If we don't do induction why don't we keep making the same observation continuously? Eg. if I observe Edgar the raven is black at time t1, and I'm trying to falsify "All ravens are black" why should I prefer to observe another raven rather than Edgar again at time t2?
It seems that if I have no induction, Edgar at time t2 seems as severe a test as Arabella at time t2. In which case why not experiment to test "all ravens are black" by continuously testing Edgar?
One answer : I have a wider scale conjecture that ravens are going to stay black. And so Arabella at time t2 is more interesting at t2. Of course, that conjecture itself may be less well corroberated, but is probably consistent with a lot of other beliefs. (ICanBelieveAnythingIWant).
Yes ... ultimately follow all the extra conjectures far enough back, and you find conjectures which aren't explicitly tested or corroborated but are, instead, built on air (or holistically / transcendentally corroborated only by the success of the whole conceptual system)