Charles Fernyhough has an essay in the Guardian about memory and the self. Nothing terribly new if you already have more than a passing familiarity with the scientific literature, but still it's nicely-written:
When we look at how memories are constructed by the brain, the unreliability of memory makes perfect sense. In storyboarding an autobiographical memory, the brain combines fragments of sensory memory with a more abstract knowledge about events, and reassembles them according to the demands of the present. The memory researcher Martin Conway has described how two forces go head to head in remembering. The force of correspondence tries to keep memory true to what actually happened, while the force of coherence ensures that the emerging story fits in with the needs of the self, which often involves portraying the ego in the best possible light….
Memory can lead us astray, but then it is a machine with many moving parts, and consequently many things that can go awry. Perhaps even that is the wrong way of looking at it. The great pioneer of memory research, Daniel Schacter, has argued that, even when it is failing, memory is doing exactly the thing it is supposed to do. And that purpose is as much about looking into the future as it is about looking into the past. There is only a limited evolutionary advantage in being able to reminisce about what happened to you, but there is a huge payoff in being able to use that information to work out what is going to happen next. Similar neural systems seem to underpin past-related and future-related thinking. Memory is endlessly creative, and at one level it functions just as imagination does.
That's how I think we should value memory: as a means for endlessly rewriting the self.