Certifying Knowledge: A Business Idea

I have written previously about things that I feel need to happen in education reform and how there could be a dramatic change in higher education that some refer to as the "education bubble". There is one big thing that I see being needed to really trigger that disruption. It is something that is clearly doable with current technology, but it hasn't been put into place yet. I feel like someone who wrote this and put it up quickly could make a lot of money. The piece that is missing is certification of knowledge with validation of the person who has the knowledge.

There are lots of sites out there now which measure various capabilities/skills. TopCoder is one of the first that jumps to my mind. They have a number of different competition based metrics to indicate how fluent a programmer is at different tasks. With all the MOOCs that are coming on line and their automatic evaluation, the ranks of these are growing quickly. However, one thing to note about the biggest MOOCs, like Cousera, is that they have a somewhat limited form of recognition of completion. Some of this is because the schools posting courses there need to still have a business model. Probably more of it though is because certifying a person would require that you really know who they are.

Universities go through a fair bit of work to verify a person's identity. At small schools that is implicit in the overhead of having small classrooms. I know all my students by name on sight. If someone else were to show up on a test day I would know something was wrong. Larger schools have other mechanisms to make certain that the person demonstrating knowledge is the same person who is getting the grade and the credit. This doesn't happen online yet, but it can, it needs to, and inevitably it will. It is just a matter of who will build the system. The rest of this blog outlines my idea for how this would be built and how it could go into a useful product.

Monitoring the Tester
The key, in my opinion, is the ubiquitous spread of webcams. Having things like biometric sensors can identify that the person is present, but a webcam and microphone makes it possible to further show that person is the one doing the work which is being evaluated. The webcam can do the same type of job as other biometrics to make certain that the person you want is the one at the computer. Facial recognition, voice recognition, potentially even information on the person's iris if they get close enough or the cam has enough resolution. In fact, the more forms of identity you have, the better it is. The site should mark each test/result with which verification techniques were used.

After the initial verification, the webcam and microphone can be used to monitor the person taking the test. Eye tracking can see where he/she is looking on the screen. Smaller body movements can show that the person seen in the image is actually the one clicking on certain answers at certain times or doing the typing. The microphone can monitor what is being said in the room to make certain that the person is not being told the answers by someone else in the room. One might even consider having 360 degree webcams to monitor the whole room.

The goal here is to have a system that can be run on a large fraction of the devices someone might want to use to do work that demonstrates their knowledge of a topic, and for that system to be roughly as accurate at preventing outside assistance as a human would be. Of course, it needs to be automated so that it can scale well. You can't be paying hundreds of people to watch other people taking their tests through a webcam. That loses the efficiency.

Certification Clearing House
To make this more valuable, you have to put it together with many forms of verification of knowledge. Many things will be done with fairly simple tests, but the ideal would be that you could have a site that not only does the verification, but which can tie in with other sites so that you might be able to do something like verify the identity of a programmer competing for TopCoder without having to rewrite TopCoder.

It would also be nice if employers could submit their own types of tests. So this could go beyond basic academic areas. The goal would be to have a significant database of areas of competency/mastery and the ability to link up employers and potential employees. In an ideal world, you would eventually use the type of oral exam method I described earlier that would use a Watson style area expert to run the tests. In addition, it could be tied in with educations tools to help teachers/learning coaches to see where people are having problems to get them over things. This type of system could literally be used for demonstrating the capabilities of people going from low grade levels up through graduate level study.

Business Model and Added Benefits
If you were to create a company on this model, it might be possible to collect money from both sides. There is clear value to individuals to have a certification of what they know and what level of mastery they have demonstrated in that knowledge. In addition, there is benefit to entities that want to verify that someone knows something.

One of the interesting perks that I see in this type of approach is that employers could actually build a description of skills for various positions. Some could be hard limits. Some might be desired attributes. Still others might be things where you need to have demonstrated a combined ability in several areas with the exact distribution being fairly insignificant. HR in this world gets a lot easier on the hiring end, and probably gets downsized a bit because there isn't nearly as much use for resumes.

In addition, if you want a job, you should be able to see the metrics in some form to see whether you qualify. I see this as a great motivator for kids in school. They could actually see what types of skills are valued for specific types and levels of jobs. They could see where there are going to be ceilings that come into play because they lack depth of a certain type of knowledge.

This goes back the other way too. Employers can do an analysis of real performance of employees to their measured skill sets and refine what types of skills really lead to better performance in a particular job. As long as enough of the information is open, this winds up being a generative system. People will find all types of new uses for the data and new meaning that can be pulled out of it.