On 23 April, ModelShare team member Sandra Gesing hosted a panel discussion with Christine Kirpatrick from the San Diego Supercomputer Center (SDSC), and Paul Brenner from the University of Notre Dame.
The workshop began with a brief presentation from Sandra, who reminded us of the importance of computational models for researchers to tackle increasingly complex scientific questions that would have been far out of their reach just a few decades ago. The stage was thus set for Paul and Christine to tackle questions about the cyberinfrastructure that enable open modeling practices, and the skills and considerations required to work in multidisciplinary research teams.
Regarding cyberinfrastructure, both Christine and Paul discussed the issue of vendor lock-in – the dependence on legacy systems that are costly to maintain, rigid, and incompatible with other tools. With this framing, Paul introduced the idea of applying design thinking to the cyberinfrastructure made available to researchers. This way, systems are designed in light of their needs and capabilities. Rather than engineers ideating solutions on their own terms, they part from an understanding of their end users. Relatedly, Christine suggested that solutions are also often domain-specific – for example, biology doesn’t require the same infrastructure as economics. Notwithstanding the diversity of solutions – tailored to researchers and/or scientific disciplines – Christine also argued for a base-level of training so that domain experts can better engage with oftentimes complex computational systems.
The intersection of researcher skills and cyberinfrastructure brought about a discussion regarding science gateways and multidisciplinarity. On the one hand, the panelists discussed how science gateways encompass tools and practices that make it easier to employ cyberinfrastructure without the need for technical knowledge of how it works. This is important because research software engineers (RSEs) require years of training on the specifics of that infrastructure. With this, it seems unnecessary for all researchers to have detailed knowledge about the systems they are using.
On the other hand, multidisciplinarity was discussed as an inherent property of scientific modeling. Consider that domain-specific expertise is necessary for data to be accurately analyzed, and models to be faithful to their target systems. On this note, audience members raised questions about the dynamics that emerge between domain experts (e.g.: biologists and economists) and RSEs. To this, Paul responded that the University of Notre Dame’s Center for Research Computings tries to meet domain experts “where they are at.” In other words, tools and practices must be delivered in ways that meet their computational skills and requirements. Christine then welcomed non-computer scientists to pursue their passion into the realm of research computing. As she explained, many of the RSEs at the SDSC were originally trained in other disciplines. With this, the line between domain specialists and RSEs may be blurred, and multidisciplinary collaborations may become easier.
Watch the panel discussion below, and sign up to our Google Group to learn how you can get involved with the ModelShare program!
📸 Image by Alan Warburton / © BBC / Better Images of AI / Plant / Licenced by CC-BY 4.0