KEEPING HUMANS CENTERED IN DIGITAL GEOSCIENCE (Invited Presentation)

In geoscience as in the rest of society, new digital methods are accelerating capabilities to generate, store, manipulate, and share geologic information. New toolsets and scientific approaches support the observation of large systems with unparalleled accuracy (e.g., remote sensing) or analysis of the geological record at scale (e.g., macrostratigraphy). However, uncritically adopted digital methods can bypass nuance and reduce interpretability. Counterintuitively, a world brimming with information may make it harder to synthesize insights. To support the continued generation of geological knowledge, we must ensure that digital geoscience workflows continue to prioritize human curation and use.

Fully using digital tools requires standardizing our representation of geologic information, with huge benefits for scale, discoverability, and reach. Community-level systems like Macrostrat, an openly accessible database that stores harmonized geologic maps and stratigraphic columns, create new opportunities to navigate the world's geologic information. But geology thrives on the integration of heterogeneous information, and standardization risks flattening some of this context. Pursuing distributed data ownership across the geoscience community, rather than centralized platforms, is one way to ensure that observational data is collected in a flexible and inclusive way.

As digital data workflows become dominant, a parallel risk looms that scientists will struggle to wield them to generate insights. While new computational tools enable extremely powerful data transformations, they tend to push data analysis towards unintuitive algorithmic methods. This is particularly true for artificial intelligence (AI) capabilities, which layer statistical processing so generously as to become "black boxes" resistant to external assessment. To advance knowledge, reasoning must be straightforward and replicable; as such, modern geoscience developed in tandem with techniques to convey geologic datasets for human consumption, through venerable graphical products such as maps, columns, and cross-sections. This legacy can be carried forward with human-usable tools to visualize and manipulate digital geologic data. Such software, though time-consuming to build, must be prioritized alongside other computational approaches.