A comparative analysis of two monumental crises in empirical philosophy: the psychological skepticism of the Scottish Enlightenment versus the linguistic pragmatism of mid-century Oxford and Harvard. 7 mins read.
Two centuries apart, two philosophers shook the foundations of empirical science by pointing out a structural vulnerability in how humans make predictions. Though they are often conflated, David Hume’s 18th-century skepticism and Nelson Goodman’s 20th-century riddle target entirely different structural levels of human knowledge.
Hume looked at the rising sun and noted that our belief that it will rise again tomorrow is not a logical certainty. It is merely a psychological habit born of repetition. Goodman, on the other hand, accepted Hume's psychological habit for the sake of argument, only to point out a deeper, linguistic trap: even if we assume the future will resemble the past, we cannot logically define *how* or in what *way* it will do so.
Hume asks us to justify our faith in the engine of induction. Goodman points out that the engine is steering us in infinite directions simultaneously, and we have no neutral map to guide us.
The Two Skeptical Horizons
To grasp the difference, consider a simple comparison of their mechanics. Hume's challenge is justificatory. He does not doubt what prediction we will make; we will predict the sun will rise and the emerald will remain green. He simply points out that we have no rational, non-circular proof to justify this expectation.
Goodman’s challenge is conceptual and linguistic. He shows that the very same past observations can be used to justify two completely opposite predictions depending on the vocabulary we employ. The problem is no longer that induction lacks a foundation, but that induction itself is radically underdetermined.
| Dimension | Hume's Problem (Old Riddle) | Goodman's Problem (New Riddle) |
|---|---|---|
| Core Question | Why should we assume the future will resemble the past? | Which patterns from the past are we allowed to project into the future? |
| Nature of Challenge | Epistemic & Justificatory (Skepticism of habit) | Linguistic & Semantic (Skepticism of predicates) |
| Proposed Escape | Custom and habit as our natural guide. | Entrenchment: prioritizing culturally and historically tested words. |
The Modern Relevance
This distinction is not merely academic. In the era of data-driven prediction, we have largely bypassed Hume's problem: we accept that our machine learning models must assume the future resembles the past to function. However, we are constantly blindsided by Goodman's problem.
An algorithm trained on historical hiring data might identify "wearing a specific type of watch" or "using certain fonts in a resume" as predictive of success simply because those arbitrary features happened to correlate with good outcomes in the past. The algorithm has constructed its own version of a "grue" predicate—a pattern that holds perfectly in historical data but has zero genuine projectibility.
Referenced Works & Texts
- David Hume, An Enquiry Concerning Human Understanding, Section IV: "Sceptical Doubts Concerning the Operations of the Understanding" (1748). Formulation of the classical problem.
- Nelson Goodman, Fact, Fiction, and Forecast, Chapter III (1954). Contrasts the old and new problems of induction.
If you found this valuable, consider supporting our work.
Join PhiloCrux community.
Unlock high-density masterclasses and investigations into ideas surviving outside the algorithmic consensus. Support independent thought and get full access to our digital library.
Join Now