David,
Henry Kim
82

There’s a powerful idea embedded in your comment, which I’m constantly drawn back to.

When we construct a theory, we’re explaining a set of data, not the world itself. The data is simply a sample of the world. Therefore, we should always be aware that a theory which accurately explains all of the data might not actually explain the world the data is taken from. Because we are limited to data samples, we can never really know what is, only what could be. If we try to create theories based on what is, we’re introducing a huge blindspot from the very beginning. Instead of creating one theory, we should be creating a set of theories based on what could be—and then sample more data from the world to either eliminate or confirm the theories in our set. And by confirm, I don’t mean prove; I mean increase our confidence that the theory could be accurate.

For example, Newtonian mechanics is accurate enough to explain any experiment we could have thrown at it 200 years ago. It would successfully explain all of the data we had gathered and could gather. But it couldn’t explain the data we were able to collect later in strong gravitational fields. To explain that data, we needed Einstein’s general relativity.

Scientists recently sampled more data, this time from the vicinity of a white dwarf star, which confirms general relativity. General relativity has been very successful at explaining any new data we’ve collected. But alternatives exist, partly because we know general relativity isn’t accurate—the theory doesn’t explain what happens on a quantum scale. And scientists keep collecting more data in edge cases where relativity might not explain things because, if we can collect data which deviates from what relativity predicts, then we’ll have a better idea of how relativity might be wrong and which of the alternatives could be a good starting point for a more likely theory.

But even if we found a theory which unified general relativity and quantum mechanics, and this theory explained all of the data we could collect, it still wouldn’t necessarily describe what is. What if there’s a time factor and all of the experiments we run today would have slightly different results if we run them 100 billion years from now? We can’t know what is, only what could be. Our theories and thinking should reflect that.

One clap, two clap, three clap, forty?

By clapping more or less, you can signal to us which stories really stand out.