Awards and citations at computing conferences

I’ve heard a surprising “fact” repeated in the CHI and CSCW communities that receiving a best paper award at a conference is uncorrelated with future citations. Although it’s surprising and counterintuitive, it’s a nice thing to think about when you don’t get an award and its a nice thing to say to others when you do. I’ve thought it and said it myself.

It also seems to be untrue. When I tried to check the “fact” recently, I found a body of evidence that suggests that computing papers that receive best paper awards are, in fact, cited more often than papers that do not.

The source of the original “fact” seems to be a CHI 2009 study by Christoph Bartneck and Jun Hu titled “Scientometric Analysis of the CHI Proceedings.” Among many other things, the paper presents a null result for a test of a difference in the distribution of citations across best papers awardees, nominees, and a random sample of non-nominees.

Although the award analysis is only a small part of Bartneck and Hu’s paper, there have been at least two papers have have subsequently brought more attention, more data, and more sophisticated analyses to the question.  In 2015, the question was asked by Jaques Wainer, Michael Eckmann, and Anderson Rocha in their paper “Peer-Selected ‘Best Papers’—Are They Really That ‘Good’?

Wainer et al. build two datasets: one of papers from 12 computer science conferences with citation data from Scopus and another papers from 17 different conferences with citation data from Google Scholar. Because of parametric concerns, Wainer et al. used a non-parametric rank-based technique to compare awardees to non-awardees.  Wainer et al. summarize their results as follows:

The probability that a best paper will receive more citations than a non best paper is 0.72 (95% CI = 0.66, 0.77) for the Scopus data, and 0.78 (95% CI = 0.74, 0.81) for the Scholar data. There are no significant changes in the probabilities for different years. Also, 51% of the best papers are among the top 10% most cited papers in each conference/year, and 64% of them are among the top 20% most cited.

The question was also recently explored in a different way by Danielle H. Lee in her paper on “Predictive power of conference‐related factors on citation rates of conference papers” published in June 2018.

Lee looked at 43,000 papers from 81 conferences and built a regression model to predict citations. Taking into an account a number of controls not considered in previous analyses, Lee finds that the marginal effect of receiving a best paper award on citations is positive, well-estimated, and large.

Why did Bartneck and Hu come to such a different conclusions than later work?

Distribution of citations (received by 2009) of CHI papers published between 2004-2007 that were nominated for a best paper award (n=64), received one (n=12), or were part of a random sample of papers that did not (n=76).

My first thought was that perhaps CHI is different than the rest of computing. However, when I looked at the data from Bartneck and Hu’s 2009 study—conveniently included as a figure in their original study—you can see that they did find a higher mean among the award recipients compared to both nominees and non-nominees. The entire distribution of citations among award winners appears to be pushed upwards. Although Bartneck and Hu found an effect, they did not find a statistically significant effect.

Given the more recent work by Wainer et al. and Lee, I’d be willing to venture that the original null finding was a function of the fact that citations is a very noisy measure—especially over a 2-5 post-publication period—and that the Bartneck and Hu dataset was small with only 12 awardees out of 152 papers total. This might have caused problems because the statistical test the authors used was an omnibus test for differences in a three-group sample that was imbalanced heavily toward the two groups (nominees and non-nominees) in which their appears to be little difference. My bet is that the paper’s conclusions on awards is simply an example of how a null effect is not evidence of a non-effect—especially in an underpowered dataset.

Of course, none of this means that award winning papers are better. Despite Wainer et al.’s claim that they are showing that award winning papers are “good,” none of the analyses presented can disentangle the signalling value of an award from differences in underlying paper quality. The packed rooms one routinely finds at best paper sessions at conferences suggest that at least some additional citations received by award winners might be caused by extra exposure caused by the awards themselves. In the future, perhaps people can say something along these lines instead of repeating the “fact” of the non-relationship.


Why organizational culture matters for online groups

Leaders and scholars of online communities tend of think of community growth as the aggregate effect of inexperienced individuals arriving one-by-one. However, there is increasing evidence that growth in many online communities today involves newcomers arriving in groups with previous experience together in other communities. This difference has deep implications for how we think about the process of integrating newcomers. Instead of focusing only on individual socialization into the group culture, we must also understand how to manage mergers of existing groups with distinct cultures. Unfortunately, online community mergers have, to our knowledge, never been studied systematically.

