Community Engagement

Christopher Moore

(Trying) to Enhance the Rate of New Scientific Discoveries through Discovery by Data Science. And an introduction to some Big Data Neuroscience Problems.

October 26, 2017
12:00pm
Foxboro Auditorium, Kassar House
Christopher Moore
Professor of Neuroscience, Brown University

Abstract: 

Our laboratory studies neural dynamics, the rapid changes in brain organization that allow processes such as enhanced attention.  We pursue these questions using a variety of imaging and electrophysiological approaches, allowing us to record from up to (in concept) a thousand neurons simultaneously. These data sets are all rich from a Data Science perspective and I will describe specific challenges that need to be met.

I will also discuss our attempt to understand, improve and reward scientific Discovery (thediscoveryengine.org). Many things are rewarded in science—citations, grant funding, patents—but, ironically, many of these are only indirectly correlated with making those findings, or proposing those theories, that change the way we think about a topic, i.e., ‘Discovery.’ A major reason for this disconnect is that Discovery as an entity has never been formally measured, and it is hard to reward what is not quantified.  

Building on a Bayesian definition of Discovery, we are having scientists rate the degree to which a paper changes their view of the topic (formally, ‘Discovery Value’) and the ‘Actionability' of the paper (e.g., its utility as a method, or as a pragmatic advance in medical treatment, etc.). 

Our newly collected data from an intensive NIH trial has shown that while Discovery Value is correlated with more traditional academic evaluation metrics such as citation rate, this new, direct measure of quality also finds a unique population of papers with the highest levels of Discovery Value that is missed by traditional methods.

Beyond these basic (if encouraging) results, this new type of data may provide unique signals that predict scientific progress, and may provide insights into a distinct ‘typology’ of scientific contribution, when Discovery Value, Actionability and other measures are analyzed with more advanced methods.

Bio: 

Dr. Moore is a Professor of Neuroscience at Brown and a co-founder of DiscoveryEngine.