Search remains the go-to information access interface and boasts a growing business intelligence footprint. But as far as I'm concerned, our reliance on search needs to end.
We all understand search's success. The search box is an easy path to rich, diverse information sources and a workaround for poor systems design. We type or speak a few words and instantly get, not just conventional search-engine results pages (SERPs), but also graphical data presentations and categorized, linked-data-enriched results.
Search has grown powerful. Unified information access extends search to fused structured and unstructured sources. Text analytics enables semantic search, and the question-answering paradigm refocuses searches on to-the-point results. These are great advances, yet they and other advances (particularly in mobile devices) may soon mean the disappearance of search as we know it. I hope they will.
Sixty years in, many searches and far too many search-driven data analyses make little sense. Searching diverts you from your business (think customer service/support, market research, competitive intelligence, and the like). And too often, the results -- SERPs, dashboards, pie charts, and data tables -- don't contain or convey the to-the-point answers you're seeking. Simply put, most searches and search-driven analyses make for ineffective decision support.
A new type of information-processing focus -- one based on signals and sensemaking -- is coming to the rescue.
I've been thinking about this stuff for several years, inspired by work that dates back beyond search. Peter Pirolli and Daniel M. Russell of PARC and Google, respectively, trace the sensemaking notion to the information scientist Vannevar Bush's work in the 1940s to develop a "mechanized private file and library" for individuals. Bush's memex sought, among other things, to "provide a more efficient means of finding new patterns and insights and sharing those discoveries with others," Pirolli and Russell wrote in an article for Human-Computer Interaction.
I gave a presentation recently on sensemaking at the Open Source Search conference. It's a theme of talks that I have included in my Sentiment Analysis Symposium, which is scheduled for Oct. 30. For example, V.S. Subrahmanian, a computer science professor at the University of Maryland, will look at emotional signals and the diffusion of sentiment in his presentation, "Sentiment and Signals."
We are immersed in a cloud of personal, social, and enterprise data sources. Call it big-data, or don't if you're sick of that term; just recognize the astuteness of social media theorist Clay Shirky's observation: "It's not information overload, it's filter failure."
The first sensemaking step is to distill a stream of essential data from the mass of sources. And here's a first distinction: Search produces point-in-time results, while signals factor in change over time, taking historical observations and ongoing data capture into account. Data may originate from servers, sensors, monitoring, or alerts, and it is selected, prepared, combined, and presented to respond to situational business needs.
Stock traders, public health workers, and car buyers look to online, social, enterprise, and personal information sources (often a mix of them) for signals about market conditions, disease outbreaks, and brand and deal information. One concept is being put to disparate applications. Think of the sensory elements and computation and control systems of a driverless car. This is a great example of machine sensemaking applied to an everyday activity.
"Situational" reflects what the searcher is doing, or would be doing if alerted, at a particular time and in a particular business context. Result presentation (for human consumption) should suit the searchers' situation. Just avoid cookie-cutter dashboards. They have become the tail that wags the BI dog, embodying the notion that an array of graphical widgets will magically deliver insights that promote business decision making. The too frequent result is a decision gap -- a disconnect. One-size-fits-all charts and gauges measure but don't guide.
Further, they aggregate disparate information without properly contextualizing it, much less fusing it. And even when they do pull from the social graph and free-text social, online, and enterprise content (especially subjective content such as sentiment and intent), they shoehorn differently structured information into displays most suited for transaction-derived data. Give me to-the-point facts or answers, augmented-reality data rendering, or business and personal systems that meld analytical outputs into the familiar working interface.
So we assemble and evaluate signals and derive quantities that can drive decisions. Professor Marti Hearst of the University of California writes in her 2009 book, Search User Interfaces, "Sensemaking refers to an iterative process of formulating a conceptual representation... of a large volume of information. Search plays only one part in this process." It also includes analysis and synthesis of the results. Pirolli and Russell write, "Sensemaking involves not only finding information but also requires learning about new domains, solving ill-structured problems, acquiring situation awareness, and participating in social exchanges of knowledge."
Social makes a special contribution. In social contexts, which include just about every modern-day activity outside sleeping and reading a book, systems look to recommendations, Likes, behaviors, and experiences from our friends, our networks, and the crowd. They computationally relate to us based on identity, location, profile, behavior, and social network, even by the device involved. (Your travels accompanied by a mobile device, your clickstream, search history, and transactions -- everything is tracked.) Sentiment and other forms of subjectivity are a huge recommendation contributor. The machines seek to do the sensemaking for us with socially and situationally infused relevance measures.
Analytics-assisted sensemaking is a work in progress. It is a future direction for search development and for the emerging world of intelligent(-seeming) machines. Search is an information access starting point, but beyond search, only analytics can systematically identify and use signals to help us evolve from search and subjectivity to sensemaking. It's going to be an exciting journey.