The Price of Big Science
Analysis
We ran the system dynamics model to match a systems model of exponential growth to actual scientific knowledge growth . The processes used are base-run simulation , extended simulation , and sensitivity analysis . In the base-run simulation model , we calibrated the simulation model with real data to see how well the assumptions about systemic features simulated within the model fit real data . Based on the results , we conducted several extended simulation analyses to estimate scientific knowledge growth direction in future . This section describes the simulations .
Base-run simulation
We first tested how accurately our system dynamics model calibrates current growth of published articles as shown in Figure 2 . The simulation test output is based on real parameter information revealing a pattern very similar to that found by Jinha ( 2010 ). Simulation outputs of system dynamics as seen in Figure 6 and Table 2 show how simulated outputs are similar to actual data .
Figure 6 shows the similarity between the actual output and the simulation output per year . The dotted line represents our simulated output , and the solid line represents the actual data . When the two lines are compared , the simulation output is almost perfectly aligned with the actual data . Table 3 shows a comparison of actual growth of published articles with growth of the published articles through the simulation numerically one by one as seen in the table . Columns 2 , 5 , 8 , and 11 represent the number of actual published articles per year ; columns 3 , 6 , 9 , and 12 show the outputs drawn from the simulation . The simulated number of published articles in 1991 and in 2009 reached 867,769 and 1,477,664 , respectively . The actual published articles in 1991 and in 2009 are 867,807 and 1,477,383 , respectively . Thus , the simulation model can be shown to have enough explanatory power in tracing the current knowledge growth to validate the model .
Extended simulation
Following validation of the base-run simulation , we simulated future growth of published articles as well as growth based on Price ’ s ( 1963 ) scientific knowledge output model . Price anticipated that scientific articles would reach a turning point to move toward the s-shape of growth after 30 years . To test Price ’ s hypothesis , our simulation model was only extended to 2040 . As a result of our model , the number of expected published articles grew to around 3.770M as seen in Figure 7 . The pattern of growth can remain at the exponential rate under the current parameters .
Sensitivity analysis
We conducted a sensitivity analysis to ensure reliability and predictive power of the model under various environmental uncertainties . In simulation models , sensitivity analysis helps to build confidence by studying the
7
For this aspect of testing , the annual global research output of articles investigated by Jinha ( 2010 ) draws from the pattern of scientific knowledge growth of accumulated articles between 1726 and 2009 . Our simulation tested the time period between 1990 and 2010 because all real parameters between the 1726 and 1990 values could not be obtained . Thus , the base year is 1990 in our simulation model . Note that the information of each parameter is described in the appendix .
119