Serendipity
[from: Serendip, a former name for Ceylon + ITY] A word coined by Horace Walpole, who says (Let. to Mann, of Jan 1754) that he had formed it upon the title of the fairy-tale 'The Three Princes of Serendip', the heroes of which 'were always making discoveries, by accidents and sagacity, of things they were not in quest of'.] The faculty of making happy and unexpected discoveries by accident.
—Oxford English Dictionary
"Did you ever observe to whom the accidents happen? Chance favors only the prepared mind."
— Louis Pasteur
I've been thinking about serendipity recently. It's a word I've liked since I first stumbled over it: the idea of unexpected discovery enabled by (as Pasteur puts it) the "prepared mind". Three events from last week are bound together in my thoughts by the idea of serendipity. I’ll spin them out in this blog, and hope you'll see why I'm interested in both serendipity and the development of the "serendipity engines".
Here's the first event. I subscribed to a newspaper. Not all that interesting on the face of it, but this is no ordinary newspaper and for the paper and its readers, it's an extraordinary moment. The New York Times has made the decision to make its full electronic edition available only to *digital subscribers*. Here's what was in my email on March 17:
"Today marks a significant transition for The New York Times as we introduce digital subscriptions. It’s an important step that we hope you will see as an investment in The Times, one that will strengthen our ability to provide high-quality journalism to readers around the world and on any platform. The change will primarily affect those who are heavy consumers of the content on our Web site and on mobile applications."
As Dieter and Sunny know all too well, I'm that heavy consumer. I read The Times over lunch and then too often attempt to amuse/enlighten/annoy my colleagues and friends with items that I find there. (Dieter fights back with Der Spiegel online.) And, of course, content from the NY Times shows up regularly in my FGS blog entries.
I also subscribe to the paper edition of the Globe and Mail. It struck me last week that I don't read the electronic NY Times in the same way I read the Globe. I read the news section of the Globe from front-to-back, picking and choosing, but reading the entire op-ed section. I then read the Arts section completely, and pick at the Sports and Business sections, again front-to-back. I don't look at the Life section. (Does that mean I don't have one?)
With the NY Times online, the html page design is cleverly thought out. When I open that "paper", I scan the major headlines at the top first (all in blue), click on links and read a few stories. But then...something completely different happens—and here's where serendipity plays in. Just a little further down the long front page are the headlines of three stories from every major and minor section of the paper, lists of the 10 most emailed, blogged, searched, and viewed stories, and via the "Times Wire" a list of the stories that were most recently updated. This online format dramatically increases the likelihood that I'll stumble onto something unexpected that will grab my eye. It does that without any customization to my particular interests (although the Times does offer that customization). By virtue of its format alone, the electronic New York Times encourages serendipity for its readers. I don't think it's accidental, and it's well worth $3.75 a week.
Here's the second event. Although I'm Acting Associate Dean here at Graduate Studies, I still have a day job in the School of Biomedical Engineering. (My musician buddies keep insisting that I keep my day job. Maybe I'm misunderstanding them...) At present, I'm teaching a graduate/undergraduate course called "Biomechanics in Physiology and Surgical Implant Design". As part of that course, each student writes a scientific review paper on an area of biomechanics that interests them. The yearly crazy quilt of topics provides oodles of fun for end-of-term reading: venus flytraps, power lifting, baseball pitching, body armor, concussion injuries, cochlear implants, insect flight, barefoot running, and so on. I learn a lot and, most of the time, so do they.
Preparing the group to produce this assignment in the form of a scientific review paper is an interesting exercise. Very few senior undergraduate engineering students (or entering BME graduate students for that matter) have a working knowledge of how to (i) search the scientific/engineering literature, and then (ii) turn that information into a cogent, organized, and engaging essay. One of the very best bits of the preparation process comes when my good friend Patrick Ellis spends a couple of hours showing my students how to use the many online search engines available to Dalhousie students (PubMed, Web of Science, Compendex, EMBASE, etc.) and how to turn those search results into electronic documents delivered to their computers. Patrick's the Director of the Kellogg Health Sciences Library and he's been a patient tutor to my biomechanics students over the years: generous with his time and willing to hold their hands when they've needed one-on-one help finding information on arcane topics like, say, the Jesus lizard.
Patrick's an expert on literature searches. After 35 years in this crazy business of science I fancy myself an expert as well—but Patrick's better. Because he understands the design and implementation of search engines—and the pragmatic choices that publishers and libraries make when cataloging and categorizing information—he anticipates what won't work and quickly structures a query that will give well-focused outcomes. That approach serves the students in my biomechanics course well given the limited time they have and the narrow focus of their papers for my course. In fact that sort of focused searching generally works just great for senior undergraduates—and even for Master's students. But—I think it fundamentally fails to do something else, something quite important for senior PhD students and post-docs.
