‘Role of automated systems, humans to be defined’

‘Role of automated systems, humans to be defined’


San Antonio, US: Advances in processing power and storage made it possible to relieve the burden on analysts. However, much work needs to be done. The challenge is greater than simply combining different intelligence products in a timely fashion to yield something that users find worthwhile and easily understandable, observed Multi-INT panellists during pre-symposium science and technology forum held at the GEOINT 2011 in San Antonio, US. They concluded that the analysts’ job lies in correctly setting the dividing line between automated systems and humans.

Dave Messinger, an associate research professor at the Center for Imaging Science at the Rochester Institute of Technology, noted that computers don’t get tired and are good at sorting through large sets of data. Where they stumble is when called upon to do sophisticated analysis. Fortunately, though, there’s another readily available system for this task. “The brain is good at complex analysis,” Messinger said. Ideally, he said, this analysis would be done using a 4D physically realistic model of the world, accurate in both space and time on an appropriate scale. Then an analyst could test a hypothesis against the model and compare it to actual data, thereby weeding out those theories that don’t fit reality and measurements. Messinger believed this capability will soon be available, and it will offer a near real time response.

In putting automated systems to use, care must be exercised, said Naval Postgraduate School remote sensing research professor Charlene Sailer. “We need to convey something about the data quality.” That subject came up several times during the presentations. Sometimes it involved data analysis and other times data integrity. For the latter, one suggestion was the use of the lowest level bit in a tamper resistance scheme. For the issue of the quality of data analysis, a proposed solution involved thresholding. More certain data would be depicted, for example, with higher intensity. But thresholding, it was pointed out, is generally problematic. It may turn out later that measurements that fell below a threshold are actually useful. For instance, a Soviet-era missile detection system was actually being able to spot Scud missiles during the first Gulf War, if attention was paid to what before had been regarded as noise. The solution, everyone agreed, was to never discard any of the rawdata. A corollary to that is to measure everything.