Science_ABM.jl Documentation

An agent-based model funding allocation in science.

Research grants are usually awarded to those projects that seem to have a good chance of success, favoring low-risk projects. Moreover, researchers are evaluated by the number and impact of their publications. At the same time, most journals are reluctant to publish negative results. All together, it is in the best interest of scientists to choose projects that are low-risk and will have results in the short term. Indeed, scientists have become more conservative in choosing their research projects in the past three decades (Rzhetsky2015).

But is this situation in the best interest of humanity? are we losing time and talent in gaining more knowledge? We propose a framework based on Shannon's mathematical theory of communication (ClaudeShannon1948) and the "Inverse Relationship Principle" (IRP) (Beavers2011). IRP states that the less probable or possible a proposition is, the more semantic information it bears. Floridi (Floridi2011) defines semantic information as "well-formed, meaningful, and truthful data", what we consider the source of knowledge. One could use this theory to quantify the amount of semantic information added by research projects to human knowledge. Low-risk projects are by definition more likely to give an expected result, thus being less informative. Risky projects, on the other hand, are more likely to fail, but if they give us a meaningful truth about the universe, they are more informative and scientifically valuable.

This package models scientific research system, i.e. production of science by scientists, publication of research findings, allocating money to the projects, and rewarding the scientists, to estimate the rate of increase in collective knowledge of humanity. Would any change in the current reward and punishment system result in faster scientific progression?

To that end, a useful approach is an agent-based evolutionary algorithm, in which there is a universe with infinite facts to be discovered. There are different probabilities of each fact to be true. Agents are rewarded/punished in different ways for finding facts, for example, by the number of publications, impact of publications, etc. This specifies the risk scientists take in choosing a project to pursue. We can then maximize parameters of the model so that the total information of the system / society is maximized and its cost minimized. If the research questions are general/interdisciplinary, then they take longer to finish.