In 1915, Einstein's general relativity and David Hilbert's competing theory both explained the same gravitational phenomena. Both matched observations perfectly, both predicted Mercury's orbital precession, yet physics chose Einstein's approach. This wasn't arbitrary—it reveals something profound about how science operates beyond simple empirical testing.

When multiple theories explain the same evidence equally well, science faces what philosophers call the underdetermination problem. The resolution involves criteria that go beyond matching observations, drawing on values that might surprise those who view science as purely objective. Understanding these criteria reveals why scientific consensus emerges even when the data alone cannot decide.

Every viable scientific theory must first pass the empirical adequacy test—it must account for observed phenomena and make testable predictions. This seems straightforward: theories that contradict observations get eliminated. But history shows this criterion alone rarely determines a winner. The geocentric model of Ptolemy, with its complex system of epicycles, predicted planetary positions as accurately as Copernicus's heliocentric model for decades.

The Duhem-Quine thesis explains why empirical adequacy proves insufficient. Any failed prediction can be saved by adjusting auxiliary assumptions rather than abandoning the core theory. When Neptune's orbit didn't match predictions, scientists didn't reject Newton's gravity—they hypothesized an undiscovered planet (eventually finding Pluto). Every theory exists within a web of assumptions, making purely empirical falsification nearly impossible.

This underdetermination means that for any body of evidence, multiple theories can always be constructed to fit the data perfectly. Scientists in the 19th century developed both wave and particle theories of light that explained all known optical phenomena. The choice between them required additional criteria beyond empirical matching—criteria that revealed deeper aspects of scientific rationality.

Scientists employ what philosophers call theoretical virtues—non-empirical criteria that guide theory choice. Simplicity (Occam's Razor) favors theories with fewer assumptions or parameters. Einstein's E=mc² triumphed partly because it unified mass and energy in one elegant equation, while competing theories required complex separate treatments. But simplicity isn't always decisive—quantum mechanics is mathematically complex yet scientifically successful.

Scope and fruitfulness often outweigh simplicity. Darwin's evolution explained not just species differences but also embryology, biogeography, and fossil patterns—its broad explanatory scope made it compelling despite initial empirical gaps. Fruitfulness means generating new research programs and unexpected predictions. Mendeleev's periodic table left gaps for undiscovered elements, correctly predicting their properties years before their discovery.

Consistency, both internal and external, acts as another arbiter. Theories must avoid self-contradiction while meshing with established scientific knowledge. When Alfred Wegener proposed continental drift, geologists rejected it partly because it conflicted with physics—no known force could move continents. Only when plate tectonics provided a mechanism consistent with physics did the geological community accept continental movement. These virtues aren't universal laws but pragmatic values that have historically led to scientific progress.

Scientific communities don't just apply theoretical virtues—they negotiate which virtues matter most in specific contexts. Particle physicists prioritize mathematical elegance and symmetry principles, while field biologists value ecological validity over mathematical simplicity. These preferences aren't arbitrary but emerge from what has worked historically in each discipline. The success of symmetry principles in predicting new particles reinforces their value in physics.

Thomas Kuhn revealed how scientific paradigms shape these negotiations. During normal science, communities agree on which problems matter and which solutions count as adequate. But during scientific revolutions, even the standards for theory choice become contested. The shift from classical to quantum mechanics involved not just new theories but new criteria for what counted as a satisfactory explanation—deterministic causation gave way to probabilistic description.

This social dimension doesn't make science arbitrary or merely conventional. Rather, it shows how scientific rationality operates through communal practices that refine and apply evaluative criteria. Peer review, replication attempts, and conference debates constitute a distributed cognitive system for theory evaluation. The convergence on theories like evolution or plate tectonics emerges from thousands of independent assessments using shared but evolving standards.

Science's method for choosing between competing theories reveals a richer rationality than simple empirical testing. The interplay of evidence, theoretical virtues, and communal standards creates a robust yet flexible system for building reliable knowledge. This process explains why scientific consensus emerges even when data alone cannot determine the winner.

Understanding these criteria illuminates not just scientific practice but rational choice in any domain where multiple explanations compete. Whether evaluating historical interpretations, policy proposals, or personal decisions, we can apply these same virtues—seeking explanations that are empirically adequate, theoretically elegant, and socially tested. Science's success comes not from a single method but from collectively refined judgment.