The Evolution of Software Publication in Astronomy

Talk given at AAS231, special session "Astronomy Software Publishing: Community Roles and Services" (Washington DC). Abstract: Software is a fundamental component of the scientific research process.  As astronomical discoveries increasingly rely on complex numerical calculations and the analysis of big data sets, publishing and documenting software is a fundamental step in ensuring transparency and reproducibility of results. I will briefly discuss the recent history of software publication and highlight the challenges and opportunities ahead.

Transcript

1. Matteo Cantiello 1,2,3,4 1 CCA, Flatiron Institute 2 Princeton University 3 Authorea Chief Scientist The Evolution of Software Publication in Astronomy 4 MESA Council Member
2. “Real astrophysical systems are 3-dimensional objects evolving in time with extremely complex physics. They can only be observed remotely. To interpret the data and model the underlying physics, astrophysicists need to build elaborate models, which are often computationally intensive” Adapted from “Working Papers: Astronomy & Astrophysics Panel Report” (1991) - National Academy Press LSST (30TB/night)
3. Astronomy has become a digital science + + … Examining 166 astronomical articles Allen et al. (2018) found 715 software instances. At least 285 unique codes (but likely many more)
4. Astronomy has become a digital science Virtually every astronomical publication stands on at least a piece of software + + …
5. Many (new) tools now enable easily versioning, sharing and deploying astronomical software. However… …
6. Software is Often Not Shared In a sample of 166 articles “over 40% of the source code for the software used was not available for download” (Allen et al. 2018)
7. Attempts of ASCL to gather the code of published work: 87% refused to share their code (66% do not answer, 20% said they won’t share the code) (Shamir et al. 2013) Software is Often Not Shared
8. The Reproducibility “Paradox” As computational methods have proliferated, research has become less transparent, reproducible, and falsifiable because these methods have not been shared or made available along with the research they enable. (Allen et al. 2018, Baker 2016, Goble et al. 2016, Stodden et al. 2016, Marwick 2015, Shamir et al. 2013, Morin et al. 2012)
9. Status of Links to Source Pepe et al. 2014 Adding external links is not a reliable solution In 2011, 44% of links published a decade earlier (in 2001) were broken
10. Challenges & Opportunities • Is it a matter of incentives? AAS, Chris Lintott Nature, Leslie Sage Springer, Ramon Khanna JOSS, Arfon Smith AstroImageJ, Karen Collins AAS Data, Gus Muench ADS, Alberto Accomazzi ASCL, Alice Allen Why Astronomers do not share their code (e.g. Weiner et al. 2009)
11. Challenges & Opportunities • Is it a matter of incentives? • Is it a matter of formats? Why Astronomers do not share their code AAS, Chris Lintott Nature, Leslie Sage Springer, Ramon Khanna JOSS, Arfon Smith AstroImageJ, Karen Collins AAS Data, Gus Muench ADS, Alberto Accomazzi ASCL, Alice Allen
12. Challenges & Opportunities • Formats ? Why Astronomers do not share their code +400 Years
13. Rethink Scientific Papers as Research Repositories Executable objects, containing the full image of research (narrative, figures, data, code…) https://www.authorea.com/users/23/articles/8762-the-paper-of-the-future E.g. The Paper of The Future
14. Thanks!