Empirical Problems
Empirical problems are questions about the world that a theory or tradition is expected to address. Does this mechanism produce this outcome? Does this intervention cause this effect? Does this phenomenon occur under specified conditions? Empirical problems have answers that can, in principle, be investigated through observation, measurement, and controlled study. They are the problems most people mean when they talk about scientific progress — the phenomena that demand explanation and the explanations that compete to address them. But Laudan insisted that empirical problems are only half the picture. A tradition that handles its empirical problems brilliantly while failing its conceptual problems is not necessarily progressive, because conceptual incoherence eventually produces empirical failures the tradition cannot accommodate.
In the AI Story
The distinction between empirical and conceptual problems is one of Laudan's most practically consequential contributions. Prior to Laudan, the philosophy of science had tended to treat all problems as empirical — questions about the world that data could settle. This framing made certain disputes intractable, because the disputants shared the data but reached opposite conclusions. Laudan's framework made the intractability legible: the disputants were facing different problem types, and empirical data alone could not resolve conceptual tensions.
Applied to the AI transition, empirical problems are abundant and investigable. What does AI do to productivity? The Berkeley study provides data. What does it do to adoption curves? The adoption curves provide data. What does it do to task seepage, to the colonization of rest by productive engagement, to the erosion of embodied knowledge? These are empirical questions with answers accumulating in the research literature.
But Laudan's framework shows that the AI discourse cannot be settled by empirical evidence alone. The triumphalist and elegist traditions share much of the empirical evidence. They disagree about what it means, because they operate with different conceptual frameworks for interpreting it. More data will not resolve the flow-compulsion problem because the traditions cannot agree on what evidence would count as disconfirming.
This is why the framework's distinction matters. Progress requires addressing both kinds of problems. Traditions that treat all problems as empirical — assuming more data will settle every dispute — systematically overlook the conceptual incoherences that will eventually undermine them. Traditions that treat all problems as conceptual — assuming careful argument alone can settle every question — systematically ignore the evidence that should be modifying their commitments.
Origin
The distinction was developed in Progress and Its Problems (1977) and elaborated in Laudan's subsequent work. It drew on his historical research into actual scientific controversies, where he observed that the most persistent disputes were almost always conceptual rather than empirical — disputes about what the data meant rather than what the data were.
Key Ideas
World-facing questions. Empirical problems ask about phenomena, mechanisms, and outcomes that can be observed.
Data-responsive but underdetermined. Empirical problems can be addressed by evidence, but evidence alone rarely settles disputes between well-developed traditions.
Necessary but insufficient. A tradition must handle its empirical problems, but doing so does not guarantee progress.
Interpretation matters. What counts as evidence for or against a theory depends on the conceptual framework of the tradition doing the evaluating.
Appears in the Orange Pill Cycle
Further reading
- Larry Laudan, Progress and Its Problems (1977), chapter 2.
- Pierre Duhem, The Aim and Structure of Physical Theory (1906) — the classical source for underdetermination.
- W.V.O. Quine, "Two Dogmas of Empiricism" (1951).
- Larry Laudan and Jarrett Leplin, "Empirical Equivalence and Underdetermination" (Journal of Philosophy, 1991).