Last week, at the SharePoint conference (and also the CEO Summit), Microsoft announced a new technology – Knowledge Network (KM) for Microsoft Office SharePoint 2007 (MOSS). What follows is an overview based on the session I attended (the presenter clearly stated we could blog at will 🙂 ) with some personal comments added in. I have not installed the product and do not have access to it. Knowledge Network is currently a closed beta, so you won’t find it on the list when MOSS beta 2 is released. I played with a very early prototype of the client-side technology 2 or 3 years ago, when it was still in MS Research and I was still at Microsoft, but the product has changed significantly since then.
KN is focused on enterprise social networking, automatic discovery and sharing of undocumented knowledge and relationships. (I copied that off the opening slide…) So what does that really mean?
The solution involves client and server elements. The KN client is installed locally and analyses email content to create an individual profile of keywords and contacts (colleagues and external contacts). The user can review the profile – for example, set privacy levels on information (e.g. choosing only to share external contacts with your direct team) and remove information you don’t want published. The user then chooses to publish the profile and it is uploaded to the KN server, i.e. this is an opt-in model (until some evil being in I.T. enforces publishing through group policy…). As multiple user profiles are published to the KN server, they are aggregated to create expertise information and form a social network (i.e. the more profiles published, the richer the network). The MOSS search service indexes the information created by the aggregated profiles and it is returned within search results.
When a person (seeker) queries for people – who knows what/whom – the results are ranked by social distance to the seeker, expertise and relationship relevance (e.g.results grouped as ‘my colleagues’, ‘know my colleagues’… <– this is similar to the LinkedIn method of being able to link to people who are linked to people you know.) The KN server also includes a feature called ‘anonymous brokering’ – it is based on the privacy field in the KN profile manager and allows people to share information with the system, but only on demand.
If you are seeking expertise, you can submit a query and will be returned a set of people results without the contact information. If you want to contact one of the experts, you click on the link to send a message via the KN broker. The KN broker forwards the email to the expert (under its own email account) and the expert can choose to accept or decline the request. If the request is accepted, the seeker and expert are hooked up.
This feature demo’d well, but I suspect the technical implementation will be far easier than the cultural implementation. In smaller organisations, it will not be difficult to guess who the expert is even without the contact information. The culture of the organisation must make it acceptable to say no to requests without fear of penalty, otherwise everyone will just say yes (or not publish their expertise) and the service becomes irrelevant. There are some configuration options, such as how many times an individual can contacted with requests during a time period – useful, but again needs to be within a culture that allows experts to say no.
The individual profile is created client-side with no server involvement. In effect, the profile is an index of the content within Outlook (keywords and contact information are extracted from emails within Outlook folders and also contact lists in IM (I’m assuming that means Live Communication Server)). After the initial profile is published, incremental updates are sent at an interval defined in the configuration. The default is 14 days. <- this is a concern. Whilst people will likely be thorough in reviewing their profile prior to the initial publish, I suspect the novelty will soon wear off and they will start to accept the defaults for incremental updates. This could lead to sensitive information being published onto SharePoint without the source user realising.
KN is designed for Microsoft environments – it requires MOSS to install (Windows Server 2003, .NET framework 2.0, SQL 2000 or 2005) and requires Active Directory and Exchange for name resolution (contact information and DLs). The client will need to be running Windows XP with Service Pack 2 or later and Office 2003 or later. The product group are already recommending that Outlook be configured with cached Exchange mode to minimise processing impact against the Exchange Server whilst inboxes are analysed to create/update the profile.
From a deployment perspective, KN is another shared service being added to MOSS (joining Excel services, Forms server, and Indexing/Search). As soon as a user elects to publish their profile to the server, the KN profile management web service takes over to calculate the expertise information and social network.
The session closed with a healthy Q&A that raised some interesting issues. A couple of specifics: Only the body of emails are indexed, not the attachments. The question was asked as to whether or not the ‘Deleted items’ folder was indexed, and I didn’t hear a clear answer. It poses an interesting challenge – how do you determine which emails contain relevant information. I delete irrelevant stuff immediately, but I delete everything eventually unless it has particular sentimental value. And that leads on to the age old challenge of auto-generating social networks based on emails we send/receive and searches we perform – how to determine when expertise is being shared versus discovery and learning versus spam (corporate as well as external) versus answering on behalf of etc. The product group are more than aware of this challenge. When asked why not mine stuff other than email (documents, IM conversations etc.) they responded that email is the richest in terms of tacit information as well as being the most pervasive source. They acknowledged that the challenge in calculating strength (relevance) was hard enough and adding data sources adds complexity, and decided the return was not worth the investment in this version (i.e. look out for extensions in the future…)
Historically, organisations have been reluctant to deploy social software tools – IM being the most recent example of irrational fears over-riding business benefits (see related post: when will IM come of age). Knowledge Network will face similar challenges, as concerns over productivity drains, privacy and culture-fit bubble to the surface. That all said, I’m glad Microsoft has finally entered this space. The power of social networks have become well documented over the past 5 years, and failure to understand them is one of the primary reasons why most KM systems fail. This will be a v1 technology and will have all sorts of flaws and challenges. But it’s a great start and this sort of capability is long overdue.
For more information:
- Microsoft product team blog for Knowledge Networks
- Craig Randall has posted his thoughts (he attended both sessions at the SharePoint conference, I just went to the session covering the details)
If you’re interested in learning more about the potential value of social networks, there are plenty of books on the subject but here are three I would recommend for starters:
- The hidden power of social networks by Rob Cross and Andrew Parker (2004)
- Linked by Albert-Laszlo Barabasi (2003)
- Six Degrees by Duncan Watts (2003)
Update: An overview has been posted up on Microsoft’s web site: http://www.microsoft.com/office/preview/servers/sharepointserver/kn.mspx complete with screenshots.
Thanks for posting such a detailed perspective on KN. The answer to the question about whether we analyze Deleted Items is "Yes". We will analyze any folder that the user wishes.Glen AndersonGPM Knowledge Network