Rapid biodiversity studies using BugBox, an AI-powered platform for arthropod identification

Entomologists are being asked to address questions on larger scales than ever before, and we need new technologies to facilitate these larger initiatives. To that end, we have developed a software platform for facilitating large-scale bioinventories of arthropods, which is powered by artificial intelligence (AI) technology for image classification. The platform, called BugBox, uses an AI algorithm to rapidly classify arthropods from specimen photographs, and exports experimental data in an analysis-ready format, complete with biodiversity indices for each field sample. We evaluated the performance of the AI algorithm over three successive model versions, by comparing the AI’s results for five biodiversity experiments with the results of a human taxonomist for the same data. We also used the AI and human data to test the hypothesis that regenerative agricultural practices increase arthropod biodiversity. We show that the AI’s performance improved over successive iterations of the model as it incorporated the human expert’s corrections. We also found agreement between human and AI results in our regenerative agriculture hypothesis tests: we found evidence that regenerative practices increase arthropod diversity in rangeland systems in central North America, whether we used AI- or human-classified data. These results demonstrate that, while the AI cannot operate without the human taxonomist’s guidance, it is still able to provide useful surrogate data very rapidly. In effect, BugBox is efficiently deploying the human’s expertise to a larger bioinventory than the human could otherwise manage. This has profound implications for the scalability of entomological science.