- From: Cory Casanave <cory-c@modeldriven.com>
- Date: Wed, 20 Jan 2010 17:14:31 -0500
- To: "Simon Miles" <simon.miles@kcl.ac.uk>, <public-xg-prov@w3.org>
Dear provenance incubator group, I have been monitoring the group for a few weeks now and would like to offer a couple of use cases into the mix. These are both based from our business and enterprise architecture work in the U.S. government by model driven solutions (modeldriven.com). Use case: Who said that? The scenario is based on a financial architecture being done for a government agency. This is a large architecture involving information, services and processes. Most of the stakeholders are non-technical, many accountants. As with any such architecture it is based on a successive set of inputs and meetings with stakeholders - not all at the same time. While this architecture was not being done with semweb tooling (it was UML), the same situation arises despite the formalism used. Near the end of the project one of the stakeholders was reviewing an information model for orders. This was not the first time this stakeholder had seen this part of the model, but they had not reviewed it in some time. The stakeholder pointed to a property on part of the model dealing with orders and asked: "Where did that come from, who told you to put it in?". Certainly a reasonable question but one we could not answer without a long dig through manual notes. There was nothing in the model to say where that property came from, when it was added or under what authority. In addition the stakeholder noted that something they though was in the model had been removed and wanted to know where it had gone. Again, the tooling could not help. Conclusion: The source (both the person entering the data and who told them to put it their), the situation (such as a meeting) and the time of each assertion in the model needs to be tracked. This should be part of the core knowledge management infrastructure and leads directly to the trustworthiness of the knowledge base as it evolves over time. Use case: Cheating dictator It seems that certain intelligence activities look at things like the college transcripts of interesting people and use these to draw conclusions about their capability and character. The story (and it may just be a story) is that Sadum Housane attended a college in Australia decades ago. The transcripts for that college were obtained and made part of his personal profile. This profile impacted important political and military activities. It became apparent that for propaganda purposes these transcripts had been modified. Analysts wanted to know what inferences had been made by human and automated means, what information was inferred and how that could change Sadum's profile and potential actions. There was no way to trace this information path, making many of the opinions questionable. This is, of course, only one small example in the world where information may be intentionally falsified or obscured and where the resulting conclusions are critically important. The source and down-stream impact of information is critical, particularly when sources and information quality are re-evaluated. Conclusion: The track of inferences may span decades and this track may be of critical strategic value. In addition, inference is a combination of human and automated activities that effect down-stream conclusions. ------------ I hope you find these use cases of interest. Regards, Cory Casanave Model Driven Solutions
Received on Wednesday, 20 January 2010 22:14:55 UTC