Filtering in Social Software: Protective Bubble v. Serendipitous Awareness

There was an interesting conversation on Twitter yesterday about the personalization of information via algorithm-based filters. It was started by Megan Murray, and Thomas Vander Wal, Gordon Ross, and Susan Scrupski quickly joined in with their viewpoints. Rachel Happe and I were late to the conversation, but we were able to interact with some of the original participants.

.The gist of the conversation was that some consumer social services (i.e. Facebook, Google Search, Yahoo News) have gotten rather aggressive about applying algorithms to narrow what we see in our personal activity streams. As a result, we aren’t able to see other information that might be useful or entertaining in our default view; we may only digest what the algorithm “thinks” is important or relevant to us. Or we must switch to a different view to see additional information (e.g. Live Feed v. News Feed in Facebook). Even worse, in some cases, the other information is simply not available to us, because the service doesn’t provide a way to override the algorithm that excluded it.

It was also noted in the Twitter conversation that the current crop of enterprise social software lacks sophisticated personalization facilities. In fact, it works the opposite way of consumer social services; the entire activity stream is usually exposed to an individual, who then has to narrow it by manually selecting and applying pre-defined filters. IBM, Jive, NewsGator, and others are beginning to use algorithms to include certain status events and updates in the stream, and to exclude others, but their efforts will require fine tuning after organizations have experimented with these nascent (or yet-to-be released) personalization features.

The default view of an enterprise activity stream should be highly personalized to the context in which an individual is working (e.g. role, business process, location, time, etc.) Optional views should allow individuals to override the algorithmically chosen results and see information relevant to a specific parameter (e.g. person, group, application, task, tag, etc.) Finally, an individual should be able to view the entire stream, if he or she so desires.

Why is the latter important? It introduces serendipity into the mix. Highly personalized information views can increase productivity for an individual as they do their job, but at the expense of awareness of what else is occurring around them (I wrote about this earlier this week, in this post.) This condition of overly-personalized information presentation has been called a “filter bubble”. The bubble is a virtual, protective barrier against information overload that is analogous to a plastic enclosure used in hospitals to shield highly vulnerable patients from potential infections.

Organizations must consciously balance the need to protect (and maximize the productivity of) their constituents from information overload with the desire to encourage and increase innovation (through serendipitous connection of individuals, their knowledge and ideas, and information they produce and consume.) That balance point is different for every organization and every individual who works in or with it.

Enterprise social software must be designed to accommodate the varying needs of organizations with respect to the productivity versus awareness issue. Personalization algorithms should be easily tunable, so an organization can configure an appropriate level of personalization (for example, InMagic’s core Presto technology features a “Social Volume Knob” that allows an an administrator to control what and how content is affected by social media. Different kinds of social content from certain people can carry different weight or influence.) More discrete, granular filters should be built into social software so individuals can customize their activity stream view on the fly (I made that case, just over a year ago, in this post.) A contextually personalized view should be the default, but enterprise social software must be designed so individuals can quickly and easily switch to a different (highly specific or broader) view of organizational activity.

What do you think? Should personalization be the default, or applied only when desired? What specific filters would you like to see in enterprise social software that aren’t currently available? What role does/could portal technology play in the personalization of organizational information and activity flows? What other concerns do you have about information overload, filter bubbles, and missed opportunities for serendipity and innovation? Please weigh in with a comment below.

This entry was cross-posted from Meanders: The Dow Brook Blog

Image © 2003 Texas Children’s Hospital

LinkedIn Signal Demonstrates The Power of Role-Based Activity Stream Filters

LinkedIn today announced Signal, a new feature (currently in beta) that lets members see an activity stream that combines LinkedIn status updates and Twitter posts from other members who have opted-in to the feature. LinkedIn has licensed the Twitter firehose to incorporate all of its members’ tweets into the site, not just tweets with the #in hashtag embedded, as is current practice.

While it is hard to imagine anyone other than corporate and independent talent recruiters will make LinkedIn their primary Twitter client, Signal does have an element that is worthy of emulation by other social networks and enterprise social software providers that incorporate an activity stream (and which of those does not these days!) That feature is role-specific filters.

I wrote previously in this post about the importance of providing filters with which individuals can narrow their activity stream. I also noted that the key is to understand which filters are needed by which roles in an organization. LinkedIn apparently gets this, judging by the screenshot pictured below.

