The paper explores the disjunctive riddle for induction: If we know the sample Ks to be P, we also know that they are P or F (for arbitrary F). Assuming that we also know that the future Ks are non-P, we can conclude that they are F, if only we can inductively infer the evidentially supported P-or-F hypothesis. Yet this is absurd. We cannot predict that future Ks are F based on the knowledge that the samples, and only they, are P. The ensuing challenge is to account for the unprojectibility of the disjunctive hypothesis. I provide an explanation in terms of epistemic dependence. The P-or-F hypothesis is unprojectible because the evidence supporting it depends epistemically on the evidence for the defeated P-hypothesis. The paper also shows that the disjunctive riddle covers the essence of Goodman's infamous grue-problem, which, therefore, can be resolved by the same means: In contrast to the green-hypothesis, the grue-hypothesis is unprojectible because the grue-evidence depends on the evidence for a defeated hypothesis.

中文翻译：

---

分离之谜和格鲁悖论

该论文探讨了归纳的析取谜语：如果我们知道样本 Ks 是 P，我们也知道它们是 P 或 F（对于任意 F）。假设我们也知道未来的 K 是非 P，我们可以得出结论，它们是 F，只要我们能够归纳推断出有证据支持的 P-或-F 假设。然而这是荒谬的。我们无法根据样本（只有它们）是 P 的知识预测未来的 K 是 F。随之而来的挑战是解释析取假设的不可预测性。我根据认知依赖提供了一个解释。P-或-F 假设是不可预测的，因为支持它的证据在认识上取决于被击败的 P 假设的证据。该论文还表明，这个分离的谜语涵盖了古德曼臭名昭著的格鲁问题的本质，因此，