WHY SCIENCE IS MORE RELIABLE THAN A HUMEAN CRITIQUE OF INDUCTION MIGHT SUGGEST:
I tend to think of it more like this:
1. Observation detects patterns
2. Patterns indicate continuous relationships
3. Continuous relationships provide information about how things work
C. Observations provide information about how things work
1. Meta-observations detect meta-patterns
2. Meta-observations provide meta-information about how things work.
3. Meta-observation meta-information becomes a meta-system of interdependent axioms.
4. If one interdependent axiom is false, the meta-system will fail to fulfill predictions.
5. We can observe the fulfillment of predictions.
C. We can observe the substantiation of a meta-system of axioms.
1. For the purposes of the meta-system, substantiation of axioms is equivalent to "in-scope" truth.
2. It is possible that all meta-system axioms are false in an "out-of-scope" way (simulation theory/demon/brain in vat).
3. Our epistemology is for our "in-scope" purposes, so "out-of-scope" issues are of no concern to our truth conclusions.
CONCLUSION:
P1. The probability of many axioms being false is exponentially less than the probability of one axiom being false.
P2. Science provides us with a multi-layered system that produces substantiation for an ever growing number of axioms.
C. Science is constantly progressing towards "in-scope" ultimate certainty.