To better understand mergers, my student Charlie Kiene spent six months in 2017 conducting ethnographic participant observation in two World of Warcraft raid guilds planning and undergoing mergers. The results—visible in the attendance plot below—shows that the top merger led to a thriving and sustainable community while the bottom merger led to failure and the eventual dissolution of the group. Why did one merger succeed while the other failed? What can managers of other communities learn from these examples?

In a new paper that will be published in the Proceedings of of the ACM Conference on Computer-supported Cooperative Work and Social Computing (CSCW) and that Charlie will present in New Jersey next month, I teamed up with Charlie and Aaron Shaw try to answer these questions.

Raid team attendance before and after merging. Guilds were given pseudonyms to protect the identity of the research subjects.

In our research setting, World of Warcraft (WoW), players form organized groups called “guilds” to take on the game’s toughest bosses in virtual dungeons that are called “raids.” Raids can be extremely challenging, and they require a large number of players to be successful. Below is a video demonstrating the kind of communication and coordination needed to be successful as a raid team in WoW.

Because participation in a raid guild requires time, discipline, and emotional investment, raid guilds are constantly losing members and recruiting new ones to resupply their ranks. One common strategy for doing so is arranging formal mergers. Our study involved following two such groups as they completed mergers. To collect data for our study, Charlie joined both groups, attended and recorded all activities, took copious field notes, and spent hours interviewing leaders.

Although our team did not anticipate the divergent outcomes shown in the figure above when we began, we analyzed our data with an eye toward identifying themes that might point to reasons for the success of one merger and the failure of the other. The answers that emerged from our analysis suggest that the key differences that led one merger to be successful and the other to fail revolved around differences in the ways that the two mergers managed organizational culture. This basic insight is supported by a body of research about organizational culture in firms but seem to have not made it onto the radar of most members or scholars of online communities. My coauthors and I think more attention to the role that organizational culture plays in online communities is essential.

We found evidence of cultural incompatibility in both mergers and it seems likely that some degree of cultural clashes is inevitable in any merger. The most important result of our analysis are three observations we drew about specific things that the successful merger did to effectively manage organizational culture. Drawn from our analysis, these themes point to concrete things that other communities facing mergers—either formal or informal—can do.

A recent, random example of a guild merger recruitment post found on the WoW forums.

First, when planning mergers, groups can strategically select other groups with similar organizational culture. The successful merger in our study involved a carefully planned process of advertising for a potential merger on forums, testing out group compatibility by participating in “trial” raid activities with potential guilds, and selecting the guild that most closely matched their own group’s culture. In our settings, this process helped prevent conflict from emerging and ensured that there was enough common ground to resolve it when it did.

Second, leaders can plan intentional opportunities to socialize members of the merged or acquired group. The leaders of the successful merger held community-wide social events in the game to help new members learn their community’s norms. They spelled out these norms in a visible list of rules. They even included the new members in both the brainstorming and voting process of changing the guild’s name to reflect that they were a single, new, cohesive unit. The leaders of the failed merger lacked any explicitly stated community rules, and opportunities for socializing the members of the new group were virtually absent. Newcomers from the merged group would only learn community norms when they broke one of the unstated social codes.

The guild leaders in the successful merger documented every successful high end raid boss achievement in a community-wide “Hall of Fame” journal. A screenshot is taken with every guild member who contributed to the achievement and uploaded to a “Hall of Fame” page.

Third and finally, our study suggested that social activities can be used to cultivate solidarity between the two merged groups, leading to increased retention of new members. We found that the successful guild merger organized an additional night of activity that was socially-oriented. In doing so, they provided a setting where solidarity between new and existing members can cultivate and motivate their members to stick around and keep playing with each other—even when it gets frustrating.

Our results suggest that by preparing in advance, ensuring some degree of cultural compatibility, and providing opportunities to socialize newcomers and cultivate solidarity, the potential for conflict resulting from mergers can be mitigated. While mergers between firms often occur to make more money or consolidate resources, the experience of the failed merger in our study shows that mergers between online communities put their entire communities at stake. We hope our work can be used by leaders in online communities to successfully manage potential conflict resulting from merging or acquiring members of other groups in a wide range of settings.

