Criteria For Explanatory Power

Criteria for explanatory power are the standards we use to decide which explanation is the "best" among several possibilities. Think of it as a mental scoring system: good explanations are simple, cover all the facts, and fit with what we already know. In the driveway scenario, "It rained" scores highest because it accounts for the wet roof and street (Scope) without needing extra assumptions.

Ask this: "Does this explanation cover all the facts, or just the convenient ones?"

Explanatory power isn't just a feeling; it relies on specific virtues that logicians and scientists use to filter hypotheses. This process is often called "Inference to the Best Explanation" (IBE). The strongest explanations typically maximize simplicity (fewest assumptions), consilience (explains the most diverse evidence), and analogy (fits established explanatory patterns).

Consider a medical diagnosis. If a patient has a cough, fever, and fatigue, the explanation "flu" has high explanatory power because it unifies all three symptoms under one cause. A competing explanation—that the patient simultaneously has allergies (cough), heat exhaustion (fever), and insomnia (fatigue)—is weak because it requires three separate coincidences.

In technology, if a website crashes for everyone, a "server outage" is a powerful explanation because it covers the scope of the problem. Assuming that "every user's computer broke at the same time" fails the simplicity test, even if it is theoretically possible.

While these criteria are powerful tools, they act as heuristics rather than guarantees. Sometimes the truth is complex, weird, or messy. However, starting with the explanation that scores highest on these criteria is the most rational way to investigate the world.

Research into explanatory power focuses on quantifying how we select the "best" hypothesis from a pool of candidates. This field intersects philosophy of science, cognitive psychology, and artificial intelligence.

• Thagard (1978) proposed three primary criteria for theory choice: consilience (how many classes of facts it explains), simplicity (lack of ad hoc auxiliary hypotheses), and analogy (similarity to established explanations) [1].
• Lipton (2004) distinguished between the "likeliest" explanation (highest probability) and the "loveliest" explanation (greatest understanding), arguing that explanatory virtues like mechanism and precision guide us toward the truth [2].
• Lombrozo (2007) demonstrated that humans have a "simplicity heuristic," often preferring simpler explanations even when complex ones are statistically more probable, suggesting a cognitive bias toward parsimony [3].

The "best" explanation is not always the true one. A common pitfall is the "Best of a Bad Lot" problem: if all available explanations are poor, the best one is still likely false. Additionally, our bias for simplicity can lead us to overlook complex systemic causes (like climate change or economic shifts) in favor of simple narratives.