Here's where the serendipity thing comes in—and bear with me while at take a bit of a side road. The use of electronic search engines makes reading the literature, planning future research projects, and writing grants much faster than it used to be. With the likes of PubMed, e-journals, e-document delivery, reference managers like EndNote, PDF paper catalogers like Papers and so on, I can get a focused literature review done very quickly. My students give me those campfire horror story stares when I describe what it was like back in the '70s when you'd troll for research papers in the big, bound volumes of Index Medicus and the Science Citation Index, and then physically go through the paper journals to find articles—often without so much as an abstract to guide you as to whether the article would be useful. I remember when, writing my first grants, I would just sit down with an entire year's paper copies of the Journal of Thoracic and Cardiovascular Surgery looking for something that might be relevant to what I had in mind. The remarkable thing was that, during those Snark hunts, I'd inevitably stumble onto things that I wasn't looking for—but which would turn out to be valuable (maybe not for what I was doing just then, but for another project, another line of inquiry altogether). Nowadays, well-structured queries under modern electronic search engines (the sort of queries that Patrick does beautifully) don't leave much room for serendipity. And the search engines create a kind of focus in the searcher that precludes long, dawdling wanderings through the halls of knowledge where one might stumble upon the next clever thing to do. One searches—one doesn’t browse.
The problem may be that we need different search engines. There has been a growing interest amongst search engine designers at places like Google in providing search tools that anticipate a user's needs by knowing the searcher well, and anticipating needs—even without a search being initiated. The idea is to produce a search tool that helps serendipity happen. Such “serendipity engines” can be location specific, anticipating what you might want to know when you are some place in particular. You might remember the scene in Minority Report when Tom Cruise's character, equipped with surgically replaced eyes, has his (new) retinas scanned on entering a shopping mall. The scan triggers advertising based on the customer profile of the previous owner of the eyes. Eric Schmidt, the CEO of Google told the TechCrunch Disrupt conference last September that the "combination of cloud computing and powerful mobile phones will also enable Google to one day tell people things they may want to know as they are walking down the street, without having to type in any search queries". The catch in all this is that users need to somehow reveal the information about themselves that will allow the engine to correctly target information to them. Mr. Schmidt acknowledged that this will require the consent of users if profile information is to be gathered to power the engine. I note that this sort of I-know-what-you-need search tool is quite different than the sort of "Related Citations" sidebar that pops up on PubMed. It finds papers that share some of the MeSH terms in your search but the engine knows nothing about your interests or needs beyond the present search.
This brings me to the third thing that happened last week. I've been working on a new operations manual and a self-study document for Dalhousie's Interdisciplinary PhD Program. That's meant looking a bit at the notion of what the descriptor "interdisciplinary" means in the context of that program—and I suppose in research generally.
I lucked onto a paper by two UK researchers, Allen Foster and Nigel Ford, entitled "Serendipity and information seeking: an empirical study" [J. Documentation, 59: 321-340, 2003]. Those authors begin by noting the role that serendipity has played in important discoveries in the humanities, in the social sciences, and in the sciences generally. While they accept the general notion of serendipity as arising from "fortuitous accidents", they ask the question (especially in the context of information retrieval) of whether serendipity happens more to some people than to others—and whether it can be helped along. One key to this understanding is to carefully consider what it means to “browse” the literature (what I did with J Thorac Cardiovasc Surg). Foster and Ford suggest that serendipity is a by-product of browsing. They quote J. Rice (Library J, 113: 38-41, 1988) as describing serendipity from a library perspective as:
“…sometimes totally fortuitous but …also, perhaps even more often, actually systematic in a convoluted way that is almost impossible to explain. But reference librarians know that what comes under the rubric of serendipity is often an actual, possibly subliminal, search strategy. Stated differently, the potential for serendipity should be directly related to the number of different access points or potential ways of retrieving from a given system.”
That is, serendipity is related to the number of possible paths by which you might reach the “fortuitous” material. Browsing, serendipity, and access are inter-related and linked to opportunities for the right sort of encounters. Foster and Ford led me to the work of Sandra Erdelez. She identifies a class of searcher/browsers whose members can be called "super-encounterers." According to Erdelez:
“[Super-encounterers] encounter information on a regular basis and perceive it as an important element of their information acquisition. [They] count on information encountering. This habit is, however, not something they are willing to talk about casually, mainly because information encountering does not adhere to the traditionally prescribed methods for finding information. Some super-encounterers are therefore concerned that they may be ridiculed because they rely on bumping into information as a "method" for information acquisition. Super-encounterers not only encounter information important to them, but also often encounter information relevant to others – friends, relatives and colleagues…It is as if information super-encounterers have channels for information perception that are more sensitive than the channels of other information users. This in turn may make them more sensitive to noticing information in their environment.”
It seems to me that the best investigators I have known, especially in the type of interdisciplinary fields where I work, display features of super-encounterers. They have browsing methods that enable them to—somehow systematically—bumble into and identify disparate pieces of information that are linked in some way that is important to them. Some of this skill may very well be innate, a feature of the way they see the world and categorize and catalog information mentally. I would love to be able to foster this sort of browsing in my students. For other types of Erdelez’s encounterers (encounterers, occasional encounterers, non-encounterers), it seems to me that tools like the front page of the New York Times online, and serendipity engines that hold or accumulate information on the searcher’s interests can act as tools to encourage serendipity during information searches—even in a searcher who is not naturally inclined to browse. They increase the number of possible paths to fortuitous material. I am also quite certain that our present, focused search engines for academic use discourage serendipity and inhibit the best browsing habits that we might encourage in senior trainees (PhD students, post-docs) who need to develop new and original approaches for their present and future endeavours. Where are the academic serendipity engines?
"When you work seven days a week, fourteen hours a day, you get lucky." (Louis Pasteur)
—Mike