Much more detail is available our paper which will be published open access and which is currently available as a preprint.


Both this blog post and the paper it is based on are collaborative work by Charles Kiene from the University of Washington, Aaron Shaw from Northwestern University, and Benjamin Mako Hill from the University of Washington. We are also thrilled to mention that the paper received a Best Paper Honorable Mention award at CSCW 2018!

What we lose when we move from social to market exchange

Couchsurfing and Airbnb are websites that connect people with an extra guest room or couch with random strangers on the Internet who are looking for a place to stay. Although Couchsurfing predates Airbnb by about five years, the two sites are designed to help people do the same basic thing and they work in extremely similar ways. They differ, however, in one crucial respect. On Couchsurfing, the exchange of money in return for hosting is explicitly banned. In other words, couchsurfing only supports the social exchange of hospitality. On Airbnb, users must use money: the website is a market on which people can buy and sell hospitality.

Graph of monthly signups on Couchsurfing and Airbnb.
Comparison of yearly sign-ups of trusted hosts on Couchsurfing and Airbnb. Hosts are “trusted” when they have any form of references or verification in Couchsurfing and at least one review in Airbnb.

The figure above compares the number of people with at least some trust or verification on both  Couchsurfing and Airbnb based on when each user signed up. The picture, as I have argued elsewhere, reflects a broader pattern that has occurred on the web over the last 15 years. Increasingly, social-based systems of production and exchange, many like Couchsurfing created during the first decade of the Internet boom, are being supplanted and eclipsed by similar market-based players like Airbnb.

In a paper led by Max Klein that was recently published and will be presented at the ACM Conference on Computer-supported Cooperative Work and Social Computing (CSCW) which will be held in Jersey City in early November 2018, we sought to provide a window into what this change means and what might be at stake. At the core of our research were a set of interviews we conducted with “dual-users” (i.e. users experienced on both Couchsurfing and Airbnb). Analyses of these interviews pointed to three major differences, which we explored quantitatively from public data on the two sites.

First, we found that users felt that hosting on Airbnb appears to require higher quality services than Couchsurfing. For example, we found that people who at some point only hosted on Couchsurfing often said that they did not host on Airbnb because they felt that their homes weren’t of sufficient quality. One participant explained that:

“I always wanted to host on Airbnb but I didn’t actually have a bedroom that I felt would be sufficient for guests who are paying for it.”

An another interviewee said:

“If I were to be paying for it, I’d expect a nice stay. This is why I never Airbnb-hosted before, because recently I couldn’t enable that [kind of hosting].”

We conducted a quantitative analysis of rates of Airbnb and Couchsurfing in different cities in the United States and found that median home prices are positively related to number of per capita Airbnb hosts and a negatively related to the number of Couchsurfing hosts. Our exploratory models predicted that for each $100,000 increase in median house price in a city, there will be about 43.4 more Airbnb hosts per 100,000 citizens, and 3.8 fewer hosts on Couchsurfing.

A second major theme we identified was that, while Couchsurfing emphasizes people, Airbnb places more emphasis on places. One of our participants explained:

“People who go on Airbnb, they are looking for a specific goal, a specific service, expecting the place is going to be clean […] the water isn’t leaking from the sink. I know people who do Couchsurfing even though they could definitely afford to use Airbnb every time they travel, because they want that human experience.”

In a follow-up quantitative analysis we conducted of the profile text from hosts on the two websites with a commonly-used system for text analysis called LIWC, we found that, compared to Couchsurfing, a lower proportion of words in Airbnb profiles were classified as being about people while a larger proportion of words were classified as being about places.

Finally, our research suggested that although hosts are the powerful parties in exchange on Couchsurfing, social power shifts from hosts to guests on Airbnb. Reflecting a much broader theme in our interviews, one of our participants expressed this concisely, saying:

“On Airbnb the host is trying to attract the guest, whereas on Couchsurfing, it works the other way round. It’s the guest that has to make an effort for the host to accept them.”

Previous research on Airbnb has shown that guests tend to give their hosts lower ratings than vice versa. Sociologists have suggested that this asymmetry in ratings will tend to reflect the direction of underlying social power balances.

power difference bar graph
Average sentiment score of reviews in Airbnb and Couchsurfing, separated by direction (guest-to-host, or host-to-guest). Error bars show the 95% confidence interval.

