Because of the growth of entropy, we have a very different epistemic access to the past than to the future. In retrodicting the past, we have recourse to “memories” and “records,” which we can take as mostly-reliable indicators of events that actually happened. But when it comes to the future, the best we can do is extrapolate, without nearly the reliability that we have in reconstructing the past.
However — the human brain, as most readers of this blog probably know, was not intelligently designed. It’s doesn’t have the high-level structure of a computer program, where all the processes are carefully planned to achieve some goal. (The lower-level structures share the mechanical features of any other physical system, but that’s of little help here.) Evolution nudges the genome in useful directions, but it can only work with the raw materials it’s given; it doesn’t have the luxury of starting from scratch. So over and over in biological organisms, we find features that were originally developed for one purpose being re-engineered for something else.
As it turns out, the way that the human brain goes about the task of “remembering the past” is actually very similar to how it goes about “imagining the future.” Deep down, these are activities with very different functions and outcomes — predicting the future is a lot less reliable, for one thing. But in both cases, the brain goes through more or less the same routine.