Online Extra: Full-length interview with Alex Bennet
Turning data into knowledge is the challenge
- By Patience Wait
- Jan 09, 2006
Alex Bennet, co-founder of the Mountain Quest Institute in Frost, W.Va., is a recognized expert in knowledge management. She was chief knowledge officer and deputy CIO for the Navy Department, and led the Federal Knowledge Management Working Group. Bennet is the recipient of the Distinguished Public Service Award, the Presidential Hammer Award, and the National Knowledge and Intellectual Property Management Award for distinguished service and exemplary leadership. She has written extensively on KM and is co-author of Organizational Survival in the New World: The Intelligent Complex Adaptive System
.GCN: Why has data become so important? Bennet:
Data has always
been important! But as we head towards the end of this decade when the computational power of supercomputers is expected to equal the computational power of the human brain'what [Ray] Kurzweil calls the singularity point'we are focusing less on technology and more on the availability of and access to the data, information and knowledge enabled by technology.
Actually, I might just as easily have used a capital 'A' on 'access,' meaning that not only are we interested in being able to get the best data and information, but we are interested in understanding it and knowing when, where and how to best apply it! You will note that the 'knowledge' word has become more and more prevalent as global access and connectivity have increased.
There are many interpretations of the terms data, information and knowledge. For purposes of this discussion, data is taken to be sequences of numbers and letters, spoken words, pictures, even physical objects when presented without context
. Information is data with some level of context. Information can be stored in computers and sent over communications lines. Knowledge is built on information, connotes meaning, heavily dependent on context, and created within the individual. It represents understanding of situations and their context, insights into the relationships within a system, and the ability to identify leverage points and weaknesses and to understand future implications of actions taken to resolve problems. In brief, knowledge is the human capacity (both potential and actual) to take effective action in varied and uncertain situations
. So it is knowledge that drives effective decision-making.GCN: Has technology matured enough to support data sharing as a fundamental system requirement? Bennet:
Certainly! It has matured enough to support information sharing
as a fundamental system requirement, and to provide the meta-data and meta-information to ensure that data and information can support knowledge growth in decision-makers. In other words, the progress we are making is finally beginning to focus on the user and usability.GCN: Are hardware and software sufficiently 'commoditized' that they have become less important? Bennet:
Okay. Let's stop the dualist nature of our thinking. Perhaps the 'focus' is changing, but the reason that focus is changing is because we're doing many things right in terms of our technology advances, and now we need to ensure that we as decision-makers profit from those advances! As long as we as a race don't destroy our technological advances, hardware and software'and technology
'will remain the underpinning of our global world. Just to set the record straight, take a look at a good Webster's dictionary and you'll discover that one of the ways technology is defined is as the sum of all the ways in which a social group (organization, country, world) provide themselves with the material objects of their civilization
. So even as we move away from competition and embrace an economy of abundance'with data, information and knowledge leading that abundance'hardware and software in some form will underpin that economy.GCN: Does the idea of 'build once/use many' apply to data? In what ways? Bennet:
Not that easy. Data, information and even knowledge are context sensitive. There's no inherent
goodness and badness, so to speak. The goodness and badness comes with the context
and how it is used in a specific situation
. For example, we could say that the terrorists that attacked the World Trade Center of 9/11 used good knowledge management in the sense that they had enough data and information and used it to succeed at some level in what they set out to do. Now, that same data might prove pretty useless for other purposes.
So data, information and knowledge become 'artifacts' and even noise, perhaps effectively used one or more times but not necessarily important or significant in different situations or as time goes by, which pretty much negates the 'build once/use many' concept!
The good news is that we're beginning to become more aware of patterns. That means that the 'type' of data used for one decision might be of use to future decisions, so we design in live feeds to authoritative data sources. Or the same data itself might be useful across different applications. Or it might mean that the patterns from data we've 'built once' in a system might provide clues for addressing future emergent issues and problems. And, of course, our advances in AI technologies helps us find those patterns.
So the 'build once/use many' term really applies to systems and patterns of data, not necessarily to the specific data itself. That being said, obviously there are some data sets that have been and will continue to be around for some time (until we discover the next best thing)! For example, it took us more than two centuries to move from Newtonian mechanics to quantum mechanics.GCN: How do you see this [attention to data] reflected in agencies' priorities? What are the big projects/programs that embody this change? How will the Data Reference Model facilitate the use and sharing of data? Where do issues of privacy fit in? Bennet:
For all the reasons above, data (and how it is used and needs to be used, i.e., information and knowledge) is the focus of attention now, whether it's the DODwide push for Web Services that rely on authoritative data sources, or the new DOD information assurance architecture which focuses on the future vision of eliminating traditional network boundaries (like Unclassified (NIPRNet) and Secret (SIPRENet)) in favor of being able to put security around objects (like data elements, Word documents, etc.). With this approach, data would be available to people based on a matching of meta data with attributes/roles of the individual, so you can see information on the network that you have a need for, but not information that isn't needed.
The result is a great emphasis on getting data tagged correctly and doing identity management right, so that:
GCN: Who should control data'and who should verify it's good data? What can be done about 'dirty data'? Bennet:
- I can see my personal record but not yours
- I can gain access to information that I need based on my security clearance, role as a CIO, etc., but other folks with different attributes/roles would not see the data as they navigated the network and
- You can separate where the data comes from and how it is used. As an example, the Marine in Iraq just wants trustworthy information, and may not need to know the highly classified source from which that data originated.
Ultimately the responsibility for data resides initially with the producer of that data and the responsibility for how it is used resides with the user.
A couple of years back the [Department of the Navy] started talking about the differences between clustering and clumping data. Clustering is when you bring data and information together that is similar or related, i.e., first- and second-cousin organization. This process of categorization by similarities is the popular approach to the organization of data and information. It supports ease of locating specific data, and can lead to innovation and insights. Clumping is organizing data driven by the decisions that will be made on that data. At the top levels of decision-making, this means identifying authoritative data fields you need from disparate locations, then linking directly to those fields for continuous real-time feed to support emerging decision-making requirements.
An example of application at the tactical level would be a database field providing information that the owners of that information realized was insufficient for the best decision, i.e., the suppliers of the information recognizing the need for additional information and supplying that additional information connected to the original information, even though the decision-maker may not know such information exists
Verifying 'good' data becomes problematic, since, as discussed above, 'good' is context and situation dependent, somewhat like 'beauty is in the eyes of the beholder.' So, as emphasized above, ultimately the responsibility for data resides initially with the producer of that data and the responsibility for how it is used resides with the user. Thus the 'beauty' of live feeds!