In the real world of scientific investigation, she said, scientists usually rely on a model-based process rather than a hypothesis-driven one. They formulate models based on what they know from previous research and then derive testable hypotheses from those models. Data from experiments don't validate or invalidate hypotheses as much as they feed back into the models to generate better research questions.
Based on her findings, Nersessian and her colleagues have developed a series of classes for science and engineering students at Georgia Tech that attempts to instill the skills necessary for a model-based approach to investigation. For example, students tackle the thankfully hypothetical scenario of a zombie invasion on campus and how officials could control and contain it. The first thing they must do, Nersessian said, is develop a model that includes relevant information on disease-spreading patterns, quarantine methods, and campus escape routes. Eventually, the students' models begin to resemble real quarantine plans based on real-world diseases. By the end of the classes, Nersessian said, students should be "spontaneous, model-based reasoners."
But it's not just the investigation itself that benefits from philosophy of science training, said Heather Douglas of the University of Waterloo. It's also important for understanding how scientists can achieve and recognize scientific integrity in their research and advocacy.
A major problem with championing integrity in science is that most definitions of integrity are logically circular: Haing scientific integrity means doing science ... with integrity. That's not particularly helpful, Douglas argued.
Instead, philosophy of science can train scientists to recognize the values inherent to doing good science. At minimum, these include internal consistency and empirical validity. But other, more subjective values creep into science as well, including whether theories have explanatory power, how simple models are, whether they fit into larger unifying theories, and whether they implicitly encourage equality and social justice.
These subjective values can be useful for determining what kinds of research questions you should be asking, but they shouldn't be employed when characterizing or interpreting data, Douglas said. Recognizing and separating out those values can help scientists determine the significance of their findings to larger issues and decide with what authority they can advocate on scientific, social, or political issues.