thanks button.” The receiver will be notified of your appreciation.

Research on thanks and love could yield powerful explanations on matters of the design and governance of crowd platforms. It might also help us understand the role of appreciation in human behaviour. In the area of design, we might be able to quantify the trade-offs between offering personalized messages and simply having a “thanks button.” Qualitative studies have shown that learners in the online creativity platform Scratch prefer personal remix attribution to automated messages, but which of these actually have the most meaningful effect overall? Emily and I are hoping to answer that question. We’re also hoping that research on the role of appreciation in commons based peer production (like Wikipedia) might contribute to general knowledge about human behaviour.

Expressions of thanks have been shown to affect performance in a variety of contexts: the timeliness of juvenile justice case workers, the size of tips in restaurants, the likelihood of repeat mentorship sessions, and the efficiency of volunteer fundraising. The Wikipedia dataset is one of the largest measurable collections of gratitude in a community anywhere, and by studying it, we might be able to learn more about how thanks functions in creative communities, especially in relation to our sense of self-efficacy, reputation, and social worth.

What We Did at CrowdCamp: Analyzing Love and Thanks

At CrowdCamp after comparing notes, the two of us wrote code to download a sample of love and thanks to explore the viability of a natural experiment (based on historical data) versus a field experiment (where we would intervene and observe the effect). At the time, we didn’t realize how easy it was to use the Wikipedia API, so we scraped 9,991 WikiLoves and 10,000 Thanks from the English language Wikipedia site using Python code and the scraper library BeautifulSoup. Across that sample of gratitude, we counted 11,367 givers and receivers. We took a random sample of 1,000 of those givers and receivers and downloaded summary statistics for each of those users.

Using data from our sample of Wikipedians, we created histograms that showed us the distribution of givers and receivers of thanks in relation to the number of edits they have made. The results were promising: it looked likely that we could look at the effect of thanks and loves across all levels of experience:

Distribution of Love and Thanks Involved Wikipedians by Edit Count

What we Did at CrowdCamp: Designing Our Study

Based on this promising analysis, we spent the rest of our weekend creating a literature review on Zotero, designing surveys that we can issue to people who have given and receive thanks, and planning the next stages of our research. Here is the presentation we shared at the end of CrowdCamp:

 

What We’re Doing Next

Since CrowdCamp, Erhardt Graeff (MIT) and Mike Greenberg have joined the project, and we’ve made substantial progress on defining our sampling strategy, acquiring more complete data from Wikipedia, stratifying the sample, and designing a set of natural experiments (within subjects designs and comparison group designs), survey approaches, and potential field experiments. We expect to have our first set of results by the end of the year.

Acknowledgments

Thanks to the CrowdCamp organizers for creating a fantastic environment, and to CMU’s HCI Loft for hosting us. We had an amazing time!

 

Thanks also to Erhardt Graeff and Mike Greenberg for joining as fantastic contributors, to Stuart Geiger, who advised our study design, and to Joseph Reagle, who hosted Nathan to discuss online gratitude with his class on online communities, offering encouragement and inspiration.