We both replicated this finding from previous work and found that, as suggested in our interviews, the relationship is reversed on Couchsurfing. As shown in the figure above, we found Airbnb guests will typically give a less positive review to their host than vice-versa while in Couchsurfing guests will typically a more positive review to the host.

As Internet-based hospitality shifts from social systems to the market, we hope that our paper can point to some of what is changing and some of what is lost. For example, our first result suggests that less wealthy participants may be cut out by market-based platforms. Our second theme suggests a shift toward less human-focused modes of interaction brought on by increased “marketization.” We see the third theme as providing somewhat of a silver-lining in that shifting power toward guests was seen by some of our participants as a positive change in terms of safety and trust in that guests. Travelers in unfamiliar places often are often vulnerable and shifting power toward guests can be helpful.

Although our study is only of Couchsurfing and Airbnb, we believe that the shift away from social exchange and toward markets has broad implications across the sharing economy. We end our paper by speculating a little about the generalizability of our results. I have recently spoken at much more length about the underlying dynamics driving the shift we describe in  my recent LibrePlanet keynote address.

More details are available in our paper which we have made available as a preprint on our website. The final version is behind a paywall in the ACM digital library.


This blog post, and paper that it describes, is a collaborative project by Maximilian Klein, Jinhao Zhao, Jiajun Ni, Isaac Johnson, Benjamin Mako Hill, and Haiyi Zhu. Versions of this blog post were posted on several of our personal and institutional websites. Support came from GroupLens Research at the University of Minnesota and the Department of Communication at the University of Washington.

Shannon’s Ghost

I’m spending the 2018-2019 academic year as a fellow at the Center for Advanced Study in the Behavioral Sciences (CASBS) at Stanford.

Claude Shannon on a bicycle.

Every CASBS study is labeled with a list of  “ghosts” who previously occupied the study. This year, I’m spending the year in Study 50 where I’m haunted by an incredible cast that includes many people whose scholarship has influenced and inspired me.

The top part of the list of ghosts in Study #50 at CASBS.

Foremost among this group is Study 50’s third occupant: Claude Shannon

At 21 years old, Shannon’s masters thesis (sometimes cited as the most important masters thesis in history) proved that electrical circuits could encode any relationship expressible in Boolean logic and opened the door to digital computing. Incredibly, this is almost never cited as Shannon’s most important contribution. That came in 1948 when he published a paper titled A Mathematical Theory of Communication which effectively created the field of information theory. Less than a decade after its publication, Aleksandr Khinchin (the mathematician behind my favorite mathematical constant) described the paper saying:

Rarely does it happen in mathematics that a new discipline achieves the character of a mature and developed scientific theory in the first investigation devoted to it…So it was with information theory after the work of Shannon.

As someone whose own research is seeking to advance computation and mathematical study of communication, I find it incredibly propitious to be sharing a study with Shannon.

Although I teach in a communication department, I know Shannon from my background in computing. I’ve always found it curious that, despite the fact Shannon’s 1948 paper is almost certainly the most important single thing ever published with the word “communication” in its title, Shannon is rarely taught in communication curricula is sometimes completely unknown to communication scholars.

In this regard, I’ve thought a lot about this passage in Robert’s Craig’s  influential article “Communication Theory as a Field” which argued:

In establishing itself under the banner of communication, the discipline staked an academic claim to the entire field of communication theory and research—a very big claim indeed, since communication had already been widely studied and theorized. Peters writes that communication research became “an intellectual Taiwan-claiming to be all of China when, in fact, it was isolated on a small island” (p. 545). Perhaps the most egregious case involved Shannon’s mathematical theory of information (Shannon & Weaver, 1948), which communication scholars touted as evidence of their field’s potential scientific status even though they had nothing whatever to do with creating it, often poorly understood it, and seldom found any real use for it in their research.

In preparation for moving into Study 50, I read a new biography of Shannon by Jimmy Soni and Rob Goodman and was excited to find that Craig—although accurately describing many communication scholars’ lack of familiarity—almost certainly understated the importance of Shannon to communication scholarship.

