Serendipity Beyond Mass Personalization | MIT Center for Civic Media
Catherine D’Ignazio is the person behind that really cute baby. She is an Assistant Professor of Data Visualization and Civic Media at Emerson College who investigates how data visualization, technology and new forms of storytelling can be used for civic engagement.
Professor D'Ignazio has conducted research on geographic bias in the news media, developed custom software to geolocate news articles and designed an application, "Terra Incognita", to promote global news discovery. She is working on sensor journalism around water quality with PublicLab, data literacy projects and various community-educational partnerships with her journalism students. Notably, she co-organized a hackathon at the MIT Media Lab called "The Make the Breast Pump Not Suck!" Hackathon.
Her art and design projects have won awards from the Tanne Foundation, Turbulence.org, the LEF Foundation, and Dream It, Code It, Win It. In 2009, she was a finalist for the Foster Prize at the ICA Boston. Her work has been exhibited at the Eyebeam Center for Art & Technology, Museo d’Antiochia of Medellin, and the Venice Biennial.
Professor D'Ignazio is a Fellow at the Emerson Engagement Lab and a Research Affiliate at (and alumna of) the MIT Center for Civic Media.
Serendipity Beyond Mass Personalization
While personalization is useful in moderation, creators that employ excessive personalization might consider a warning message like the above.
This blog post is part of my epic (only to me) quest to write a thesis about serendipity and online information discovery. Here I detail the strategies that creators are using to inspire curiosity and retain user attention that do not require mass personalization. These include novel news delivery, conceptual experiments, wide spectrum/narrow format and including participatory voices.
Many contemporary news applications are following the lead of social websites (read: advertising agencies) and tending towards personalization (read: surveillance) in the vein of Nicolas Negroponte's Daily Me. Examples include News.me, Zite, Pugmarks, News 360 and Match the News among many others. These depend on gathering fairly extensive user information in the form of browsing history and user preferences. Just as Facebook tries to match users to ads, these applications use data collection and algorithms to try to predict what content users will enjoy. While some personalization is certainly helpful for narrowing the spectrum of available information, the concerns with personalization are well-documented and include the idea that personalization reinforces homophily and creates echo chambers, filter bubbles and a Tyranny of the Majority by prioritizing what is popular throughout the system. Researchers Daniel Fleder and Kartik Hoasanager recently showed that in even if recommendation systems create more individual diversity they can actually decrease group diversity of information. I would add to this list the growing public concern about personal data collection, use and ownership. A news recommendation system, particularly a browser extension or app, could easily be a way to simply collect mountains of personal data about users and package it for resale or cross-reference it with other forms of data to create even more detailed user portraits. As public understanding and outrage over data practices mounts, companies will face growing pressure to limit the personal data they collect, disclose how it is used and give consumers an easy way to opt out or risk serious damage to their brand. Facebook is weathering this storm right now.
So what alternatives to personalization (surveillance) exist for news innovation and serendipitous information discovery?
First of all, we can look back to the pre-Internet ways of delivering the news. In my last post, I referenced Ethan Zuckerman's blog post response to Why Audio Never Goes Viral regarding how radio is potentially our most serendipitous medium. And in an overloaded infosphere, news apps that put a premium on brevity, context and tight curation like Yahoo! News Digest are making a comeback. The path to serendipity in this case is as it was previously - a "professional" editor curates the most important things of the day for you. You might see articles about Neymar's back injury next to articles about the insurgence in Iraq. In this case, the "professionals" have the help of algorithms to surface content and multimedia to provide context but it's an interesting return to the model of media as shared experience; Everyone sees the same thing.
The Accidental News Explorer app uses search and topic browsing to encourage users wander through the news. Images from http://accidentalnewsexplorer.com/.
There are plenty of other experiments in novel news delivery that do not rely on excessive personalization. Some are predicated on chance encounters like the Accidental News Explorer app which uses a combination of search and topic browsing to lead the user to new destinations. Umano is a mobile and tablet app that combines some personalization (tracking a user's past reads in the system) with the idea of listening to articles read aloud. Narrators from the community submit news articles they have read aloud which are then recommended to other users for listening. News aggregation apps like Feedly, Paper and Flipboard are pushing the visual dimension of newsreading, providing a lush (but low-commitment, since you can always swipe to the next article) magazine-style interface to the news. And here at the Center for Civic Media, Alexis Hope and Kevin Hu have created Fold, a platform that supports telling news stories with added context and supporting information without having to leave the main thread of the story.
Do these novel news delivery mechanisms support serendipitous information discovery?
Not necessarily. But they represent some very interesting paths news innovators are taking towards capturing and retaining users' attention that do not require mass personalization. If we look a little further afield, there are some really interesting experiments in serendipitous information discovery systems that fall into three other categories: Conceptual Experiments, Wide Spectrum/Narrow Format, and Serendipity Through Participatory Voices.
Jonas Lund, We See In Every Direction web browser, 2013
There are conceptual experiments that posit alternative ways to browse and search for information. In general, these experiments are not meant for widespread adoption but serve to make a point or suggest a different way of discovering information. For example, We See in Every Direction is a collaborative web browser by artist Jonas Lund that allows any number of users into a single browser window. You see potentially thousands of cursors at the same time and have to deal with many users negotiating which links to click on and what to type in the browser bar . Variations on this idea of simultaneous collective browsing have come out of the Boston Globe and New York Times newsrooms. News Intermix, inspired by Zuckerman's book Rewire, is a browser extension that will periodically redirect your browser away from your most visited news sites and towards a set of 900 sites that are more global . So if I have visited nytimes.com too much, News Intermix automatically redirects me to http://www.kashmirnewz.com/, a site run by five independent journalists that has English-language news stories about the disputed region of Kashmir. The redirection happens at a rate that the user decides, but even when you have chosen to be redirected it is quite surprising and unexpected in the moment. Other projects have a humorous take. Automatic Browser, a Chrome extension by artist Brian House, observes your web browsing and then eventually takes over your browser for you . It tells you to "Sit back and relax" while the extension goes to the sites you normally go to. And finally, there are numerous examples of alternate search sites like the glorious bananaslug.com , a "Long-tail Search Engine" which adds a random word to your Google search, or million short which removes the million most popular websites from your search results .