LinkedIn Signal screenshot courtesty of TechCrunch

Notice the left-hand column, labeled “Filter by”. LinkedIn has most likely researched a sample of its members to determine which filters would be most useful to them. Given that recruiters are the most frequent users of LinkedIn, the set of filters displayed in the screenshot makes sense. They allow recruiters to see tweets and LinkedIn status updates pertaining to LinkedIn members in specific industries, companies, and geographic regions. Additionally, the Signal stream can be filtered by strength of connection in the LinkedIn network and by post date.

The activity stream of every enterprise social software suite (ESS) should offer such role-based filters, instead of the generic ones they currently employ. Typical ESS filtering parameters include individuals, groups or communities, and workspaces. Some vendors offer the ability to filter by status as a collaborator on an object, such as a specific document or sales opportunity. A few ESS providers allow individuals to create custom filters for their activity stream. While all of these filters are helpful, they do not go far enough in helping individuals narrow the activity stream to view updates needed in a specific work context.

The next logical step will be to create standard sets of role-based filters that can be further customized by the individuals using them. Just as LinkedIn has created a filter set that is useful to recruiters, ESS providers and deploying organizations must work together to create valuable filter sets for employees performing specific jobs and tasks. Doing so will result in increased productivity from, and effectiveness of, any organization’s greatest asset – it’s people.

Enterprise Social Software and Portals: A Brief Comparison of Deployment Patterns

In my last post, I examined whether or not Enterprise Social Software (ESS) is the functional equivalent of enterprise portal applications as they existed ten years ago. My conclusion was:

From a functional perspective, ESS is quite similar to enterprise portal software in the way that it presents information, but that does not tell the whole story. ESS lacks critical personalization capabilities, but provides better collaboration, process, publication and distribution, categorization, and integration functionality than portals. In my judgment, ESS is somewhat similar to portal software, but mainly in appearance. It makes more functionality available than portals did, but needs to add a key missing piece – personalization.

In this post, I will focus on the observation that ESS resembles enterprise portals in another regard – how and why it is deployed.

Enterprise v. Smaller Deployments

Portals were initially marketed as a tool for enterprise-wide communication and interaction, with each internal or external user role having its own personalized set of resources available in the user interface. While there were some early enterprise-wide deployments, portal software was deployed far more often at the functional level to support specific business processes (e.g. sales, procurement, and research portals) or at the departmental level to support operations.

Enterprise social software has also been touted as most valuable when deployed across an organization. However, like portal software, ESS has most often been deployed at the functional level in support of activities such as marketing, customer service, and competitive intelligence. As a result, the promised network effects of enterprise-wide deployments have not been realized to-date, just as they were not with most portal deployments.

Internal- v. External-Facing Deployments

Most early portal deployments were internally-focused, as shown in this InformationWeek summary of market research conducted in 2001. Not only was there a smaller number of externally-focused deployments, mixed-audience deployments did not begin to appear until the portal market was extremely mature. ESS deployments have followed this same pattern, and we are just now seeing early efforts to blend inward- and outward-facing business activity in common ESS environments.

Internal Use Cases

Portal software was often deployed in response to a specific business need. Among the most common were:

• intranet replacement/updgrade
• self-service HR
• application aggregation
• document/content management
• expertise location
• knowledge sharing
• executive dashboards

ESS has been deployed for many of the same reasons, especially intranet replacement, application aggregation, expertise location, and knowledge sharing.

External Use Cases

Portal software was deployed externally to provide self-service access to corporate information. In some cases, access to selected application functionality was also provided to key business partners. Retail and B2B portals enabled customers to purchase goods and services online. Process acceleration, revenue growth, and cost reduction were the key business drivers behind nearly all external portal uses.

ESS doesn’t seem to have the same goals. I have seen some, but little, evidence that external communities are being leveraged to accelerate business processes or reduce costs. Peer support communities are a good example of cost reduction via ESS. The goal of most outward-facing ESS deployments seems to be customer engagement that translates (eventually) into increased innovation and revenue for the deploying organization.

Conclusions

ESS deployments today strongly resemble portal projects that were undertaken ten years ago. Few, if any, ESS deployments have been enterprise-wide. Instead, ESS is deployed to many of the same department and functional groups, to support the same business processes, and to drive many of the same business results as portals were a decade ago (and still are.)

What does this commonality with early portal deployments mean for ESS? I will examine that in my next post. Until then, I would love to hear your reaction to what I have presented here.