George Blumberg told us about talking with his father, Medicine Nobel laureate Baruch (Barry) Blumberg about the importance of exploratory research in science.
We discussed these statements from Barry Blumberg's autobiography:
“Arriving at a medical diagnosis can be viewed as an example of scientific process. […]
In the induction phase, we first collect data and with them formulate a hypothesis. […]
Pure induction is always modified by some preconception of where the investigation will go. The fact that a decision was made to collect data in a particular place (the beach at Cape Cod, the surface of the moon, the glaciers of Greenland) and within a particular category of nature (plants, mollusks, clouds, quarks, nucleic acids, etc.) indicate the existence of a working hypothesis. […]
In the deductive phase, the hypothesis is stated first, and then experiments are devised and observation made in an attempt to support (“prove”) or reject it.” (Pages 24-25)
“It is a common experience of scientists that unexpected data are often the most interesting because they generate totally new kinds of ideas. Recognizing this we began to organize our study design so as to produce unexpected results. This might seem semantically facetious – if you expect something unexpected, can it really be unexpected? – but it works.” (Page 26)
Blumberg, B.S., 2002. Hepatitis B: The hunt for a killer virus. Princeton University Press.
Helen Berman also described her discussions with Barry about data collection and discovery-based research..
Reading these reflections now, some questions are: Which types of hypotheses and data collection are more productive in AI-based science? What are the roles best played by human scientists and by AI?