Learning From Grant Applications

There is definite value in mining the rich content of research grant applications for insight into how we can improve funding practices, as Stuart Buck and Christopher Steven Marcum describe in “Let Unfunded Grant Applications See the Light of Day” (Issues, Spring 2025). But even as I am friendly with the authors, I see their rather cavalier rejection of the downsides of open access to all grant applications as myopic.

With the extraordinary advances that have been made in machine learning and artificial intelligence, there is a fantastic opportunity to mine grant applications to address some vexing challenges. Among other things, experimenting with analysis of grant applications could allow us to develop tools to better predict what we will likely know in the future, allow program managers and leadership to set real goals for the US scientific enterprise, take an evidence-based approach to allocating dollars to promising fields of research, move from preferentially funding incremental research to more groundbreaking work, and evaluate the impact of institutions, funding programs, individual researchers, and even particular review boards and program managers. For sure, controlled access should be granted to the corpus of applications for the development and testing of such predictive and evaluative tools.

Agencies, however, have been more than reticent to allow access to grant applications under the guise of protecting privacy even when controlled in a walled garden where access is limited for specific research purposes. The aggressive protectionist approach has impeded access to diverse resources including research data and public records obtained under the Freedom of Information Act. After all, if you can block access to operational information, there is no cause to evolve and scrutiny can be dodged.

But somewhere around Barstow Buck and Marcum concluded that full open access to all research grant applications to public scrutiny would add value to the research enterprise that eclipses the obvious disservice. They are wrong. First, brushing aside the very real danger of having your research ideas and approaches cribbed by others ignores reality, and comparing access to grant applications to preprints and preregistration is perplexingly twisted. The confidential nature of your ideas and approaches is your competitive advantage and fuel for the aggressive competitiveness that drives American science. Enabling such access would absolutely disadvantage smaller labs and tip your hand to larger, better funded labs that could put more resources into a project. Further, it is naïve to think that open applications would not add more fuel to politicalize science by driving accusations that an agency or review panel is biased against a lane of research when they might be just avoiding poor experimental design or funding research into crackpot fringe theories. We should all favor more funding agency transparency, but blind faith in transparency is not governance and open for open’s sake does not consider the larger costs and implications.

Rather, agencies should be allowing liberal access to grant application corpus for the explicit purpose of improving how we fund science, setting real goals for funding agencies, developing the tools that allow us to resolve vexing challenges in research, limiting redundancy, and moving toward allowing scientists to lean in on new frontiers.

Michael Stebbins