For example, the book form of Shannon’s 1948 article was published by University Illinois on the urging of and editorial supervision of Wilbur Schramm (one of the founders of modern mass communication scholarship) who was a major proponent of Shannon’s work. Everett Rogers (another giant in communication) devotes a chapter of his “History of Communication Studies”² to Shannon and to tracing his impact in communication. Both Schramm and Rogers built on Shannon in parts of their own work. Shannon has had an enormous impact, it turns out, in several subareas of communication research (e.g., attempts to model communication processes).

Although I find these connections exciting. My own research—like most of the rest of communication—is far from the substance of technical communication processes at the center of Shannon’s own work. In this sense, it can be a challenge to explain to my colleagues in communication—and to my fellow CASBS fellows—why I’m so excited to be sharing a space with Shannon this year.

Upon reflection, I think it boils down to two reasons:

  1. Shannon’s work is both mathematically beautiful and incredibly useful. His seminal 1948 article points to concrete ways that his theory can be useful in communication engineering including in compression, error correcting codes, and cryptography. Shannon’s focus on research that pushes forward the most basic type of basic research while remaining dedicated to developing solutions to real problems is a rare trait that I want to feature in my own scholarship.
  2. Shannon was incredibly playful. Shannon played games, juggled constantly, and was always seeking to teach others to do so. He tinkered, rode unicycles, built a flame-throwing trumpet, and so on. With Marvin Minsky, he invented the “ultimate machine”—a machine that’s only function is to turn itself off—which he kept on his desk.

    A version of the Shannon’s “ultimate machine” that is sitting on my desk at CASBS.

I have no misapprehension that I will accomplish anything like Shannon’s greatest intellectual achievements during my year at CASBS. I do hope to be inspired by Shannon’s creativity, focus on impact, and playfulness. In my own little ways, I hope to build something at CASBS that will advance mathematical and computational theory in communication in ways that Shannon might have appreciated.


  1. Incredibly, the year that Shannon was in Study 50, his neighbor in Study 51 was Milton Friedman. Two thoughts: (i) Can you imagine?! (ii) I definitely chose the right study!
  2. Rogers book was written, I found out, during his own stint at CASBS. Alas, it was not written in Study 50.

Disappointment on the new commute

Imagine my disappointment when I discovered that signs on Stanford’s campus pointing to their “Enchanted Broccoli Forest” and “Narnia”—both of which that I have been passing daily on my new commute—merely indicate the location of student living groups with whimsical names.

Heading to the Bay Area

On September 4th, I’ll be starting a fellowship at the Center for Advanced Studies in the Behavioral Sciences (CASBS), a wonderful social science research institute at Stanford that’s perched on a hill overlooking Palo Alto and the San Francisco Bay. The fellowship is a one-year gig and I’ll be back in Seattle next June.

A CASBS fellowship is an incredible gift in several senses. In the most basic sense, it will mean time to focus on research and writing. I’ll be using my time there to continuing my research on the social scientific study of peer production and cooperation. More importantly though, the fellowship will give me access to a community of truly incredible social social scientists who be my “fellow fellows” next year.

Finally, being invited for a CASBS fellowship is a huge honor. I’ve been preparing by reading a list of Wikipedia articles I built about the previous occupants of the study that I’ll be working out of next year (the third fellow to work out of my study was Claude Shannon!). It’s rare for junior faculty like myself to be invited and I’m truly humbled.

The only real downside of the fellowship is that it means that I’ll be spending the academic year away from Seattle. I’m going to miss working out of UW, my department, and the Community Data Science Collective lab here enormously.

In a personal sense, it means I’ll be leaving a wonderful community in Seattle in and around my home at Extraordinary Least Squares. I’m going to miss folks deeply and I look forward to returning.

Of course, I’m also pretty excited about moving to Palo Alto. It will be the first time either Mika or I have lived in California and we hope to take advantage of the opportunity.

Please help us do so!  If you’re at Stanford, in Silicon Valley, or are anywhere in the Bay Area and want to meet up, please don’t hesitate to get in contact! We’ll be arriving with very little community and I’m really interested in meeting and making friends  and taking advantage of my nine-months in the area to make connections!

Forming, storming, norming, performing, and …chloroforming?

