Every scientist knows that science advances only if knowledge is shared.  Mathematically, this statement implies that the advance of science is a function of both the sharing of research results, as well as doing the original research. In principle, therefore, decision makers face the problem of deciding how much to spend on original research and how much to spend on sharing the knowledge that comes out of research.
Consider the accompanying graph with the x-axis being the fraction of research resources expended on spreading knowledge. The scale would range from 0% to 100%. The y-axis is the pace of scientific discovery. One can imagine a curve plotting the pace of discovery as a function of the fraction of resources expended on sharing knowledge.
When the fraction of resources is 0%, the pace of science advance is zero, as nothing is shared. When the fraction of resources is 100%, the pace of advance is also zero, as nothing is spent on the research itself. In between these endpoints, the plot will have a maximum. The plot is the Knowledge Investment Curve.
While we show a conceptualization of the Knowledge Investment Curve, we know very little about the actual form of this curve, or even how much is currently invested in sharing.
Most knowledge sharing activities are not funded directly as budget items. These include writing an estimated one million research papers and reports a year worldwide, as well as finding and reading them. It includes preparing for and participating in conferences, as well as writing and reading emails, blogs, etc. It also includes training postdocs and Ph.D. students, plus an untold number of colleague to colleague conversations.
These myriad activities are centuries old, as old as science itself. What each costs in the aggregate we have little idea. We do know that scientific journals cost several billion dollars a year, because they depend on a central infrastructure that has a visible price. We also know the budgets of organizations whose purpose is to share knowledge such as the Office of Scientific and Technical Information and sister organizations across the government such as Defense Technical Information Center, The National Library of Medicine, The National Agricultural Library and others.
We also know that the Internet, especially the World Wide Web, is changing the nature of the equation, because the unit cost of sharing is so much less than the traditional means. The Web has made sharing global, or at least potentially so. One of the thrusts of the Office of Scientific and Technical Information is to develop Web tools, like World Wide Science.org, to promote global discovery. Like the journals, the Web has a visible price in its portals.
We can ask then what the federal investment should be in Web-based science sharing? Conceptually, points on the Knowledge Investment Curve to the left of the optimum imply that the pace of science discovery would be accelerated by increasing the percentage of funding for sharing results. One thing we know is that the investment in sharing is highly uneven across the various sciences. The fraction of health science research funding dedicated to sharing knowledge is greater than for physical and energy sciences. The latter is unlikely to be near the optimum.