Data and observation is theory-dependent
Y&L argue that we “cling on to the notion that humanity is endowed with the wisdom to intuit new insights by philosophizing, independent of data and observations.” So, where exactly do new ideas and hypotheses come from, if not from data (as Y&L argue)?
Data does not have any qualities—whether important, surprising, or funny—without some kind of hypothesis or theory. There is nothing inherently meaningful or interesting about an apple falling, or any other data point, without a hypothesis or theory. It is only when the apple’s fall is met by a question and hypothesis that it takes on meaning. The very idea of analyzing an apple’s fall—the process of selecting that fall as data to be considered (and relating it to the moon!)—illustrates the central role of hypothesis in scientific discovery.
Now, perhaps Y&L agree that data is meaningless without a theory. After all, they soften their original “a hypothesis is a liability”-argument by allowing that various forms of “background” are important, even in hypothesis-free data exploration and scientific discovery. They refer to the importance of “theoretical background,” “mental background,” and “conceptual background.” This accumulated background, which allows scientists to build on the work of their predecessors, is undoubtedly important. And as Y&L further suggest, the scientific process is recursive, indeed, a conversation between data and theory.
However, this does not mean that scientific discovery and progress are deterministic or inevitable, where “each new question or hypothesis [is] triggered by the analysis of an earlier dataset” (Y&L). This cycle is not automatic. Science is not an all-seeing eye that is observer-independent [1]. It is necessarily punctuated by the human generative capacity to conjecture and hypothesize. It is conjecture, hypothesis and theory—rather than hypothesis-free exploration of data—that allow us to see something in a new way. Without this generative capacity, it is hard to fathom how we could know anything at all. As noted by the philosopher Charles Peirce, “man’s mind has a natural adaptation to imagining correct theories of some kinds…If man had not the gift of a mind adapted to his requirements, he could not have acquired any knowledge” [2].
Hypotheses can of course lead scientists astray, a point Y&L emphasize. We agree. But there is no meaningful, hypothesis-free alternative. The alternative to a bad or blinding hypothesis is a new or better one. Data, empirical findings, and obvious facts can also lead scientists astray. All observation is necessarily hypothesis-laden, no matter how informal these hypotheses might be. There is nothing inherent about data that tells us what to hypothesize. Again, data does not speak for itself. That is our point in focusing on the gorilla experiment, which appears to offer evidence of so-called human blindness. It similarly looks like the sun orbits the earth. But appearances and associated data can be deceiving. Therefore, a hypothesis tells us what data to look for, what experiments to construct, and how to interpret findings.