In 1965, Bruce Tuckman proposed a “developmental sequence in small groups.” According to his influential theory, most successful groups go through four stages with rhyming names:

  1. Forming: Group members get to know each other and define their task.
  2. Storming: Through argument and disagreement, power dynamics emerge and are negotiated.
  3. Norming: After conflict, groups seek to avoid conflict and focus on cooperation and setting norms for acceptable behavior.
  4. Performing: There is both cooperation and productive dissent as the team performs the task at a high level.

Fortunately for organizational science, 1965 was hardly the last stage of development for Tuckman’s theory!

Twelve years later, Tuckman suggested that adjourning or mourning reflected potential fifth stages (Tuckman and Jensen 1977). Since then, other organizational researchers have suggested other stages including transforming and reforming (White 2009), re-norming (Biggs), and outperforming (Rickards and Moger 2002).

What does the future hold for this line of research?

To help answer this question, we wrote a regular expression to identify candidate words and placed the full list is at this page in the Community Data Science Collective wiki.

The good news is that despite the active stream of research producing new stages that end or rhyme with -orming, there are tons of great words left!

For example, stages in a group’s development might include:

  • Scorning: In this stage, group members begin mocking each other!
  • Misinforming: Group that reach this stage start producing fake news.
  • Shoehorning: These groups try to make their products fit into ridiculous constraints.
  • Chloroforming: Groups become languid and fatigued?

One benefit of keeping our list in the wiki is that the organizational research community can use it to coordinate! If you are planning to use one of these terms—or if you know of a paper that has—feel free to edit the page in our wiki to “claim” it!


Also posted on the Community Data Science Collective blog. Although credit for this post goes primarily to Jeremy Foote and Benjamin Mako Hill, the other Community Data Science Collective members can’t really be called blameless in the matter either.

How markets coopted free software’s most powerful weapon (LibrePlanet 2018 Keynote)

Several months ago, I gave the closing keynote address at LibrePlanet 2018. The talk was about the thing that scares me most about the future of free culture, free software, and peer production.

A video of the talk is online on Youtube and available as WebM video file (both links should skip the first 3m 19s of thanks and introductions).

Here’s a summary of the talk:

App stores and the so-called “sharing economy” are two examples of business models that rely on techniques for the mass aggregation of distributed participation over the Internet and that simply didn’t exist a decade ago. In my talk, I argue that the firms pioneering these new models have learned and adapted processes from commons-based peer production projects like free software, Wikipedia, and CouchSurfing.

The result is an important shift: A decade ago,  the kind of mass collaboration that made Wikipedia, GNU/Linux, or Couchsurfing possible was the exclusive domain of people producing freely and openly in commons. Not only is this no longer true, new proprietary, firm-controlled, and money-based models are increasingly replacing, displacing, outcompeting, and potentially reducing what’s available in the commons. For example, the number of people joining Couchsurfing to host others seems to have been in decline since Airbnb began its own meteoric growth.

In the talk, I talk about how this happened and what I think it means for folks of that are committed to working in commons. I talk a little bit about the free culture and free software should do now that mass collaboration, these communities’ most powerful weapon, is being used against them.

I’m very much interested in feedback provided any way you want to reach me including in person, over email, in comments on my blog, on Mastodon, on Twitter, etc.


Work on the research that is reflected and described in this talk was supported by the National Science Foundation (awards IIS-1617129 and IIS-1617468). Some of the initial ideas behind this talk were developed while working on this paper (official link) which was led by Maximilian Klein and contributed to by Jinhao Zhao, Jiajun Ni, Isaac Johnson, and Haiyi Zhu.

Honey Buckets

When I was growing up in Washington state, a company called Honey Bucket held a dominant position in the local portable toilet market. Their toilets are still a common sight in the American West.

Honey Bucket brand portable toilet. Photo by donielle. (CC BY-SA)

They were so widespread when I was a child that I didn’t know that “Honey Bucket” was the name of a company at all until I moved to Massachusetts for college. I thought “honey bucket” was just the generic term for toilets that could be moved from place-to-place!

So for the first five years that I lived in Massachusetts, I continued to call all portable toilets “honey buckets.”

Until somebody asked me why I called them that—five years after moving!—all my friends in Massachusetts thought that “honey bucket” was just a personal, idiosyncratic, and somewhat gross, euphemism.

