New Ways of Doing Science?

download report

Transcript New Ways of Doing Science?

New Approaches and Tools for
Doing Networked Science
David Baker, Sid Banerjee, David
Cooper, Ashish Goel, Elizabeth Lorns,
Sandeep Neema, Andrew Sallans
The Need
• Social networks and Internet communications have
revolutionized many other collaborative tasks
• Spectacular success from early experiments
– Opt-In science: anyone can choose to participate if they have
the right expertise
• Growing number of collaborators in scientific work
• Data sources are growing exponentially
• Automated tools for discovering scientific information (eg.
DeepDive) show early promise.
The time seems ripe for DARPA investment in a program.
Automated Knowledge Assistant
– Exploring or learning a new field
• What are key papers and concepts in a new field?
• Unsolved problems and experts
– Concept Maps and Visualization Tools
– Augmented search, where searching reveals
related concepts, key research results, other
communities which have studied the same
– Using humans to direct the creation of this map
– Identifying research that needs reproduction
Incentives and Mechanisms for Opt-In
A formal understanding of incentives in collaborative research
vs competition, and innovative funding mechanisms
Understanding the nature of rewards: intellectual credit, intellectual
property, funding
Endogenous creation of rewards
Team formation
How to encourage confirmatory as well as exploratory research?
Understanding and designing large-scale crowdsourced
research frameworks — drawing lessons from Fold-It, etc.
Incentivizing diversity/exploration in research
Theoretical models? Connections with bandit problems?
Team formation
Experiments in Fold-It/Topcoder competitions
Platforms and Experimentation
• Existing Examples
– The Science Exchange Network
– Topcoder/FoldIt
– Polymath/K Base
• A platform for collaborative opt-in research among experts
• A platform for opt-in research among the general public
– Eg. Expanding the FoldIt paradigm to drug discovery and neurodegenerative disease
– Collaborative design and visualization
• Experiments with different incentive mechanisms
• Common standards and web APIs for data access and
Validation and integrity of research
– Goal: Improving reproducibility of biological
– How
• Replication studies
• Experiment with separation of experiment creator and
• Characterization of biological protocols in terms of
• Automating Reproducibility
• Tools for capturing workflow
Some Potential Participants
Astronomy community (eg. Chris Lintott – Zooniverse)
David Baker and collaborators (FoldIt/eteRNA)
Michael Bernstein (Stanford -- Crowdsourcing platforms. Eg. Collaborative writing)
Center for Open Science
Distributed Biology Team
Yiling Chen (Harvard – User Generated Content)
Ashish Goel (Stanford – markets and social algorithms)
Arpita Ghosh (Cornell – User Generated Content)
Karim Lakhani (Harvard – Topcoder experiments)
Sandeep Neema (Vanderbilt – collaborative visualization)
Chris Re (DeepDive)
[email protected] home and [email protected] home and [email protected]
The Science Exchange network
Terry Tao (UCLA), Tim Gowers (Polymath)
Luis von Ahn (CMU – Crowdsourcing platforms, eg. Human Computation)