Google has introduced a new service to facilitate knowledge sharing. Google describes it this way:
“The Knol project is a site that hosts many knols — units of knowledge — written about various subjects. The authors of the knols can take credit for their writing, provide credentials, and elicit peer reviews and comments. Users can provide feedback, comments, and related information. So the Knol project is a platform for sharing information, with multiple cues that help you evaluate the quality and veracity of information.”
Knols are indexed by the big search engines, of course. And well-written knols become popular the same as regular web pages. The Knol site allows anyone to write and manage knols through a browser on any computer.”
It seems strange to claim that a blog post or an article is “a unit of knowledge.” This is true first, because an article is a container composed of many statements, assertions, or knowledge claims. But, second even if we interpret an article as referring to its content, the set of abstract objects (propositions, arguments, problems, theories, etc. but never “concepts”) asserted or expressed by the article, then a few things seem clear.
— A given article may assert no knowledge at all, but only known falsehoods;
— A given article may contain any number assertions that, as far we know, may be true and that therefore are knowledge: and
— Assertions that are knowledge will differ in both their logical and empirical content. Some assertions will be relatively vacuous and will be nearly empty of content, making them very difficult to refute. Other assertions will be riskier and prima facie should be much easier to test, and, in principle, to refute. Both assertions may be knowledge at any particular point in time. Which has more “units of knowledge?”
Third, the above comments and Google’s idea, as well, assume that expressed linguistic content can be knowledge, i.e. that though knowledge can only be created by humans, once created, it can exist as “knowledge without a knower.” However, for many people, there is only one sort of knowledge, and that is “knowledge in the mind.” If one believes that, then Google’s idea is way off the mark, because “knols” would have to be viewed as “units of mental belief,” and perhaps even as “units of mental justified true belief.” Of course, such units are not directly accessible through articles, and therefore such articles can’t be units of knowledge.
Fourth, even though the idea of a concrete unit of knowledge, in the form of an article, doesn’t hold up, a more measurable idea is the amount of knowledge expressed in an article. So Google could set up a system to compare articles in terms of the “knol score” of what they express, where the knol score is defined as the value of an article on the knol ratio scale. Measuring the knol score involves an easy application of Thomas L. Saaty’s Analytic Hierarchy Process (AHP), and the theory and methodology behind such an application have been tested and applied for roughly 35 years now in thousands of studies.
To create ratio-scaled knol scores, all Google would have to do would be to use the following procedure.
Step 1: When people register for the knol site, and using a graphical method of comparison splitting a pie, get their pairwise comparison ratings of the categories with respect to the amount of knowledge they connote relative to one another other: Greatest Amount of Knowledge possible in a single article; Great Amount of Knowledge, Good Amount of Knowledge; Fair Amount of Knowledge; Some Knowledge; Least Amount of Knowledge. This step will require [(n)(n-1)]/2 or 15 judgments by each person. These graphical comparisons are directly translatable to ratios. Software to easily perform such ratings over the web is called Comparion and is available from Expert Choice.
Step 2: Use Saaty’s Eigenvalue method to compute the relative priority ratio scale scores of the six categories. The scores emerging from the method are meaningful ratio scale numerical scores implicitly compared to an absolute zero knowledge value which no article will ever actually reach. Moreover, the logical consistency of the judgments underlying the ratio-scaled scores are tested using the method, and the test results are derived along with the ratio scale scores.
Step 3: As individuals add ratings, multiply each individual’s vector of ratio scale scores by the ratio of 100 to the value of “Greatest Amount of Knowledge possible in a single article,” in each individual’s vector. This will have the effect of mathematically stretching the length of all individual vectors so that all scores are in the interval of 0 to 100.
Step 4: Average the ratio scale values for all categories across all individual ratings. It’s easy to weight the individual ratio scale scores for inconsistency levels, so that the scores reflecting greater inconsistency have less weight in the overall average.
Step 5: Ask individuals who want to rate articles for amount of knowledge to categorize the articles according to the six categories. Transform their categorization to a ratio scale score by mapping to the category to the average category value for the whole population.
Step 6: Average the numerical ratings across all individuals to derive a numerical rating for an article at any point in time.
Step 7: Keep updating the average category ratings and the average article ratings as new data becomes available.
This method will produce a ratio scale score of amount of knowledge in each article posted on the knol site. The ratio scale units created as part of the process can also be called “knols.” However, the choice of unit size is arbitrary, and is due to the adjustment of each vector so that it has a high score of 100, even though the absolute zero and original “Greatest Amount of Knowledge possible in a single article,” ratings are not.
Well that’s it. Knols are not yet units of knowledge, but Google can measure the units of knowledge expressed in them with a methodology like the foregoing.