While most of these sites and tools are not meant for daily use, they serve the important purpose of pointing out overlooked aspects of the discovery of information in everyday life. They lead us to provocative questions like "Why isn't our web more collaborative?", "What sites surface once we remove the most popular from our search results?", "Am I so predictable that an algorithm could do my browsing?" These experiments use chance, intervention and alternative algorithms to increase the likelihood that we will find information we are not seeking. They use lightness, surprise and humor to increase the likelihood that we will pay attention to it.
StumbleUpon gives the user a large header bar at the top of the page where she can "stumble" to the next website.
Wide Spectrum/Narrow Format
Other approaches to serendipitous discovery tighten up the form of information presentation while simultaneously widening the spectrum of topics that appear in that form. The website and social network StumbleUpon , for example presents the user with a large black header bar across the top of the page. While the system chooses what website might appear below the header, the header bar and large red button standardize the experience and shift the user into clicking through websites ("stumbling") in an experience similar to a slide show. While the range of information the system offers you might be quite broad, the standardized header gives it a common format and an easy out to see if the next "find" is better. To curate its recommendations, the site combines user-specified topic preferences, collaborative filtering (thumbs-up and thumbs-down voting combined with machine learning) and social (following other SU users) curation. In 2007, StumbleUpon introduced the "StumbleThru" service whereby you can limit your stumbles to a particular domain. News sites you can stumble through include the BBC, the Financial Times, the New York Times and Yahoo! News.
LongReads is a website and a hashtag (#longreads) where editors and readers share high-quality, long-form journalistic storytelling about any topic. The editors of the site curate the top five longreads each week and present them with their word count and expected reading time. This week's stories all clock in at more than ten minutes but range in topic from an in-depth portrait of the trapped Chilean miners to a tour of books authored by ex-lovers of rock stars.
Similarly, delanceyplace is a service that offers a narrow, standard experience in the format of a daily email. Each day's email is an excerpt from a recently released, non-fiction book that the editors have deemed to be interesting. The topics of the recent excerpts range from Napolean's march on Moscow to a history of the practice of bloodletting to an examination of how babies learn. Curation in this case is by humans.
These sites and services approach serendipity by widening the spectrum of relevant information. They are predicated on the idea that people want to see information that they are not seeking and use various curatorial strategies to locate their content. But they mitigate the risk of disorienting users with new information by narrowing the format of information presentation. I previously discussed the balance between novelty and normalization. In these cases the form of delivery - header bar with big red button, hashtag, or email - becomes standard and predictable for the user while the content varies across a very wide spectrum. The organizing feature of these services is the form of information presentation rather than the topical content of the information.
20 Day Stranger is an app by the Playful Systems group at the MIT Media Lab that connects you to an anonymous person's everyday life for 20 days.
Serendipity Through Participatory Voices
Other services consciously reconfigure or redistribute the responsibility of curation in order to introduce novelty and variation in online information systems. This is the logic behind the The Listserve , an email lottery system and list created out of the Interactive Telecommunications program at NYU in 2012, that gives one person a day the opportunity to broadcast to the email list. If your email address gets chosen then you have three days to craft a message to 24,562 potential listeners. People send poems, simple greetings, project promotions and news items.
The government of Sweden took a similar approach with their official twitter account. In a program titled "Curators of Sweden" launched in December 2011, the government began handing over the twitter handle @sweden to a different Swede each week. You must be nominated by someone else and, if chosen, each citizen has seven days to tweet as Sweden. The idea behind the program is that individual voices would curate the web differently and collectively provide a portrait of Sweden "different to that usually obtained through traditional media."
Finally, there are more experimental projects like the mobile app 20 Day Stranger (out of the Playful Systems group at the MIT Media Lab) that create chance connections between individuals that are not part of each other's social networks. When you download 20 Day Stranger you are connected to a stranger in a different location. It's not ChatRoulette - You never know the stranger's name or speak to them directly. You do see what they do and how they move through their day based on the data points that they exhaust over GoogleMaps, FourSquare and Instagram.
In these applications and other experiments in participatory archives (61Fresh, 18 Days in Egypt), the potential for serendipitous encounter is heightened by the widened spectrum of voices participating in the conversation and through the structures of chance (lottery, stranger-matching) introduced by the systems. And in an increasingly corporatized web filled with brands and bots, each of these systems is also predicated on the idea of listening to the individual voices of "real people". Chance, participation and human voices are used hand-in-hand to drive desire, curiosity and encounters with new information.
This post tracks the numerous ways that news innovators are experimenting with serendipity beyond mass personalization. There are many strategies to employ to pique interest, inspire curiosity, and retain attention in relation to news stories that do not require collecting and storing massive amounts of user data. We've explored novel news delivery mechanisms, conceptual experiments like collective browsing and surprise interventions, narrowing the format of information presentation, and experimenting with serendipity based on opening up the field of participation. If you've made it to the bottom of this post I'd love to extend my thanks and also hear your comments and feedback.