Hi all, When people seek guidance from science, they expect it to work more than once.
In a new article with Cassidy Waldrip and David Lazer, we ask if early fears about corporate control of computational social science have become true. To investigate this question, we analyzed over 200 papers in *Science* , *Nature*, and *PNAS* that involved tech firms and considered whether it's possible to repeat them, comparing studies with corporate collaborations, APIs/Scraping, and papers that relied on Wikipedia. The answer? Few papers that rely on tech firms can be replicated, papers reliant on Wikipedia have *much* more replication feasibility, and where papers based on Wikipedia are not repeatable, it's often because the authors included a commercial dataset. This is one of two new papers I have out this summer on transparent, reliable science that might be of interest to this community. - The blog post summary is here: New Studies: How Commercial Forces Make Science Less Reliable <https://citizensandtech.org/2026/07/commercial-determinants-science/> - Here are the papers: - Matias, J. N., Waldrip, C., & Lazer, D. (2026). Commercial Determinants of Replication Infeasibility in Computational Social Science <https://journalqd.org/article/view/9661>. *Journal of Quantitative Description: Digital Media*, *6*. - McLoughlin, K. L., Palmer, A., Matias, J. N., & Crockett, M. J. (2026). Proprietary LLMs impede scientific transparency and reproducibility <https://www.nature.com/articles/s44159-026-00593-8>. *Nature Reviews Psychology*, 1-2. I hope this is helpful and encouraging to the Wiki research community! -- J. Nathan Matias <http://natematias.com/> : Cornell University : Citizens and Technology Lab <https://citizensandtech.org> : Coalition for Independent Technology Research <http://independenttechresearch.org/> : social.coop/@natematias : blogs <https://natematias.com/external-posts/> : daylight time photos <https://social.coop/tags/outdoormoments> _______________________________________________ Wiki-research-l mailing list -- [email protected] To unsubscribe send an email to [email protected]
