NSF Workshop on Discovery Informatics Vasant Honavar

1. NSF Workshop on Discovery Informatics Vasant Honavar Program Director Information & Intelligent Systems (IIS) Division Computer & Information Science & Engineering (CISE) Directorate National Science Foundation vhonavar@nsf.gov

2. A Cartoon of the scientific process Synthesis, Integration Review, Replication, Communication, Publication Hypotheses Model Abduction Background Knowledge Deduction Induction Predictions Data Observation Experimentation

3. Current State of Scientific Discovery • Increasingly sophisticated instruments of observation • Increasing automation of • Data acquisition e.g., using robotic instruments e.g., for measuring gene expression, etc. • Aspects of data analysis e.g., using statistical machine learning to build predictive models from data • Exponential increase in • Databases • Scientific literature • Ontologies and Knowledge Bases • Analysis tools

4. Current State of Scientific Discovery Humans still largely responsible for • Generating and prioritizing questions • Designing and prioritizing experiments • Interpreting results • Forming hypotheses • Drawing conclusions • Replicating studies • Documenting studies • Communicating results • Reviewing results • Integrating results into the larger body of knowledge

5. Ross King’s robot scientist Adam gives a glimpse of some possibilities http://www.wired.com/wiredscience/2011/04/robot-scientist-language/

6. Science 3 April 2009: Vol. 324 no. 5923 pp. 85-89 The Automation of Science Ross D. King, Jem Rowland, Stephen G. Oliver, Michael Young, Wayne Aubrey, Emma Byrne, Maria Liakata, Magdalena Markham, Pinar Pir, Larisa N. Soldatova, Andrew Sparkes, Kenneth E. Whelan and Amanda Clare ABSTRACT The basis of science is the hypothetico-deductive method and the recording of experiments in sufficient detail to enable reproducibility. We report the development of Robot Scientist “Adam,” which advances the automation of both. Adam has autonomously generated functional genomics hypotheses about the yeast Saccharomyces cerevisiae and experimentally tested these hypotheses by using laboratory automation. We have confirmed Adam's conclusions through manual experiments. …

7. Goals of the Workshop • Identify research challenges in Discovery Informatics • Computational modeling of discovery processes • Supporting efficient experimentation and discovery • Enabling collaborative and integrative discovery • Identify opportunities for advancing Discovery Informatics through research in AI, Informatics, and Robotics • Foster a Discovery Informatics Research Community

8. A Cartoon of the scientific process Synthesis, Integration Review, Replication, Communication, Publication Hypotheses Model Abduction Background Knowledge Deduction Induction Predictions Data Observation Experimentation

9. Supporting efficient experimentation and discovery How to facilitate • Identifying and prioritizing questions • Gathering facts and background knowledge e.g., through machine reading • Generating hypotheses from knowledge and data • Designing and prioritizing experiments • Building predictive or causal models from knowledge and data • Communicating models, hypotheses, predictions, and observations • Generating predictions • Testing predictions • Documenting and sharing workflows • Reproducing studies • Collaborating with humans and robots

10. Thanks • To Yolanda Gil and Haym Hirsh for organizing the workshop • To the invited attendees from academia and industry for attending the workshop • To my NSF IIS colleagues for supporting the workshop • Le Gruenwald • Maria Zemankova • Sylvia Spengler • PetrosDrineas • To the program managers from NSF, DARPA, DOE, ONR, IARPA, NIH, NITRD for their strong show of interest