Natural experiment showing how “wide walls” can support engagement and learning

Seymour Papert is credited as saying that tools to support learning should have “high ceilings” and “low floors.” The phrase is meant to suggest that tools should allow learners to do complex and intellectually sophisticated things but should also be easy to begin using quickly. Mitchel Resnick extended the metaphor to argue that learning toolkits should also have “wide walls” in that they should appeal to diverse groups of learners and allow for a broad variety of creative outcomes. In a new paper, Sayamindu Dasgupta and I attempted to provide an empirical test of Resnick’s wide walls theory. Using a natural experiment in the Scratch online community, we found causal evidence that “widening walls” can, as Resnick suggested, increase both engagement and learning.

Over the last ten years, the “wide walls” design principle has been widely cited in the design of new systems. For example, Resnick and his collaborators relied heavily on the principle in the design of the Scratch programming language. Scratch allows young learners to produce not only games, but also interactive art, music videos, greetings card, stories, and much more. As part of that team, Sayamindu was guided by “wide walls” principle when he designed and implemented the Scratch cloud variables system in 2011-2012.

While designing the system, Sayamindu hoped to “widen walls” by supporting a broader range of ways to use variables and data structures in Scratch. Scratch cloud variables extend the affordances of the normal Scratch variable by adding persistence and shared-ness. A simple example of something possible with cloud variables, but not without them, is a global high-score leaderboard in a game (example code is below). After the system was launched, we saw many young Scratch users using the system to engage with data structures in new and incredibly creative ways.

cloud-variable-script
Example of Scratch code that uses a cloud variable to keep track of high-scores among all players of a game.

Although these examples reflected powerful anecdotal evidence, we were also interested in using quantitative data to reflect the causal effect of the system. Understanding the causal effect of a new design in real world settings is a major challenge. To do so, we took advantage of a “natural experiment” and some clever techniques from econometrics to measure how learners’ behavior changed when they were given access to a wider design space.

Understanding the design of our study requires understanding a little bit about how access to the Scratch cloud variable system is granted. Although the system has been accessible to Scratch users since 2013, new Scratch users do not get access immediately. They are granted access only after a certain amount of time and activity on the website (the specific criteria are not public). Our “experiment” involved a sudden change in policy that altered the criteria for who gets access to the cloud variable feature. Through no act of their own, more than 14,000 users were given access to feature, literally overnight. We looked at these Scratch users immediately before and after the policy change to estimate the effect of access to the broader design space that cloud variables afforded.

We found that use of data-related features was, as predicted, increased by both access to and use of cloud variables. We also found that this increase was not only an effect of projects that use cloud variables themselves. In other words, learners with access to cloud variables—and especially those who had used it—were more likely to use “plain-old” data-structures in their projects as well.

The graph below visualizes the results of one of the statistical models in our paper and suggests that we would expect that 33% of projects by a prototypical “average” Scratch user would use data structures if the user in question had never used used cloud variables but that we would expect that 60% of projects by a similar user would if they had used the system.

Model-predicted probability that a project made by a prototypical Scratch user will contain data structures (w/o counting projects with cloud variables)

It is important to note that the estimated effective above is a “local average effect” among people who used the system because they were granted access by the sudden change in policy (this is a subtle but important point that we explain this in some depth in the paper). Although we urge care and skepticism in interpreting our numbers, we believe our results are encouraging evidence in support of the “wide walls” design principle.

Of course, our work is not without important limitations. Critically, we also found that rate of adoption of cloud variables was very low. Although it is hard to pinpoint the exact reason for this from the data we observed, it has been suggested that widening walls may have a potential negative side-effect of making it harder for learners to imagine what the new creative possibilities might be in the absence of targeted support and scaffolding. Also important to remember is that our study measures “wide walls” in a specific way in a specific context and that it is hard to know how well our findings will generalize to other contexts and communities. We discuss these caveats, as well as our methods, models, and theoretical background in detail in our paper which now available for download as an open-access piece from the ACM digital library.


This blog post, and the open access paper that it describes, is a collaborative project with Sayamindu Dasgupta. Financial support came from the eScience Institute and the Department of Communication at the University of Washington. Quantitative analyses for this project were completed using the Hyak high performance computing cluster at the University of Washington.