If we look closely, The Primary Problem in Modern Epistemology closely resembles the famous problem of induction.
The problem of induction, first formulated by David Hume in 1739, states that we cannot reasonably infer that the future will resemble the past. That is to say that no matter how many times we see a pattern repeat over time, we cannot say for sure that it will happen again. The sun may not come up tomorrow.
The problem of induction threw the foundations of the philosophy of science into a crisis for over two centuries. If witnessing (seeming) cause/effect relationships doesn’t warrant a definitive conclusion, what is it good for? Of course, science went on evolving and producing value for those who practiced it, but its rational foundation was in disarray. (This is a crude interpretation and lacks nuance, but it’s sufficient for my purposes here.)
The problem of induction is considered to be insurmountable. That is, Hume’s reasoning is impenetrable, and his conclusions are undeniable.
However, in 1972 “answered” the problem of induction sufficiently to put science back on solid philosophical ground. While he accepted the imperviousness of the problem of induction, he sidestepped it, claiming that it doesn’t strike at the heart of scientific endeavor. He reformulated science as the endeavor to determine false theories, thereby strengthening the claim of theories that survive experimentation. (Again, crude and lacking nuance. Please read about the problem of induction and Popper’s response in more detail here, here, and here.)
In effect, this makes science probabilistic. It never gives us 100% confidence in any theory, and instead, it strengthens theories to functionally reliable probabilities by testing them over time.
So, what can this teach us about trust and how to determine reliable sources?
When determining what sources of information are reliable, we face a problem similar to the problem of induction. No matter how many times we find a source to be accurate or reliable, we can’t deduce that it will be correct next time. However, as Popper did with science, we can redirect the goal away from certainty and toward reliable probability.
No source regularly producing content will be correct 100% of the time (and retractions are essential). Though we can’t write off a source for one misplaced claim or conclusion, we can keep track of their record. It’s easy, then, to say that sources with long records with high reliability ought to be trusted while sources with short records or low reliability ought not.
It also suggests that sources without a record—those that have not been time-tested—ought to be taken with a grain of salt.
