Dr. Konrad Kording – a dually-appointed PIK professor of neuroscience and bioengineering – will be the first speaker in the Penn Libraries’ Fall 2019 Research Tea series.
In this series, the Libraries invites Penn faculty to raise interesting – even provocative – issues at the intersection of faculty research and the role of libraries. Dr. Kording will consider various problems in science research, including image fraud, epistemological failures, tools failures, and replicability.
Read more about Dr. Kording’s work in the Q&A below, register for his September 10 talk here, and stay tuned for additional faculty Research Teas scheduled on September 17, October 15, and December 3.
Your research focus has shifted from computational neuroscience to causality in data. Can you describe your lab’s current scientific approach?
My lab wants to understand how we can obtain datasets, use some limited knowledge about the data domain, and then figure out the relationship between different data elements – how things truly affect one another. For example, let’s say you’re randomly assigned to a doctor who has some preference for a certain treatment. In such cases, we can figure out how well a treatment works. Similar ideas may allow us to determine whether publishing in an expensive “glam” journal actually gets you more readers.
The overarching focus is data that matters. What studying brains and studying cancer have in common is that we have large datasets and we want to understand the underlying logic (and, ideally, causality) based on that data. My trainees acquire a deep understanding of how we can use data to ask questions. I push them to apply those skills wherever they can maximally serve the good, beauty, or the truth.
What is one of the more unusual transdisciplinary partnerships that you’ve formed?
I work with molecular biologists on the properties of DNA polymerases; those polymerases copy the DNA in our cells when they divide. The project promises to allow us to record from all neurons in a brain. And there are always new possibilities for transdisciplinary research. For example, I’m currently curious about whether evolution can help us to understand progress in machine learning.
As you stated, the overarching focus of your lab is “data that matters.” How does this intersect with your work on the automated detection of image plagiarism and your rules for structuring scientific papers?
If papers are fraudulent, they make further research less efficient by leading scientists on a wild goose chase. If papers are not readable, they make the reader waste time (and coffee). As such, doing something about these problems promises to make science a bit more efficient. Science is already sprawling, so making it more efficient matters.
At the research tea, you’ll be discussing “epistemological failures.” Can you describe this concept and explain why it might be important to discuss in the context of a library?
Scientists have always turned to libraries to access the literature, but that literature is created using certain epistemic commitments [on the part of researchers]. While scientists tend to be mum about their commitments, bad commitments lead to misleading literature; in fact, some of my lab’s research suggests that large areas of neuroscience may be fallacious. Libraries should allow us to not just understand the research, but also the underlying set of assumptions. I think that libraries need to develop tools that will allow us to search for epistemic commitments.
How else might libraries support scientific research?
I wish librarians were more active in modern research. For instance, why was Google Scholar not built by librarians? And given that Google Scholar has lots of negative externalities, why is the library community not actively working on better tools to replace it? Why are they letting computer scientists steer the ship? We should be mindful in how we negotiate who sees what information, and librarians can really help with that.