The Data Briefing: Better Crowdsourced Federal Government Projects Through Cognitive Computing

Cognitive computing has been receiving a good deal of attention lately as more companies have been building intelligent agents. Ever since IBM Watson’s 2011 appearance on Jeopardy, cognitive computing has spread into healthcare, investing and even veterinary medicine. However, it is only recently that cognitive computing has spread into government applications.

As the name implies, cognitive computing is where computers operate much like the way people think. Computers use data mining techniques, pattern recognition algorithms and natural language processing to search a large set of unstructured data to find solutions. Cognitive computers excel in making connections between seemingly unrelated data and assisting people in searching large collections of information. Cognitive computers are also good at solving process problems when the cognitive computer software is paired with Internet of Things sensors.

Another good application of cognitive computing is pairing it with crowdsourcing projects. Two recent research papers describe how cognitive computing can improve the management of crowdsourcing projects and improve the results developed by crowdsourcing participants. As more federal agencies harness the power of crowdsourcing, pairing cognitive computing with chatbots can better manage the entire crowdsourcing process.

The first research paper describes how to use cognitive computing to help prevent a common problem in crowdsourced projects. Crowdsourced projects work by slicing up a large project into many small microtasks. This prevents the participants from seeing more than just their slice of the project. For many crowdsourcing projects, there have to be moderators who can see the big picture. Wikipedia is a good example of this problem, where many participants contribute edits to parts of articles or even entire articles. However, there is a small cadre of editors who coordinate the contributions over the entire Wikipedia site. Thus, the bottleneck problem arises because the number of moderators is rarely enough to handle the greater number of contributors.

Cognitive computing can take the place of human moderators. This is done by distributing the sensemaking process to the crowdsourcing participants. For example, the cognitive computing application randomly selects articles for participants to sort into a category. The sorting results are then aggregated into a single sorting result. This sorting exercise can be repeated until consistent results are established. The purpose is to help the participants arrive at a big picture view by assembling their specific views of the larger project. The advantage of using cognitive computing instead of human moderators is that it removes the bottleneck problem.

The second research paper deals with a second problem caused by the inability of crowdsourcing participants to see the big picture. Imagine that the crowdsourcing project is to sort a large box of blocks into categories. Each participant is given a small pile of blocks to categorize. The issue is that some participants may receive piles that contain too many blocks of the same type. Other participants may receive a pile of blocks that contain too much variety. The point is that participants may create categories that are not truly representative of the entire box. This is because the participants are limited to their small slice of reality.

The cognitive computing application solves this problem by giving the participants a view of the entire box of blocks through a “sample and search” process. This way, each participant has a more realistic view of the entire box. The categories that are created are classified with a “high-confidence” rating and “low-confidence” rating based on aggregating all of the participants’ work. This helps to create categories that are truer representations of actual categories in the big picture reality of the crowdsourcing project. Again, the problem of not seeing the big picture is solved by increasing the probability that the small slice of reality is more representative of the bigger picture.

DigitalGov has some great examples of crowdsourcing helping federal government projects. As more and more agencies turn to crowdsourcing, cognitive computing can help better manage the projects and produce better results. Even among federal employees, pairing cognitive computing with crowdsourcing can greatly improve internal agency processes. I look forward to the innovations produced by harnessing crowdsourcing through chatbots powered by cognitive computing applications.Each week, The Data Briefing showcases the latest federal data news and trends.
Dr. William Brantley is the Training Administrator for the U.S. Patent and Trademark Office’s Global Intellectual Property Academy. You can find out more about his personal work in open data, analytics, and related topics at BillBrantley.com. All opinions are his own and do not reflect the opinions of the USPTO or GSA.