Cataloging Concepts by Learning Level

OSTI has fielded the beta version of a new science education portal: http://www.scienceeducation.gov [3] or simply SE.gov. Like Science.gov, the SE.gov portal aggregates content from across the Federal government, in this case educational content. But it also has a new, unique feature, in that it sorts content by learning level. My SBIR team is proud to have developed the basis for this feature. It is all a matter of cataloging scientific concepts by learning level.

What does this mean and how did we do it? To begin with, note that simply searching by a word or phrase does not work well for educational content. This is because educational content comes in many different grade levels, from kindergarten through graduate school.

Once a concept is learned, such as “battery,” it is used thereafter. On average the concept “battery” is learned in 4th grade. So searching on “battery” may yield results across a wide range of learning levels, from elementary school to graduate school. But the user always has a specific learning level, so most of these results will be unsuitable. Someone with a college learning level has no need of elementary content, and vice versa.

What is needed is a way to automatically sort content by the user’s learning level. That is what we have done. Basically we have cataloged most of the scientific concepts learned in elementary, secondary, and postsecondary science education. We have also estimated just when, on average, each concept is learned. This is the learning level of the concept.  We have been able to do this because many states have published detailed standards specifying which concepts will be learned in which grades.

Given this catalog of concepts and their learning levels, it is relatively easy for a computer to estimate what the learning level of each search result might be. If the document uses mostly elementary concepts then it is suitable for an elementary learner. If it uses a lot of high school concepts, but few college concepts, then it is suitable for a high school learning level user, and so on.

Mind you this approach is not foolproof. For example, elementary level content may include instructions for teachers, written at a college level. Computers don’t know the difference between content and instructions, so such as document may be scored as college level. Conversely, advanced content (from economics or math for example) may get an elementary ranking, because so far we are only cataloging concepts from the physical sciences. But our simple learning level ranking system is much better than none at all. It really works.

Note by the way that learning levels are not just for students. I have a Ph.D. but my biology learning level is just high school, because I never took a biology course in college. I took lots of physics and a little chemistry, but no biology. For this reason many scientists, engineers, and even science teachers need learning level search just as much as students do. Researchers often find themselves exploring areas they were not trained in, where they have a great need for educational content.

We are also seeing other applications. For example, we frequently find content intended for a given grade that includes concepts that are much more advanced. Such material is unsuitable for many users, and unusable in schools in many states because it is not consistent with state standards. We also find that different content publishers label their content in different ways, so the learning level is not clear when the material is found.

Our catalog of concepts provides a basis for standardized labeling and improved document design. It is, in effect, a controlled vocabulary, where the control is exercised by the state standards. In many states, what is taught and when it is taught is no longer up to the teacher, or the textbooks, it is mandated by state law. Our standards based catalog can be used to design the most useful content.

What is really needed is content that clearly teaches each concept in our catalog, on a concept by concept basis. There is remarkably little of this sort of content around, and what does exist is widely scattered. Science education will benefit greatly if more of this kind of content is created, aggregated and cataloged.

David Wojick